Intro To Logic 20.20
Uploaded by
Dallas VaughnIntro To Logic 20.20
Uploaded by
Dallas VaughnIntro to Logic
Collection Editors:
Ian Barland
Phokion Kolaitis
Moshe Vardi
Matthias Felleisen
John Greiner
Intro to Logic
Collection Editors:
Ian Barland
Phokion Kolaitis
Moshe Vardi
Matthias Felleisen
John Greiner
Authors:
Ian Barland John Greiner
Fuching Chi Phokion Kolaitis
Matthias Felleisen Moshe Vardi
Online:
< http://cnx.org/content/col10154/1.20/ >
CONNEXIONS
Rice University, Houston, Texas
This selection and arrangement of content as a collection is copyrighted by Ian Barland, Phokion Kolaitis, Moshe
Vardi, Matthias Felleisen, John Greiner. It is licensed under the Creative Commons Attribution 1.0 license (http://creativecommons.org/
Collection structure revised: January 29, 2008
PDF generated: October 25, 2012
For copyright and attribution information for the modules contained in this collection, see p. 132.
Table of Contents
1 Introduction
1.1 90 = 100: A Proof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 the need for proofs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 de ning a proof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 Propositional Logic
2.1 A formal vocabulary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Reasoning with truth tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 Reasoning with equivalences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.4 Reasoning with inference rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.5 Exercises for Propositional Logic I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.6 Exercises for Propositional Logic II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3 Relations and Models
3.1 relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.2 properties of relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.3 interpretations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.4 Nonstandard Interpretations (optional) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.5 modeling with relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4 First-Order Logic
4.1 A formal vocabulary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.2 Reasoning with equivalences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.3 Reasoning with inference rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.4 Exercises for First-Order Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5 Conclusion, Acknowledgements
5.1 Logic: Looking Back . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.2 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6 Appendices and Reference Sheets
6.1 propositional equivalences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.2 propositional inference rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.3 rst-order equivalences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.4 rst-order inference rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.5 propositional axioms for WaterWorld . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.6 rst-order axioms for WaterWorld . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.7 browser support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Attributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
iv
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
Chapter 1
Introduction
1.1 90 = 100: A Proof1
Construct a four-sided gure ABED as follows:
◦
• |∠ABE| = 90
◦
• |∠DEB| = 100
• |AB| = |ED|
Using that as a starting point, we now tinker a bit to show that 90=100:
• Draw the perpendicular bisectors to BE and AD; call the point where they meet C.
note: Actually, we must prove that those two perpendicular bisectors really do meet at all (i.e.,
that the point C even exists). In this case, it turns out to be pretty clear it's not hard to argue
that lines AD and BE aren't parallel, and therefore their perpendicular bisectors aren't parallel,
and so they must intersect (in Euclidean geometry). Still, be alert for people making glib assertions
in proofs.
1 This content is available online at <http://cnx.org/content/m12727/1.8/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
1
2 CHAPTER 1. INTRODUCTION
Figure 1.1: A construction to help prove that 90=100
Looking at this gure, some warning ags should be going up: How do we know C lies below BD? Might
it lie above BD? Or exactly on BD? It turns out that the argument below is the same in all of these cases,
though you'll certainly want to verify this to yourself later.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
3
1 |AB| = |ED| By construction.
2 |BC| = |EC| C is on the perpendicular
bisector of BE (thus 4BEC is
isosceles).
3 ∠CBE∼
=∠BEC Base angles of isosceles triangle
BEC are congruent.
4 |∠CBE| = |∠BEC| Congruent angles have equal
measures; line 3.
5 |AC| = |DC| C is on the perpendicular
bisector of AD (thus 4ADC is
isosceles).
6 4ABC∼
=4DEC (!!) Triangles with three congruent
sides are congruent (Euclid's
Side-Side-Side congruence
theorem); lines 1,2,5.
7 (From here, it's just routine
steps to conclude 90=100:)
8 ∠ABC∼
=∠DEC Corresponding parts of
congruent triangles are
congruent; line 6.
9 |∠ABC| = |∠DEC| Congruent angles have equal
measures; line 8.
10 |∠ABC| = |∠ABE| + |∠CBE| By construction.
11 |∠DEC| = |∠DEB| + |∠BEC| By construction.
12 |∠DEC| = |∠DEB| + |∠CBE| Substituting equals with equals;
lines 11 and 4.
13 |∠ABC| = |∠DEB| + |∠CBE| Substituting equals with equals;
lines 12 and 9.
14 |∠ABE| + |∠CBE| = |∠DEB| + Substituting equals with equals;
|∠CBE| lines 13 and 10.
15 |∠ABE| = |∠DEB| Subtracting equals from equals
remains equal.
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
4 CHAPTER 1. INTRODUCTION
16 90 = |∠DEB| By construction, and
substituting equals with equals;
line 15.
17 90 = 100 By construction, and
substituting equals with equals;
line 16.
Table 1.1
A useful corollary: 0=1.
1 90 = 100 Previous theorem.
2 0 = 10 Subtracting equals (90) from
equals remains equal.
3 0 = 1 Dividing equals by non-zero
equals (10) remains equal.
Table 1.2
Exercise 1.1.1 (Solution on p. 15.)
If you feel this result is incorrect, then the challenge for you is to nd the rst line which is false.
You may have noticed that the proof given here has some very minuscule stepse.g. Congruent angles
have equal measure. Usually such simple steps can be omitted, since they are obvious to any reader. We
include them for a few reasons:
• As a careful thinker, you should recognize that such small steps really are part of the complete reasoning,
even if they're not worth mentioning continually.
• If a computer is checking a proof, it needs to actually include those steps.
• Programmers do need to be concerned with distinctions about (abstract) typesthe dierence between
angles and their measures, in this case.
• Sometimes a line's justi cation is glibly given as by construction, when that may not even be correct
!-).
In this course, we'll spend a few weeks working with proofs which do include all the small, pedantic steps,
to instill a mental framework for what a rigorous proof is. But after that, you can relax your proofs to leave
out such low-level steps, once you appreciate that they are being omitted.
1.2 the need for proofs2
1.2.1
The ancient Greeks loved to hang around on the stoa
3 , sip some wine, and debate. But at the end of the day,
they wanted to sit back and decide who had won the argument. When Socrates claims that one statement
follows from another, is it actually so? Shouldn't there be some set of rules to o cially determine when an
argument is correct? Thus began the formal study of logic.
aside: The three fundamental studies were the Trivium grammar (words), logic (reasoning),
and rhetoric (eective communication). These allowed study of the Quadrivium arithmetic (pat-
terns in number), geometry (patterns in space), music (patterns in tone), and astronomy (patterns
in time). All together, these subjects comprise the seven liberal arts
4 .
2 This content is available online at <http://cnx.org/content/m10714/2.36/>.
3 http://education.yahoo.com/reference/dictionary/entry/stoa
4 http://members.aol.com/oldenwilde/members/diu/quadriv.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
5
These issues are of course still with us today. And while it might be di cult to codify real-world arguments
about (say) gun-control laws, programs can be fully formalized, and correctness can be speci ed. We'll look
at three examples where formal proofs are applicable:
• playing a simple game, WaterWorld;
• checking a program for type errors;
• circuit veri cation.
Many other areas of computer science routinely involve proofs, although we won't explore them here. Man-
ufacturing robots rst prove that they can twist and move to where they need to go before doing so, in
order to avoid crashing into what they're building. When programming a collection of client and server
computers, we usually want to prove that the manner in which they communicate guarantees that no clients
are always ignored. Optimizing compilers prove that, within your program, some faster piece of code behaves
the same as and can replace what you wrote. With software systems controlling more and more life-critical
applications, it's important to be able to prove that a program always does what it claims.
1.2.1.1 WaterWorld
Consider a game called WaterWorld, where each location is either empty sea or contains a pirate. When you
enter a location, you must correctly anticipate whether or not it contains pirates.
• If you correctly anticipate open sea, you are able to enter and determine how many of the (up to 3)
adjacent locations contain a pirate.
• If you correctly anticipate a pirate, the location is tagged as dangerous, and you gather no further
information.
Furthermore, there are really two types of moves: guesses, and assertions. If you make an assertion, then
even if you happen to be correct but it is possible you could have been wrong, then it is an error. Also,
it is an error if you make a guess about a location if it is actually possible to assert a location's contents.
The interesting fact about these types of games is that while sometimes guesses are necessary (when?),
surprisingly often an assertion can be made.
(You can freely download WaterWorld
5 .)
5 http://www.teachlogic.org/WaterWorld/download.shtml
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
6 CHAPTER 1. INTRODUCTION
(a) (b)
Figure 1.2: Glimpses of two dierent WaterWorld boards
For instance, in the rst board, what assertions can we be sure of ? What, exactly, is your reasoning?
How about in the second board? You can certainly envision wanting a computer player that can deduce
certain moves, an make those for you automatically.
1.2.1.2 Type Checking
When writing a program, we'd like to simply look at the program and determine whether it has any bugs,
without having to run it. We'll see in the future, however, that such a general problem cannot be solved.
Instead, we focus on nding more limited kinds of errors. Type checking determines whether all functions
are called with the correct type of inputs. E.g., the function + should be called with numbers, not Booleans,
and a function which a programmer has declared to return an integer really should always return an integer.
Consider the following program:
// average:
// Simply divide sum by N, but guard against dividing by 0.
//
real-or-false average( real sum, natNum N ) {
if (N != 0)
return sum / N;
else
return false;
}
One reason programmers are required to declare the intended type of each variable is so that the computer
(the compiler) can prove that certain errors won't occur. How can you or the compiler prove, in the above,
that average returns a real number or false, but never returns (say) a string, and doesn't raise an exception?
Deductions are made based on premises about the types that are passed in, along with axioms about the
input and return types of the built-in functions if, !=, and /, as well as which exceptions those built-ins
might raise.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
7
Consider this variant:
// augment-average:
// Given an old sum and N, compute the average if one more
// datum were included.
//
real augment_average( real old_sum, natNum old_N, real new_datum ) {
return average( old_sum + new_datum, old_N + 1 );
}
Most compilers will rejectaugment-average, claiming that it may actually return false. However, we're
real, by using some knowledge about natural numbers and
able prove that it really will only return a
adding 1, plus some knowledge of what average returns. (Note that our reasoning uses aspects of average's
interface which aren't explicitly stated; most
6 type systems aren't expressive enough to allow more detailed
type contracts, for reasons we'll allude to later.) So we see that many compilers have overly conservative
type-checkers, rejecting code which is perfectly safe, because they are reasoning with only a narrow set of
type-rules.
This example alludes to another use of logic: Not only is it the foundation of writing proofs (ones that
can be created or checked by computers), but logic can also be used as an unambiguous speci cation
language. Observe that while a function's implementation is always speci ed formally and unambiguously
in a programming language the interface is speci ed entirely English, aside from a few type declarations.
Many bugs stem from ambiguities in the English, that dierent humans interpret dierently (or, don't think
about). Being able to use logic to specify an interface (and cannot be modi ed even if the somebody later
tunes the implementation) is an important skill for programmers, even when those logic formulas aren't used
in proofs.
1.2.1.3 Circuit Veri cation
Given a circuit's blueprints, will it work as advertised? In 1994, Intel had to recall ve million of its Pentium
half a billion dollars, lots7 of
processors, due to a bug in the arithmetic circuitry: This cost Intel nearly
8
bad publicity , and it happened after intensive testing. Might it have been possible to have a program
try to prove the chip's correctness or uncover an error, before casting it in silicon?
Software and hardware companies are increasingly turning to the use of automated proofs, rather than
semi-haphazard testing, to verify (parts of ) large products correct. However, it is a formidable task, and
how to do this is also an active area of research.
There are of course many more examples; one topical popular concern is verifying certain security prop-
erties of electronic voting machines (often provided by vendors who keep their source software a proprietary
secret).
Having proofs of correctness is not just comforting; it allows us to save eort (less time testing, and
also able to make better optimizations), and prevent recall of faulty products. But: who decides a proof is
correct the employee with best SAT scores?!? Is there some trusted way to verify proofs, besides careful
inspection by a skilled, yet still error-prone, professional?
Many highly intelligent people are poor thinkers. Many people of average intelligence are skilled thinkers.
The power of the car is separate from the way the car is driven. Edward De Bono, consultant, writer, and
speaker (1933- )
6 http://download.plt-scheme.org/doc/300/html/mzlib/mzlib-Z-H-13.html#node_chap_13
7 http://www.netfunny.com/rhf/jokes/95q1/pentiumd3.html
8 http://www.netfunny.com/rhf/jokes/94q4/pentiumd2.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
8 CHAPTER 1. INTRODUCTION
1.3 de ning a proof9
1.3.1 What are proofs? (informal)
Example 1.1
The following submission from an anonymous engineer to the January, 1902 edition of Popular
Mechanics caught my eye. Seems like something every Boy/Girl Scout and Architect should know.
HOW TO USE THE WATCH AS A COMPASS: Very few people are aware of the fact that
in a watch they are always provided with a compass, with which, when the sun is shining, the
cardinal points can be determined. All one has to do is to point the hour hand to the sun and
south is exactly half way between the hour and the gure 12 on the watch. This may seem strange
to the average reader, but it is easily explained. While the sun is passing over 180 degrees (east to
west) the hour hand of the watch passes over 360 degrees (from 6 o'clock to 6 o'clock). Therefore
the angular movement of the sun in one hour corresponds to the angular movement of the hour
hand in half an hour; hence, if we point the hour hand toward the sun the line from the point
midway between the hour hand and 12 o'clock to the pivot of the hands will point to the south.
Engineer.
They give an argument of correctness; is that really a proof ? Well, there are some ambiguities: Do I hold
the watch vertically, or, in the plane of the sun's arc? Certainly I can't hold it up-side down, even though this
isn't explicitly stated. Furthermore, the correctness of the reasoning relies on some unstated assumptions.
E.g., the sun is at its highest (northernmost) point of its transit at noon. Is this actually true? Does it
depend on the time of year? I'm not exactly sure (and will have to sit down and scratch my head and draw
pictures of orbits, to convince myself ). Certainly there are at least a couple of caveats: even beyond account
for Daylight Savings Time, the solar-time and clock-time only align at time-zone boundaries, and they drift
up to an hour apart, before the next boundary recti es the dierence. Is this presuming I'm in the northern
hemisphere? What if I'm on the equator?
To be fair, the intent of this anecdote was to give enough evidence to convince you, not necessarily to
be a complete, stand-alone self-contained proof. But in writing out a careful proof, one is forced to consider
all the points just made; being forced to understand these can lead you to better understand the procedure
yourself. But be careful to distinguish between something which sounds reasonable, and something that
you're certain of.
1.3.1.1 An argument by form
How can we tell true proofs from false ones? What, exactly, are the rules of a proof ? These are the questions
which will occupy us.
Proofs are argument by form. We'll illustrate this with three parallel examples of a particular proof form
called syllogism.
Example 1.2
1 All people are mortal. Premise
2 Socrates is a person. Premise
continued on next page
9 This content is available online at <http://cnx.org/content/m12072/1.12/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
9
3 Therefore, Socrates is mortal. Syllogism, lines 1,2
Table 1.3
Example 1.3
1 All [substitution ciphers] are Premise
[vulnerable to brute-force
attacks].
2 The [Julius Caesar cipher] is a Premise
[substitution cipher].
3 Therefore, the [Julius Caesar Syllogism, lines 1,2
cipher] is [vulnerable to
brute-force attacks].
Table 1.4
Note that you don't need to know anything about cryptography to know that the conclusion
follows from the two premises. (Are the premises indeed true? That's a dierent question.)
Example 1.4
1 All griznoxes chorble happily. Premise
2 A oober is a type of griznox. Premise
3 Therefore, oobers chorble Syllogism, lines 1,2
happily.
Table 1.5
You don't need to be a world-class oober expert to evaluate this argument, either.
aside: Lewis Carroll, a logician, has developed many whimsical examples
10 of syllogisms and
simple reasoning. (Relatedly, note how the social context of Carroll's examples demonstrates some
feminist issues in teaching logic
11 .)
As you've noticed, the form of the argument is the same in all these. If you are assured that the rst two
premises are true, then, without any true understanding, you (or a computer) can automatically come up
with the conclusion. A syllogism is one example of a inference rule that is, a rule form that a computer
can use to deduce new facts from known ones.
10 http://home.earthlink.net/∼lfdean/carroll/puzzles/logic.html
11 http://www.indiana.edu/∼koertge/rfemlog.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
10 CHAPTER 1. INTRODUCTION
1.3.1.2 Some non-proofs
Of course, not all arguments are valid proofs. Identifying invalid proofs is just as interesting as identifying
valid ones.
note:
Homer: Ah, not a bear in sight. The Bear Patrol must be working.
Lisa: That's specious12
reasoning, Dad.
Homer: Thank you, honey.
Lisa: By your logic, this rock keeps tigers away.
Homer: Oh? How does it work?
Lisa: It doesn't work.
Homer: Uh-huh.
Lisa: It's just a stupid rock.
Homer: Uh-huh.
Lisa: But I don't see any tigers around here, do you?
[pause]
Homer: Lisa, I want to buy your rock!
[A moment's hesitation . . . and money changes hands.]
(From The Simpsons
13 Much Apu About Nothing
14 .)
If Lisa isn't around, who will identify specious reasoning for us? We can certainly use her approach of
nding other particular examples that follow the same argument, yet lead to a clearly erroneous conclusion.
Example 1.5
Suppose that my friend makes the following argument:
1 Warm cola tastes bad. Premise
2 Warm salt-water tastes bad. Premise
3 Therefore, mixing them together Common-sense conclusion,
tastes bad. lines 1,2
Table 1.6
I'm skeptical, so I have a sip; sure enough, the conclusion is indeed true. But is the proof
correct does the common-sense conclusion rule actually hold? In order to refute the form of
the argument, we can try similar arguments which have the same form but a false conclusion (as
Lisa did).
1 Ice-cold coke tastes good. Premise
continued on next page
12 http://education.yahoo.com/reference/dictionary/entry/specious
13 http://snpp.com/
14 http://www.snpp.com/episodeguide/season7.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
11
2 Ice coee tastes good. Premise
3 Therefore, mixing them together Common-sense conclusion,
tastes good. lines 1 and 2.
Table 1.7
After another unfortunate sip, I verify that this conclusion is not true, and therefore my friend's
reasoning is at fault.
My friend responds by claiming that the common-sense conclusion is too valid; the rule is that
bad-taste is preserved upon mixing, not that any taste is preserved. While I'm inclined to believe
that, we realize we can still test this more re ned rule: can you come up with an instance of mixing
together bad-tasting things and ever getting a yummy result? (Say, salt and our, which can be
mixed and baked to get delicious saltines! The argument continues, about whether the form of the
argument precludes baking, and so on.)
The end result (after I take some antacid) is that we have a clearer understanding of the
initially vague common-sense conclusion, and stricter rules about when it applies. Thus, re ning
the argument has led us to a greater understanding.
The above examples are a bit frivolous, but the procedure of looking for counterexamples applies to many
real-world dilemmas. It also highlights the dierence between a correct proof, and a faulty proof that might
still happen to lead to a true result. (By the way, this is the exact same skill used when trying to come up
with an algorithm for a problem: well, the algorithm works for this input, but can I nd a something that
makes one of the steps fail? If so, you then try re ning your algorithm well, I can add a test to take care
of that problem; is that enough so that it always works?)
Exercise 1.3.1 (Solution on p. 15.)
Solve this statement for [X]: It is wrong to ban [X]. Such a ban would punish those reasonable
citizens who would use [X] responsibly, while those who really want to abuse [X] will be able to get
it anyway, through a black market which will only subsidize other criminal activities.
In real-world issues, there are often many subtleties, and short arguments that sound airtight might be
glossing over factors which are important in practice.
Example 1.6
During daylight, there is no need to have headlights (or running lights) on: there's already plenty
of light for everybody to see each other by. Even during the day, headlights slightly increase how
quickly other drivers see you during (say) a routine, tenth-of-a-second glance in their mirror.
Example 1.7
When in a turn-only lane, there is absolutely no need to signal since there's only one way to turn,
a signal can't communicating any information to other drivers! Glib, but not true: Other defensive
drivers presumably know you have only one legal option, but they don't know that you know
that, and they are planning reactions in case you surprise them with a sudden illegal maneuver.
By signaling, you give them information which helps them better plan for yet other contingencies.
Furthermore, it also gives you more con dence that other drivers are expecting your turn, reducing
your suspicion that they're about to pull a surprise maneuver on you. (True, these are all low-
probability events which almost always turn out to be unnecessary. But avoiding accidents is all
about minimizing risks for the one moment events do spiral out of control.)
Example 1.8
You'll lose weight if and only if you burn more calories than you take in. All those diet-plan books
can never get around this, and all their details are pointless.
True, calorie intake and expenditure solely determine weight loss/gain. But after some thought,
we can get examples where the above logic overlooks some relevant dierences: If your friend told
you they were switching from a diet of 2000 calories of balanced short-term and long-term energy
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
12 CHAPTER 1. INTRODUCTION
sources (sugars, proteins, and carbs) to a diet of 2000 calories worth of Pixy Stix at breakfast plus
a Flintstones multivitamin, would you be optimistic that they would have the willpower to strictly
follow the new plan? The two plans are equal when counting calories, but in actuality one really
is a better plan. (Even more exaggeratedly, consider a daily plan of 2000 calories of sugar while
never drinking any watersince water has no calories, it can't aect your calorie count, according
to the above claim.)
These contrived counterexamples help illustrate that it's conceivable that there can be a dier-
ence between diet plans, so the initial claim isn't technically true.
The point illustrated is that often real-world arguments incorrectly imply that their result follows from
the form of the argument, when in fact the form is not valid in the way a syllogism is. This fallacy can
be illuminated by nding a dierent domain in which the argument fails. The practice of searching for
domains which invalidate the argument can help both sides of a debate hone in on bringing the unspoken
assumptions to light. The original argument, if its conclusion is indeed true, must be patched either by
adding the unspoken assumptions or xing the invalid form.
Exercise 1.3.2 (Solution on p. 15.)
Mistakes in syllogisms are hard to make: what are the only two ways to have an error in a syllogism?
1.3.1.3 Other Inference Rules
Of course, there are more ways to deduce things, beyond a syllogism.
• Who decides what the valid inference rules are?
• Is it always clear when people have used the inference rules correctly?
(a) (b)
Figure 1.3: Glimpses of two dierent WaterWorld boards
Consider the following argument about WaterWorld boards:
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
13
1 (A) is next to exactly onepirate. Premise, from either sub gure
2 (A) has only one unexplored Premise, from either sub gure
neighbor.
3 If you are an unexpected Incorrect conclusion
location next to (A), then you
contain apirate.
Table 1.8
This conclusion is not valid; while it is correct for the rst board shown (Figure 1.3(a)), it is incorrect for
the second (Figure 1.3(b)). (I make this mistake all the time when playing WaterWorld too quickly, arrggh!
The Author.)
The problem is that the author of the argument presumably meant to conclude all explored neighbors
of (A) contain a pirate.
Before we can study exact proofs, we need a way of writing exactly what we mean. This
will occupy us for the next section.
1.3.1.4 The need for a precise language
These previous glitches in the WaterWorld arguments both arise, of course, because we were sloppy about
what each sentence meant exactly. We used informal English a ne language for humans, who can cope
with remarkable amounts of ambiguity but not a good language for specifying arguments.
aside: Laws and contracts are really written in a separate language from English legalese
full of technical terms with speci c meanings. This is done because, while some ambiguity is
tolerable in 99% of human interaction, the remaining 1% can be very problematic. Even so, legalese
still contains intentionally ambiguous terms: When, exactly, is a punishment cruel and unusual?
What exactly is the community standard of indecency? The legal system tries to simultaneously
be formal about laws, yet also be exible to allow for unforeseen situations and situation-speci c
latitude. (The result of this tension is the position of Judge.)
aside: Court decisions
15 , while dense reading, are often the model of well-presented arguments.
Consider, from a previous example (Example 1.1), the statement . . .[this is something] every Boy/Girl
Scout and Architect should know. Does this mean all people who are both a scout and architect, or
everybody who is at least one or the other? Genuinely ambiguous, in English! (Often, and/or is used to
mean one or the other or possibly both.)
We'll next look at a way to specify some concepts non-ambiguously, at least for WaterWorld. We need
to be more careful about how we state our facts and how we use these known facts to deduce other facts.
Remember, faulty reasoning might not just mean losing a silly game. Hardware and software bugs can lead
to signi cant bodily harm (Imagine software bugs in an airplane autopilot or surgical robot system), security
loopholes (e.g., in Mozilla
16 or IE17 ), or expensive recalls (p. 7).
One reaction to the above arguments is Well, big deal somebody made a mistake (mis-interpreting or
mis-stating a claim); that's their problem. (And sheesh, they sure are dolts!) But as a programmer, that's
not true: Writing large systems, human programmers will err, no matter how smart or careful or skilled
they are. Type-checkers catch some errors upon compilation, and test suites catch their share of bugs, but
many still remain in real-world software. Thus we are looking for systemic ways to reduce and catch errors,
with the ultimate ideal of being able to prove programs correct.
15 http://caselaw.lp. ndlaw.com/scripts/getcase.pl?court=us&navby=year&year=recent
16 http://www.mozilla.org/projects/security/known-vulnerabilities.html
17 http://www.microsoft.com/technet/security/current.aspx
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
14 CHAPTER 1. INTRODUCTION
aside: Other professions have checklists, protocols, and regulations to minimize human error;
programming is no dierent, except that the industry is still working on exactly what the checklists
or training should be. Someday, a license will be required for practicing software, at least for
software involved with life-safety.
In our study of formal logic, we'll need three things:
• Syntax (language) a precise syntax and vocabulary for expressing concepts without ambiguity,
· Propositional logic,
· First-order logic (propositional logic, plus relations and quanti ers)
• Semantics (meaning) and modeling how to connect these formal languages to whatever topic we
want to reason about (including our software).
• Reasoning (proofs) methods of deducing new facts from old. We'll see three types of reasoning, and
how to use them for each of our two logics:
· Truth tables
· Boolean Algebra
· Inference Rules
We'll visit these topics in an interleaved manner rst propositional logic (immediately with its semantics)
and three methods of reasoning for it; then rst-order logic and an in-depth look at its interpretations, and
nally the methods of reasoning for rst-order logic.
We'll begin with a particular syntax propositional logic for the game of WaterWorld before using
this syntax to formally deduce safe moves.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
15
Solutions to Exercises in Chapter 1
Solution to Exercise 1.1.1 (p. 4)
The aw is extremely hard to nd. We won't actually give the solution, but here's a hint on how to go
about attacking the puzzle:
Note that nding the bug in the proof is the same skill as debugging a program. A good approach is to
try various degenerate inputs. In this case, there are a couple of inputs to the constructionthe length of
◦
CD is arbitrary; no matter how long or short the proof should apply equally well. Similarly, the angle 100
seems arbitrary; ddling with inputs like these (making them very small or very large) might give you some
very careful drawing will clear things up.
clues as to where the bug is. A
Solution to Exercise 1.3.1 (p. 11)
This argument is or has been commonly used for varying topics
• marijuana,
• alcohol,
• all drugs,
• handguns,
• birth control,
• prostitution,
• encryption technology.
The interesting part, is that the traditional Left and Right political positions each use this argument for
some of these items, while rejecting the argument when used for other items.
A more rational response is to either accept all the above, or none of the above, or to realize that
the stated argument wasn't everything that there might be implicit assumptions or arguments which
actually do distinguish between these cases (the dierent interpretations of [X]). Being able to articulate
the dierences is essential. The more re ned arguments may be more nuanced, and less able to t into a
sound-bite, but lead to a better understanding of one's own values. And sometimes, upon re ection, one
may realize that some of the implicit values or premises are things they actually disagree with, once they
are precisely spelled out.
Solution to Exercise 1.3.2 (p. 12)
1. The argument isn't actually in syllogism form. For example, the following is an incorrect syllo-
gism:
1 All people don't know my le's Premise (Equivalent to
password. Nobody knows my le's
password, but reworded to be
of the required form All
somethings have some
property..)
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
16 CHAPTER 1. INTRODUCTION
2 All hackers are people. Premise
3 Therefore, my le is secure Incorrect syllogism, lines 1,2
from hackers.
Table 1.9
To be a syllogism, the conclusion would have to be all hackers don't know my le's password. The
le might or might not be secure, but the above doesn't prove it.
2. One of the two premises is wrong.
1 All people don't know my le's Premise, but possibly false
password.
2 All hackers are people. Premise, but possibly false
3 Therefore, all hackers don't Syllogism, lines 1,2
know my le's password.
Table 1.10
This proof fails, of course, if some hackers are non-people (e.g., programs), or if some people know the
password. (In fact, presumably you know the password!)
Of course, even if a proof fails, the conclusion might be true for other reasons. An incorrect argument doesn't
prove the conclusion's opposite!
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
Chapter 2
Propositional Logic
2.1 A formal vocabulary
2.1.1 propositions1
Recall examples of where we'd like proofs:
• WaterWorld (Is a certain location guaranteed safe?)
• type checking (Does a program call functions in the proper way?)
• circuit veri cation (Does a circuit always work as advertised?)
After seeing the reasons why proofs are important, we ended with a call for rst needing a precise language
for writing down statements without the ambiguity of English.
aside: Might a programming language be a good way to specify formal concepts without ambi-
guity? Programming languages are usually motivated by specifying how to do something (imple-
mentation), rather than formally specifying what is being done (interface). While there is a deep
relation between these two, logic is more appropriate for specifying the what.
2.1.1.1 A formal vocabulary
Imagine an oer where, for a mere $6.99, you can get: EE, (FF or CF or OB or HB) or CC and PH and
BR and GR or WB and PJ. Some ne print clari es for us that BR includes T (Whi, Whe, Ra, or Hb),
FT, HM (Bb, Ba, or Ca), EM, B with CrCh, BB (GR from 6-11am). Unfortunately, it's not clear at all
how the and and ors relate. Fundamentally, is x and y or z meant to be interpreted as (x and y )
or z , or as x and (y or z )? With some context, we might be able to divine what the author intended:
the above oer is the direct translation from the menu of a local diner
2 : 2 eggs, potatoes (french fries,
cottage fries, O'Brien or hashed brown) or cottage cheese and peach half (grits before 11am) and choice of
bread with gravy or whipped butter and premium jam. Bread choices include toast (white, wheat ,raisin or
herb), hot our tortillas, homemade mu n (blueberry, banana or carrot), English mu n, bagel with cream
cheese, homemade buttermilk biscuits. Grits available from 6:00am to 11:00am. (In a brazen display of
understatement, this meal was called Eggs Alone.) Even given context, this oer still isn't necessarily clear
to everybody: can I get both french fries and a peach half ? Happily, coee is available before having to
decipher the menu. In this example, parentheses would have clari ed how we should interpret and, or.
But before we discuss how to connect statements, we will consider the statements themselves.
1 This content is available online at <http://cnx.org/content/m10715/2.58/>.
2 http://www.houseofpies.com/Breakfast.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
17
18 CHAPTER 2. PROPOSITIONAL LOGIC
De nition 2.1: proposition
A statement which can be either true or false.
Example
Your meal will include hashbrowns.
De nition 2.2: propositional variable
A variable that can either be true or false, representing whether a certain proposition is true or
not.
Example
HB
We will often refer to propositional variables as just plain ol' propositions, since our purpose in
studying logic is to abstract away from individual statements and encapsulate them in a single variable,
thereon only studying how to work with the variable.
For a proposition or propositional variable X, rather than write X is true, it is more succinct to simply
write X . Likewise, X is false is indicated as ¬X .
aside: Compare this with Boolean variables in a programming language. Rather than (x ==
true) or (x == false), it's idiomatic to instead write x or !x.
Observe that not all English sentences are propositions, since they aren't true/false issues. Which of the
following do you think might qualify as propositions? If not, how might you phrase similar statements that
are propositions?
• Crocodiles are smaller than Alligators.
• What time is it?
• Pass the salt, please.
• Hopefully, the Rice Owls will win tomorrow's game.
®
• Mr. Burns is lthy rich.
• Fresca is the bee's knees.
2.1.1.1.1 A particular vocabulary for WaterWorld
When playing WaterWorld, what particular propositions are involved? To consider this, we think of a generic
board, and wonder what the underlying statements are. They are statements like location A contains
a pirate ( A − unsafe), location G has 2 adjacent pirates ( G − has − 2) and so on. Each of these
statements may be true or false, depending on the particular board in question.
Here are all the WaterWorld propositions (Section 6.5) that we'll use.
Remember that B − unsafe doesn't mean I'm not sure whether or not B is safe; rather it means B is
unsafe it contains a pirate. You may not be sure whether (the truth of ) this proposition follows what
you see, but in any given board the variable has one of two values, true or false.
Every WaterWorld board has the same set of propositions to describe it: A − unsafe, B − has − 2, etc.
However, dierent boards will have dierent underlying values of those propositions.
2.1.1.1.2 Connectives
− has − 2. Some statements
Some statements in the above proof were simple, e.g., the single proposition A
− unsafe and G − unsafe). We build these more complicated statements
had several parts, though, e.g., (F
out of propositions. If you know both F − unsafe is false, and G − unsafe is false, what can you deduce
about the truth of the statement (F − unsafe and G − unsafe)? Clearly, it is also false. What about when
F − unsafe is false, but G − unsafe is true? What about when both propositions are true? In fact, we can
ll in the following table:
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
19
Truth table for ∧ (AND)
a b (a ∧ b)
false false false
false true false
true false false
true true true
Table 2.1
De nition 2.3: truth table
A truth table for an expression has a column for each of its propositional variables. It has a row
for each dierent true/false combination of its propositional variables. It has one more column for
the expression itself, showing the truth of the entire expression for that row.
Exercise 2.1.1.1 (Solution on p. 64.)
What do you think the truth table for a or b looks like? Hint: To ll out one row of the table,
say, for a = true and b = false, ask yourself For this row, is it true that (a is true, or b is true)?
Exercise 2.1.1.2 (Solution on p. 64.)
The above proof also used subexpressions of the form not b-unsafe. What is the truth table for
not a?
Exercise 2.1.1.3 (Solution on p. 64.)
What is the truth table for the expression (not a) or b?
De nition 2.4: connective
1. The syntactic operator combining one or more logical expressions into a larger expression.
Example
Two operators are ∧ and ∨.
2. A function with one or more Boolean inputs and a Boolean result. I.e., the meaning of a syntactic
operator.
Example
The meaning of ∧ and ∨, e.g., as described by their truth tables.
Example
nand (mnemonic: not and), written ↑, takes in two Boolean values a and b, and returns true
exactly when a ∧ b is not true that is, a ↑ b ≡ ¬ (a ∧ b).
The following are the connectives we will use most often. At least some of these should already be familiar
from Boolean conditional expressions.
Connectives
Connective Pronunciation Meaning Alternative pronun-
ciations / notations
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
20 CHAPTER 2. PROPOSITIONAL LOGIC
¬ not ¬a: a is false -a; !a
∧ and a ∧ ba and b are both a*b; ab; a&&b; a&b
true
∨ or a ∨ b: at least one of a+b; a||b; a|b
{a, b} is true
⇒ implies a ⇒ b: equivalent to a→b; a⊃b; if a then b;
¬a ∨ b a only if b; b if a; b is
necessary for a; a is suf-
cient for b
Table 2.2
Many other connectives can also be de ned. In fact, it turns out that any connective for propositional
logic can be de ned in terms of those above.
Example 2.1
Another connective is if-and-only-if or i, written as a ⇔ b, which is true whena and b have
the same truth value. So, as its name implies, it can be de ned as (a ⇒ b) ∧ (b ⇒ a). It is also
commonly known as a is equivalent to b and a is necessary and su cient for b.
Exercise 2.1.1.4 (Solution on p. 64.)
Another connective is exactly-one-of , which is more traditionally called exclusive-or or xor
(since it excludes both a and b holding, unlike the traditional inclusive or.) How would you de ne
a xor b in terms of the above connectives?
Note that the conventional a ∨ b is sometimes called inclusive-or, to stress that it includes the case where
both a and b hold.
In English, the word or may sometimes mean inclusive-or, and other times mean exclusive-or, depending
on context. Sometimes the term and/or is used to emphasize that the inclusive-or really is intended.
Exercise 2.1.1.5 (Solution on p. 65.)
For each of the following English sentences, does or mean inclusive-or or exclusive-or?
1. Whether you are tired or lazy, caeine is just the drug for you!
2. Whether you win a dollar or lose a dollar, the dierence in your net worth will be noticed.
3. If you own a house or a car, then you have to pay property tax.
4. Give me your lunch money, or you'll never see your precious hoppy taw
3 again!
2.1.2 formulas4
2.1.2.1 Well-Formed Formulas
If we want to develop complicated expressions about breakfast foods like eggs, hashbrowns, and so on, we
will want an exact grammar telling us how to connect the propositions, what connections are allowed, and
when parentheses are necessary (if at all). We will choose a grammar so that all our formulas are fully
parenthesized:
De nition 2.5: Well-Formed formula (WFF)
1. A constant: true or false. (If you prefer brevity, you can write T or F.)
2. A propositional variable.
3 http://www.niftycool.com/hotawhoma.html
4 This content is available online at <http://cnx.org/content/m12073/1.16/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
21
Example
a
3. A negation ¬φ, where φ is a WFF.
Example
¬c
4. A conjunction φ ∧ ψ, where φ and ψ are WFFs.
Example
a ∧ ¬c
5. A disjunction φ ∨ ψ, where φ and ψ are WFFs.
Example
¬c ∨ a ∧ ¬c, or equivalently, (¬c) ∨ (a ∧ ¬c)
6. An implication φ ⇒ ψ, where φ and ψ are WFFs.
Example
¬c ∨ a ∧ ¬c ⇒ b, or equivalently, ((¬c) ∨ (a ∧ ¬c)) ⇒ b
The last two examples illustrate that we can add parentheses to formulas to make the precedence explicit.
While some parentheses may be unnecessary, over-parenthesizing often improves clarity. We introduced the
basic connectives in the order of their precedence: ¬ has the highest precedence, while ⇒ has the lowest.
Furthermore, ∧ and ∨ group left-to-right: a ∧ b ∧ c ≡ (a ∧ b) ∧ c, whereas ⇒ groups right-to-left.
Example 2.2
We can combine these ways of forming WFFs in arbitrarily complex ways, for example,
¬ ((¬a ∧ c ∨ (b ⇒ a ⇒ c)) ∧ ¬ (a ⇒ ¬b))
While large WFFs are common, and we will use some, ones with this much nesting are not.
note: φ, ψ , and θ are meta-variables standing for any WFF. The literal character φ doesn't
actually show up inside some WFF; but instead, any particular formula can be used where we write
φ. It is a variable which you the reader must substitute with some particular WFF, such as
a⇒b . Similarly, a, b, and c are meta-variables to be replaced with a proposition, such as b.
Variations of well-formed formulas occur routinely in writing programs. While dierent languages might
vary in details of what connectives are allowed, how to express them, and whether or not all parentheses are
required, all languages use WFFs.
Example 2.3
When creating the homeworks' web pages, the authors keep the problems and solutions together in
one le. Then, a program reads that le, and creates a new one which either excludes the solution
(for the problem set), or includes it (for the solution set, and for practice-problems). The condition
for deciding whether to include the solutions is a WFF.
;; is-a-solution?: paragraph -> boolean
;; A function to tell if we are looking at a "solution" paragraph.
;; Assume this is provided.
;; is-in-a-practice-prob?: paragraph -> boolean
;; A function to tell if Is the current problem a practice problem?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
22 CHAPTER 2. PROPOSITIONAL LOGIC
;; Assume this is provided.
;; include-all-solutions?: boolean
;; A variable for the entire file.
;; Assume this is provided.
;; show-or-hide-soln: paragraph -> paragraph
;; Either return the given paragraph,
;; or (if it shouldn't be revealed) return a string saying so.
;;
(define (show-or-hide-soln a-para)
(if (and (is-a-solution? a-para)
(not (or include-all-solns? (is-in-a-practice-prob? a-para)))
"(see solution set)"
a-para))
Note that the Boolean variable include-all-solutions? and Boolean values of (is-a-solution?
a-para) and (is-in-a-practice-prob? a-para) play the part of propositions (is − soln,
include − solns, is − practice), respectively. The if's condition boils down to the WFF is − soln ∧
¬ (include − solns ∨ is − practice).
Keep in mind that a WFF is purely a syntactic entity. We'll introduce rules later for re-writing or reasoning
with WFFs, but it's those rules that will be contrived to preserve our meaning of connectives like ∧ or ¬.
The truth value of a WFF depends on the truth values we assign to the propositions it involves.
When writing a program about WFFs, verifying syntactic property, calculating a value, counting the
number of negations or bs, etc., such programs exactly follow the de nition of WFF given.
2.1.2.2 Some formulas are truer than others
Is the formula A − unsafe ∨ A − has − 2 true? Your response should be that it depends on the partic-
ular board in question. But some formulas are true regardless of the board. For instance, A − unsafe ∨
¬A − unsafe: this holds no matter what. Similarly, A − unsafe ∧ ¬A − unsafe can never be satis ed (made
true), no matter how you try to set the variable A − unsafe.
De nition 2.6: truth assignment
An assignment of a value true or false to each proposition being used.
Example
For the formula a ⇒ a ∧ b, one possible truth assignment is a = true and b = false. With that
truth assignment, the formula is false.
aside: We've used three dierent symbols to describe equality in some sense:
• a ⇔ b is a formula. The symbol ⇔ is a logical connective.
• φ ≡ ψ is a statement that two formulas are equivalent − − − that is, the same for all truth
assignments.
• a = true de nes the value of a proposition. We also use the symbol for de ning variables,
b = ψ , and meta-variables, φ = ψ .
Of these, only ⇔ occurs within a formula.
Commonly, people use symbols such as ≡ for multiple purposes. This is problematic when part
of what we are studying are the syntactic formulas themselves.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
23
De nition 2.7: tautology
A WFF which is true under any truth assignment (any way of assigning true/false to the proposi-
tions).
Example
A − unsafe ⇒ A − unsafe
Example
a⇒a ∨ b
De nition 2.8: unsatis able
A WFF which is false under any truth assignment.
Example
¬ (A − unsafe ⇒ A − unsafe)
Example
a ⇒ ¬a
Note that in algebra, there are certainly formulas which are true (or similarly, false) for all values, but
they don't get special names. For example, over the real numbers, any assignment to x makes the formula
x2 ≥ 0 true, so it's similar to a tautology. Similarly, x = x+1 is unsatis able, since it can't be made true
for any assignment to x.
Some people use the term contingency to mean formulas in between: things which can be either true or
false, depending on the truth assignment. Really, tautologies and unsatis able formulas are boring. However,
trying to determine whether or not a formula is a tautology (or, unsatis able) is of interest. That's what
proofs are all about!
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
24 CHAPTER 2. PROPOSITIONAL LOGIC
Figure 2.1: Kinds of formulas: tautologies, contingencies, unsatis ables
Identify the following Yogi Berra quotes either as tautologies, unsatis able, or neither. (Take these
exercises with a grain of salt, since the English statements are open to some interpretation.)
Exercise 2.1.2.1 (Solution on p. 65.)
Pitching always beats batting −−− and vice-versa.
Exercise 2.1.2.2 (Solution on p. 65.)
You can observe a lot just by watchin'.
Exercise 2.1.2.3 (Solution on p. 65.)
Nobody goes there anymore. . . it's too crowded.
Exercise 2.1.2.4 (Solution on p. 65.)
It sure gets late early out here.
Exercise 2.1.2.5 (Solution on p. 65.)
Always go to other people's funerals; otherwise they won't come to yours.
2.1.2.3 Finding Truth
Now that we've seen how to express concepts as precise formulas; we would like to reason with them. By
reason, we mean some automated way of ascertaining or verifying statements −−− some procedure
that can be carried out on an unthinking computer that can only push around symbols. In particular, for
propositional logic, we'll restrict our attention to some (closely related) problems:
• TAUTOLOGY: given a formula φ, is it a tautology?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
25
• SATis ability: Give a formula φ, is it satis able? (Is there some truth assignment to its variables,
that makes it true?)
• EQUIV: Given two WFFs φ and ψ , are they equivalent? (Do they give the same result for all possible
truth assignments to their variables?
2.1.2.4 Game-speci c rules
Is x ∨ y ∨ z a tautology? Clearly not. Setting the three propositions each to false, the formula is false. But
now consider: Is A − has − 0 ∨ A − has − 1 ∨ A − has − 2 a tautology? The answer here is yes of course,
... well, as long we're interpreting those propositions to refer to a WaterWorld board. We'll capture this
notion by listing a bunch of domain axioms for WaterWorld: formulas which are true for all WaterWorld
boards.
There are a myriad of domain axioms which express the rules of WaterWorld. Here are a few of them:
• A − has − 0 ⇒ B − safe ∧ G − safe
• A − has − 2 ⇒ B − unsafe ∧ G − unsafe
• ...
A more complete list is here (Section 6.5.2: The domain axioms). Whenever we deal with WaterWorld, we
implicitly take all these domain axioms as given.
2.2 Reasoning with truth tables
2.2.1 using truth tables5
Seeing how we can express some concepts as some formulas, and how some formulas are tautologies while
others might be true or false depending on the truth assignment, we come to a question: how can we
determine when a formula is a tautology? How can we tell if two dierent formulas are equivalent for all
truth assignments? We'll look at three dierent methods of answering these questions:
• reasoning with truth tables (this section),
• reasoning with equivalences (Section 2.3.1), and
• reasoning with inference rules (Section 2.4.1).
2.2.1.1 Using Truth Tables
Is ⇒ associative? In other words, is a ⇒ (b ⇒ c) always equivalent to a ⇒ b ⇒ c? What is a methodical
way of answering questions of this type? We can make a truth table (De nition: "truth table", p. 19) with
two output columns, one for each formula in question, and then just check whether those two columns are
the same.
Exercise 2.2.1.1 (Solution on p. 65.)
Use truth tables to show that a ⇒ (b ⇒ c) and (a ⇒ b) ⇒ c aren't equivalent.
Thus we see that truth tables are a method for answering questions of the form Is formula φ equivalent to
formula ψ ? We make a truth table, with a column for each of φ and ψ, and just inspect whether the two
columns always agree. A bit of a brute-force solution, but certainly correct.
What about the related question, Is formula θ a tautology?. Well, obviously truth tables can handle
this as well: make a truth table for the formula, and inspect whether all entries are true. For example, in the
above problem (Exercise 2.2.1.1), we could have made a truth table for the single formula a ⇒ (b ⇒ c) ⇔
(a ⇒ b) ⇒ c. The original question of equivalence becomes, is this new formula a tautology?
5 This content is available online at <http://cnx.org/content/m10716/2.35/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
26 CHAPTER 2. PROPOSITIONAL LOGIC
The rst approach is probably a tad easier to do by hand, though clearly the two approaches are equiva-
lent. Another handy trick is to have three output columns you're computing: one for φ = a ⇒ (b ⇒ c), one
for ψ = (a ⇒ b) ⇒ c, and one for φ ⇔ ψ; lling out the rst two helper columns makes it easier to ll out
the last column.
tip: When making a truth table for a large complicated WFF by hand, it's helpful to make
columns for sub-WFFs; as you ll in a row, you can use the results of one column to help you
calculate the entry for a later column.
Exercise 2.2.1.2 (Solution on p. 66.)
Is it valid to replace the conditional
int do_something(int value1, int value2)
{
bool a = . . .;
bool b = . . .;
if (a && b)
return value1;
else if (a || b)
return value2;
else
return value1;
}
with this conditional?
int do_something(int value1, int value2)
{
bool a = . . .;
bool b = . . .;
if ((a && !b) || (!a && b))
return value2;
else
return value1;
}
After all, the latter seems easier to understand, since it has only two cases, instead of three.
So, how would do we use truth tables to reason about WaterWorld? Suppose you wanted to show that
G − safe was true on some particular board. Clearly a truth table with the single column G − safe alone
isn't enough (it would have only two rows false and true and just sit there and stare at you). We
need some way to incorporate both the rules of WaterWorld (Section 6.5) and the parts of the board that
we could see.
We can do that by starting with a huge formula that was the conjunction of all the WaterWorld domain
axioms; call it ρ. We would encode the board's observed state with another formula,ψ . Using these, we can
create the (rather unwieldy) formula that we're interested in: ρ ∧ ψ ⇒ G − safe. (Notice how this formula
eectively ignores all the rows of the the truth-table that don't satisfy the rules ρ, and the rows that don't
correspond to the board we see ψ : because of the semantics of ⇒, whenever ρ ∧ ψ is false, the overall
formula ρ ∧ ψ ⇒ G − safe is true.)
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
27
2.2.2 the limitations of truth tables6
2.2.2.1 Are we done yet?
Are we done with propositional logic, now that we can test for equivalence and tautologies, using truth tables?
Possibly. Truth tables can answer any question about propositional logic, but not always conveniently.
Consider the following code:
bool do_something(int value)
{
bool a = . . .;
bool b = . . .;
if (a && !b)
return true;
else if (!a && !b)
return false;
else if (a)
return a;
else if (b)
return false;
else
return true;
}
Clearly, this is very ugly and should be simpli ed. But to what? We could build a truth table for the
corresponding WFF, but so far we don't have any better way of nding a simpler equivalent formula than
testing equivalence with whatever comes to mind. We need some way to generate formulas, given either an
equivalent formula or a truth table.
There is another practical di culty with truth table: they can get unwieldy.
Exercise 2.2.2.1 (Solution on p. 66.)
How many rows are there in a truth table with 2 input variables? 3 variables? 5 variables? 10
variables? n variables?
Exercise 2.2.2.2 (Solution on p. 66.)
(Optional) Now, how many such boolean functions are possible, with 2 inputs? With 3? For
fun, sit down and name all the possible two-input functions. You'll nd that some of them are
rather boring, such as the constant function true, and many are just permutations on ⇒.
When discussing a circuit with 100 wires (each corresponding to a single proposition), truth tables are
clearly infeasible
7 . Modern processors have millions of wires and transistors. It is still an area of active
research to cope with such a huge number of possibilities. (The key idea is to break things down into small
sections, prove things about the small sections, and hopefully have a small set of sentences formally capturing
the interface between sections.)
So truth tables are intractable for analyzing circuits of more than a few wires. But will they su ce for
answering WaterWorld questions? Image a (large) table with all the neighbor propositions: A − has − 0,
B − has − 0, . . ., A − has − 1, B − has − 1, . . . Now, determine which rows which entail B − safe. To answer
this, we end up looking at rows involving many clearly-irrelevant propositions such as Z − has − 2.
aside: Hmm, considering every possible board and then counting what proportion of boards
entail B − safe hmm, this is the brute-force de nition of probability! Since such truth tables
enumerates all possible boards, it's like looking for probability 1 the brute-force way.
6 This content is available online at <http://cnx.org/content/m12074/1.7/>.
7 http://www.cs.rice.edu/∼ian/Misc/exponential-is-big.shtml
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
28 CHAPTER 2. PROPOSITIONAL LOGIC
Also, this method of playing WaterWorld via huge truth tables would be unsatisfying for another reason:
it doesn't actually re ect our own reasoning. As a general principle of programming, your program should
always re ect how you conceive of the problem. The same applies to logic.
aside: Consider the dierence between using truth tables and actually reasoning. The philosopher
Bertrand Russell
8 , trying to pin down what exactly constitutes knowledge, suggested that he
knows that the last name of Britain's prime minister begins with a 'B'. While Gordon Brown is
prime minister, making Bertrand is correct, we hesitate to say he actually knows the fact he
wrote his example when the prime minister was Arthur Balfour
9 (1902-1905). So while he is correct
in a truth-table sense, his reasoning isn't, and we tend to say that he does not actually know the
prime minister's last initial.
So, no: we're not yet nished with propositional logic. We want to look for (hopefully) more feasible ways
to determine whether a formula is a tautology (or, whether two formulas are equivalent). As a clue, we'll try
to discover methods which are based on the way we naively approach this. We'll look rst at equivalences
(Section 2.3.1), and then at inference rules (Section 2.4.1).
2.3 Reasoning with equivalences
2.3.1 propositional equivalences10
2.3.1.1 Propositional Equivalences
What are the roots of x3 − 4x? Well, in high-school algebra you learned how to deal with such numeric
formulas:
x3 − 4x
= x x2 − 4 factor out x
= x (x − 2) (x + 2) The identity a2 − b2 = (a + b) (a − b) with a being x, and b being 2.
Table 2.3
This last expression happens to be useful since it is in a form which lets us read o the roots 0, +2, -2.
dierent formulas are all equivalent. In fact, our very de nition
The rules of algebra tell us that these three
of two formulas being equivalent is that for any value of x the two formulas return the same value. We
are distinguishing between syntax (the expression itself, as data), and semantics (what the expression
means). Usually, when presented with syntax, one is supposed to bypass that and focus on its meaning
(e.g., reading a textbook). However, in logic and post-modern literature alike, we are actually studying the
interplay between syntax and semantics. The general gist is that in each step, you rewrite subparts of your
formula according to certain rules (replacing equals with equals).
Well, we can use a similar set of rules about rewriting formulas with equivalent ones, to answer the
questions of whether two formulas are equal, or whether a formula is a tautology. George Boole
11 was the
rst to realize that true and false are just values in the way that numbers are, and he rst codi ed the rules
for manipulating them; thus Boolean algebra is named in his honor.
aside: The term algebra
12 comes from the values true, false and operators ∧, ∨ having some
very speci c properties similar to those of numbers with ×, +.
8 http://www.humanities.mcmaster.ca/∼bertrand/
9 http://www.britannia.com/gov/primes/prime39.html
10 This content is available online at <http://cnx.org/content/m10717/2.40/>.
11 http://kerryr.net/pioneers/boole.htm
12 http://planetmath.org/encyclopedia/Algebra.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
29
Figure 2.2: George Boole (1815-1864)
Again, each individual step consists of rewriting a formula according to certain rules. So, just what are
the rules for manipulating Boolean values? We'll start with an example.
Example 2.4
1 a ∧ false ∨ b ∧ true
2 ≡ false ∨ b ∧ true Dominance of false over ∧
3 ≡ b ∧ true ∨ false Commutativity of ∨
4 ≡ b ∧ true Identity element for ∨ is false
5 ≡b Identity element for ∧ is true
Table 2.4
Thus we have a series of equivalent formulas, with each step justi ed by citing a propositional equivalence
(Section 6.1). By and large, the equivalences are rather mundane. A couple are surprisingly handy; take a
moment to consider DeMorgan's laws.
¬ (φ ∧ ψ) ≡ ¬φ ∨ ¬ψ ¬ (φ ∨ ψ) ≡ ¬φ ∧ ¬ψ
Table 2.5
(Try φ being Leprechauns are green, and ψ being Morgana Le Fay likes gold. Do these laws make
sense, for each of the four possible truth assignments?) Augustus DeMorgan
13 was also an important gure
in the formalization of logic.
13 http://www-gap.dcs.st-and.ac.uk/∼history/Mathematicians/De_Morgan.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
30 CHAPTER 2. PROPOSITIONAL LOGIC
Figure 2.3: Augustus DeMorgan (1806-1871)
Here is another example. For a statement φ ⇒ ψ, the contrapositive of that formula is ¬ψ ⇒ ¬φ. We
can show that a formula is equivalent to its contrapositive:
Example 2.5
Contrapositive
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
31
1 φ⇒ψ
2 ≡ ¬φ ∨ ψ De nition of ⇒
3 ≡ ψ ∨ ¬φ Commutativity of ∨
4 ≡ ¬¬ψ ∨ ¬φ Double Complementation
5 ≡ ¬ψ ⇒ ¬φ De nition of ⇒
Table 2.6
Don't confuse the contrapositive of a statement with the converse of a formula: The converse of φ⇒ψ is
the formula ψ ⇒ φ; in general a formula is not equivalent to its converse!
This next example is actually a proof of one of the laws from the given list, using (only) others from the
list.
Example 2.6
Absorption of ∨
1 φ ∧ ψ ∨ ψ
2 ≡ φ ∧ ψ ∨ ψ ∧ true Identity of ∧
3 ≡ ψ ∧ φ ∨ ψ ∧ true Commutativity of ∨
4 ≡ ψ ∧ (φ ∨ true) Distributivity of ∧ over ∨
5 ≡ ψ ∧ true Dominance of ∨
6 ≡ψ Identity of ∧
Table 2.7
Exercise 2.3.1.1 (Solution on p. 66.)
Show that the Absorption of ∧ equivalence holds, given the other equivalences. I.e., show
(a ∨ b) ∧ b ≡ b.
Compared to proofs using truth tables, Boolean algebra gives us much shorter proofs. But, determining
which equivalence to use in the next step of a proof can be di cult. In this case, compare the solution for
this exercise to the previous absorption proof. These two proofs have a special dual relationship described
in the next section.
Exercise 2.3.1.2 (Solution on p. 66.)
Show that the modus ponens rule, a ∧ (a ⇒ b) ⇒ b always holds. I.e., show that it is a tautology,
and thus equivalent to true.
So, what would it mean to use Boolean algebra as reasoning for WaterWorld? That is, if you wanted to
show that G − safe was true, how would you do that using Boolean algebra? As with truth-tables, we would
take the conjunction of all the WaterWorld domain axioms (call it ρ), and the board's observed state (ψ ).
We would then want to show that asserting G − safe was already equivalent to the rules-and-observed-state:
ρ ∧ ψ ≡ ρ ∧ ψ ∧ G − safe.
2.3.1.1.1 Duals (optional)
Duals: a symmetry between ∧, ∨ mediated by ¬.
Looking at the provided propositional equivalences (Section 6.1), you should notice a strong similarity
between those for ∨ and those for ∧. Take any equivalence, swap ∨s and ∧s, swap trues and falses, and
you'll have another equivalence! For instance, there are two avors of DeMorgan's law, which are just duals
of each other:
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
32 CHAPTER 2. PROPOSITIONAL LOGIC
¬ (φ ∧ ψ) ≡ ¬φ ∨ ¬ψ ¬ (φ ∨ ψ) ≡ ¬φ ∧ ¬ψ
Table 2.8
aside: In terms of circuit diagrams, we can change each AND gate to an OR gate and add
negation-bubbles to each gate's inputs and outputs. The principle of duality asserts that this
operation yields an equivalent circuit.
The idea of duality is more general
14 than this. For example, polyhedra have a natural dual15 of inter-
changing the role of vertices and faces.
2.3.2 normal forms16
2.3.2.1 CNF, DNF, . . . (ENufF already!)
In high school algebra, you saw that while x3 − 4x and x (x − 2) (x + 2) are equivalent, the second form is
particularly useful in letting you quickly know the roots of the equation. Similarly, in Boolean algebra there
are certain canonical normal forms which have nice properties.
A formula in Conjunctive Normal Form, or CNF, is the conjunction of CNF clauses. Each clause
is a formula of a simple form: a disjunction of possibly-negated propositions.
Example 2.7
c ⇒ a ∧ b is equivalent to (a ∨ ¬c) ∧ (b ∨ ¬c). This latter formula is in CNF, since it is
the conjunction of disjunctions, and each disjunction consists only of propositions and negated
propositions.
Example 2.8
The conjunctions and disjunctions need not be binary. The following formula is also is CNF.
¬a ∧ (a ∨ b ∨ ¬c) ∧ (b ∨ ¬d ∨ e ∨ f )
Note that its rst clause is just one negated proposition. It is still appropriate to think of this
as a disjunction, since φ ≡ φ ∨ φ.
Another format, Disjunctive Normal Form, or DNF is the dual of conjunctive normal form. A DNF
formula is the disjunction of DNF clauses, each a conjunction of possibly-negated propositions.
Example 2.9
a ∧ b⇒c is equivalent to ¬a ∨ ¬b ∨ c which is in DNF: three disjunctions, each being a clause
with only one term. (It also happens to be in CNF a single clause with three terms!) It is
also equivalent to the more eshed out DNF formula where we insist that each clause include all
except a ∧ b ∧ ¬c:
three variables. We end up with a formula that includes each possible clause
(a ∧ b ∧ c) ∨ (a ∧ ¬b ∧ c) ∨ (a ∧ ¬b ∧ ¬c) ∨ (¬a ∧ b ∧ c) ∨
That is, the formula
(¬a ∧ b ∧ ¬c) ∨ (¬a ∧ ¬b ∧ c) ∨ (¬a ∧ ¬b ∧ ¬c).
aside: Electrical Engineering courses, coming from more of a circuit perspective, sometimes call
CNF product-of-sums, and call DNF sum-of-products, based on ∨,∧ being analogous to +,*.
Any Boolean function can be represented in CNF and in DNF. One way to obtain CNF and DNF formulas
is based upon the truth table for the function.
14 http://carbon.cudenver.edu/∼hgreenbe/glossary/duals.html
15 http://www.georgehart.com/virtual-polyhedra/duality.html
16 This content is available online at <http://cnx.org/content/m12075/1.12/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
33
• A DNF formula results from looking at a truth table, and focusing on the rows where the function is
true: As if saying I'm in this row, or in this row, or . . .: For each row where the function is true,
form a conjunction of the propositions. (E.g., for the row where a is false, and b is true, form ¬a ∧ b.)
Now, form the disjunction of all those conjunctions.
• A CNF formula is the pessimistic approach, focusing on the rows where the function is false: I'm not
in this row, and not in this row, and . . .. For each row where the function is false, create a formula
for not in this row: (E.g., if in this row a is false and b is true form ¬ (¬a ∧ b); then notice that by
DeMorgan's law, this is a ∨ ¬b a disjunct. Now, form the conjunction of all those disjunctions.
Example 2.10
Truth table example
a b c Unknown function
false false false false
false false true false
false true false true
false true true true
true false false false
true false true true
true true false false
true true true false
Table 2.9
For CNF, the false rows give us the following ve clauses:
• a ∨ b ∨ c
• a ∨ b ∨ ¬c
• ¬a ∨ b ∨ c
• ¬a ∨ ¬b ∨ c
• ¬a ∨ ¬b ∨ ¬c
and the full formula is the conjunction of these. Essentially, each clause rules out one row as being
true.
For DNF, the true rows give us the following three clauses:
• ¬a ∧ b ∧ ¬c
• ¬a ∧ b ∧ c
• a ∧ ¬b ∧ c
and the full formula is the disjunction of these. Essentially, each clause allows one row to be true.
This shows that, for any arbitrarily complicated WFF, we can nd an equivalent WFF in CNF or DNF.
These provide us with two very regular and relatively uncomplicated forms to use.
Exercise 2.3.2.1 (Solution on p. 67.)
The above example (Example 2.10) produced CNF and DNF formulas for a Boolean function, but
they are not the simplest such formulas. For fun, can you nd simpler ones?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
34 CHAPTER 2. PROPOSITIONAL LOGIC
2.3.2.1.1 Notation for DNF, CNF
Sometimes you'll see the form of CNF and DNF expressed in a notation with subscripts.
• DNF is ∨i ψi , where each clause φi is ∧j λj , where each λ is a propositional variable (Prop), or a
negation of one (¬Prop).
• CNF is ∧i ψi , where each clause ψi is ∨j λj , where each λ is again a propositional variable (Prop), or a
negation of one (¬Prop).
For example, in the CNF formula (a ∨ b) ∧ (¬a ∨ b ∨ c) ∧ (¬a ∨ ¬b) we have φ2 = ¬a ∨ b ∨ c
within that clause we have λ1 = ¬a.
One question this notation brings up:
• What is the disjunction of a single clause? Well, it's reasonable to say that ψ ≡ ψ. Note that this is
also equivalent to ψ ∨ false.
• What is the disjunction of zero clauses? Well, if we start with ψ ≡ ψ ∨ false and remove the ψ, that
leaves us with false! Alternately, imagine writing a function which takes a list of booleans, and returns
the ∨ of all of them the natural base case for this recursive list-processing program turns out to be
false. Indeed, this is the accepted de nition of the empty disjunction. It follows from false being the
identity element for ∨. Correspondingly, a conjunction of zero clauses is true.
Actually, that subscript notation above isn't quite correct: it forces each clause to be the same length,
which isn't actually required for CNF or DNF. For fun, you can think about how to patch it up. (Hint:
double-subscripting.)
Note that often one of these forms might be more concise than the other. Here are two equivalently verbose
ways of encodingtrue, in CNF and DNF respectively: (a ∨ ¬a) ∧ (b ∨ ¬b) ∧ . . . ∧ (z ∨ ¬z) is equivalent
to (a ∧ b ∧ c ∧ . . . ∧ y ∧ z) ∨ (a ∧ b ∧ c ∧ . . . ∧ y ∧ ¬z) ∨ (a ∧ b ∧ c ∧ . . . ∧ ¬y ∧ z) ∨
. . . ∨ (¬a ∧ ¬b ∧ . . . ∧ ¬y ∧ ¬z). The rst version corresponds to enumerating the choices for each
location of a WaterWorld board; it has 26 two-variable clauses. This may seem like a lot, but compare it to
the second version, which corresponds to enumerating all possible WaterWorld boards explicitly: it has all
possible 26-variable clauses; there are 226 ≈ 64 billion of them!
2.3.3 soundness and completeness17
2.3.3.1 Are we done yet?
We have shown procedures, using both truth tables and equivalences, for solving two dierent logic problems:
• Equivalence: Show whether or not two WFFs φ and ψ are equivalent (the same under any truth
assignment);
• Tautology: Show whether or not a given WFF φ is a tautology (true under all truth assignments).
Exercise 2.3.3.1 (Solution on p. 67.)
Which of these two logic problems seems harder than the other? That is, suppose you have a
friend who can solve any Equivalence problem e ciently. But you want to open a business which
will solve any Tautology problem e ciently. Can you open your business and, by subcontracting
out speci c Equivalence problems to your friend, really solve any Tautology problem brought to
you? This question is sometimes phrased as Does Tautology reduce to Equivalence? Or, does
it work the other way: does Equivalence reduce to Tautology?
But we have a more fundamental question to ask, about the method of using Boolean algebra (propositional
equivalences) to prove something: Where does the initial list of allowable equivalences come from, and how
do we know they're valid? The answer is easy −−− each equivalence can be veri ed by a truth table!
17 This content is available online at <http://cnx.org/content/m12076/1.10/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
35
Exercise 2.3.3.2 (Solution on p. 67.)
Using a truth table, show the validity of conjunctive Redundancy: φ ∧ (¬φ ∨ ψ) ≡ φ ∧ ψ
This is called soundness of Boolean algebra: If, using our propositional equivalence rules, we derive that
φ and ψ are equivalent, then truly they are equivalent. (Whew!)
By the way, there is one subtle point: our truth table tells us that a ∧ b and b ∧ a are equivalent.
But then suddenly we generalize this to saying that for any formulas φ andψ , φ ∧ ψ and ψ ∧ φ are also
equivalent. What lets us justify that step? It's because any given formula will be either true or false, so we
can reduce the entire formula to a single true/false proposition.
Is Boolean algebra enough? Does our list of allowable propositional equivalences include everything you'll
need? That is, could I have asked as a homework problem to show some two formulas equivalent (using
Boolean algebra), and even though they really are equivalent, there aren't enough rules to on our list to let
you nish the homework? Hmm, good question! The property we desire here is called the completeness
of Boolean algebra: any equivalence which is true can be proved.
It turns out that, given any two formulas which really are equivalent, Boolean algebra is indeed su ciently
powerful to show that. Put both formulas into CNF (or, DNF); if the truth tables are equal then the
CNF formulas will be equal. (Well, there are a few details to take care of: you have to order the clauses
alphabetically, eliminate any duplicate clauses, and include all variables in each clause. This might be
tedious, but not di cult.) Thus, Boolean algebra is complete, since (we state without proof ) this procedure
can always be carried out.
The concepts of soundness and completeness can be generalized to any system.
De nition 2.9: soundness
If the system (claims to) prove something is true, it really is true.
De nition 2.10: completeness
If something really is true, the system is capable of proving it.
2.4 Reasoning with inference rules
2.4.1 propositional inference rules18
2.4.1.1 Inference
Truth tables and equivalences are useful and powerful tools, but they do not correspond to how we usually
reason about things. What we will do now is look at more familiar reasoning and how to formalize that.
For example, with Boolean algebra it is awkward to prove that a ∧ b implies a. For that, it is necessary
to reword the problem in terms of equivalences, as a ∧ b ⇒ a ≡ true. Our next tool provides a more
straightforward way to reason about implications.
Example 2.11
Given the following piece of a WaterWorld board, how would you conclude that G is unsafe?
18 This content is available online at <http://cnx.org/content/m10718/2.35/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
36 CHAPTER 2. PROPOSITIONAL LOGIC
Figure 2.4: A glimpse of a WaterWorld board
Since H − has − 2, at least two of H 's three neighbors must be unsafe. But, since we know that
one of these, J , isn't unsafe, then the two others, including G, must both be unsafe. Let's write
this out more explicitly:
1 H − has − 2would imply one of WaterWorld domain axiom, i.e.,
the following is true: de nition of H − has − 2
(P − unsafe and G − unsafe), or
(J − unsafe and P − unsafe), or
(G − unsafe and J − unsafe).
2 H − has − 2 is true. Premise (by inspection of this
particular board)
3 One of the following is true: lines 1,2
(P − unsafe and G − unsafe), or
(J − unsafe and P − unsafe), or
(G − unsafe and J − unsafe).
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
37
4 not J − unsafe Premise (by inspection)
5 (P − unsafe and G − unsafe) lines 3,4
6 G − unsafe line 5
Table 2.10
Whew! A lot of small steps are involved in even this small deduction. It's apparent we'd want to automate
this as much as possible! Let's look at some other short examples, which we'll formalize in a moment.
Exercise 2.4.1.1 (Solution on p. 68.)
How do you know that A − has − 2 proves B − unsafe ?
Exercise 2.4.1.2 (Solution on p. 68.)
Similarly, how do you reason that A − has − 1 and G − safe prove B − unsafe?
2.4.1.1.1 Formal inference rules and proofs
In the above examples, we relied on common sense to know what new true formulas could be derived from
previous ones. Unfortunately, common sense is imprecise and sometimes wrong. So, we need to formalize
how we form proofs.
We now de ne a formal proof of θ from the premises φ, . . . , ψ , written
φ, . . . , ψ ` θ (2.1)
(A proof with no premises simply means there is nothing on the left of the turnstile: ` θ.) For example,
we'll show shortly that H − has − 2 ` G − unsafe. A proof consists of a sequence of WFFs, each with a
justi cation for its truth. We will describe four permissible justi cations for each step:
• A premise.
• An axiom.
• An inference rule.
• A subproof.
aside: O cially we might want to annotate the turnstile with ww, to mean proves within the
WaterWorld inference system, indicating our use of the WaterWorld domain axioms. If you're
proving things about other domains, you'd use dierent domain axioms.
Example 2.12
We can formalize the above examples to show each of the following:
• H − has − 2 ` G − unsafe
• A − has − 2 ` B − unsafe
• A − has − 1, G − safe ` B − unsafe
See below for formal proofs of some of these.
Stating an axiom, a simple assumed truth, is a rather trivial, boring way of coming up with a true formula.
Some axioms are domain axioms: they pertain only to the domain you are considering, such as WaterWorld.
In our case, we don't have any axioms that aren't domain axioms. If our domain were arithmetic, our axioms
would describe how multiplication distributes over addition, etc.
Just using axioms is not enough, however. The interesting part is to deduce new true formulas from
axioms and the results of previous deductions.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
38 CHAPTER 2. PROPOSITIONAL LOGIC
note: The point of philosophy is to start with something so simple as not to seem worth stating,
and to end with something so paradoxical that no one will believe it. Bertrand Russell
19 , The
Philosophy of Logical Atomism
An inference rule formalizes what steps are allowed in proofs. We'll use this list of valid inference rules
(Section 6.2) as our de nition, but, this is just one set of possible inference rules, and other people could use
slightly dierent ones.
First, let's look at some simple examples, using the simpler inference rules.
Example 2.13
We'll formalize a previous exercise (Exercise 2.4.1.1) to show A − has − 2 ` B − unsafe.
1 A − has − 2 Premise
2 A − has − 2 ⇒ WaterWorld domain axiom
B − unsafe ∧ F − unsafe
3 B − unsafe ∧ F − unsafe ⇒Elim, lines 1,2, where
φ = A − has − 2, and
ψ = B − unsafe ∧ F − unsafe
4 B − unsafe ∧Elim (left), line 3, where
φ = B − unsafe, and
ψ = F − unsafe
Table 2.11
What we mean in line 3, for example, is that we are using the domain axiom ⇒Elim. That
states that if we know φ ⇒ ψ , and we know φ, then we can conclude ψ. In line 3, we have de ned
φ = A − has − 2 and ψ = B − unsafe ∧ F − unsafe, so that φ⇒ψ corresponds to the conclusion
of line 2 and φ corresponds to that of line 1. Thus, this domain axiom applies, and we get the
conclusion ψ .
That's almost exactly like the steps we took in the previous informal proof, but now we're a bit
pickier about our justi cations for each step.
Formally, when using a domain axiom, the justi cation is a combination of the name of that inference rule,
the line numbers of which previous WFFs are being used, and a description of how those WFFs are used in
that inference rule in this particular step. Later, we'll often omit the description of exactly how the speci c
inference rule is used, since in many cases, that information is painfully obvious.
Example 2.14
In this system, commutativity of ∧ and ∨ are not among the inference rules. However, they do
follow. For example, consider the following proof of A ∧ B ` B ∧ A.
1 A ∧ B Premise
2 A ∧Elim (left), line 1, where φ=A
3 B ∧Elim (right), line 1, where
ψ=B
continued on next page
19 http://www.humanities.mcmaster.ca/∼bertrand/
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
39
4 B ∧ A ∧Intro, lines 3,2, where φ = B,
and ψ = A
Table 2.12
Does this example (Example 2.14) also show that C ∧ D ` D ∧ C ? Well, yes and no. That proof does
not have anything to do with propositions C
D. But, clearly, we could create another nearly identical
and
proof for C ∧ D ` D ∧ C , by substituting C and D for A and B , respectively. What about proving the
other direction of commutativity: B ∧ A ` A ∧ B ? Once again, the proof has exactly the same form,
but substituting B and A for A and B , respectively. Stating such similar proofs over and over is technically
necessary, but not very interesting. Instead, when the proof depends solely on the form of the formula and
not on any axioms, we'll use meta-variables to generalize.
Example 2.15
Generalized ∧ commutativity: χ ∧ υ`υ ∧ χ
1 χ ∧ υ Premise
2 υ ∧Elim (left), line 1, where φ=χ
3 χ ∧Elim (right), line 1, where
ψ=υ
4 υ ∧ χ ∧Intro, lines 3,2, where φ = υ,
and ψ = χ
Table 2.13
Exercise 2.4.1.3 (Solution on p. 68.)
Similarly, associativity of ∧ and ∨ are not among the inference rules. This is a particularly
important detail, since our WaterWorld domain axioms frequently use formulas of the form a ∧
b ∧ c, which isn't technically legal according to our de nition of WFFs. What we'd like to show
is that χ ∧ υ ∧ ω`χ ∧ υ ∧ ω and χ ∧ υ ∧ ω`χ ∧ υ ∧ ω as well as the equivalent for ∨.
Thus, when we see three, four, or more terms in a conjunction (or disjunction), we can legitimately
group them as we see t.
These deductions are straightforward and should be unsurprising, but perhaps not too interesting. These
simple rules can carry us far and will be used commonly in other examples.
Example 2.16
The case-elimination rule is easy enough for a dog! Rico
20 has a vocabulary of over 200 words, and if
asked to fetch an unknown toy, he can pick it out of a group of known toys by process-of-elimination.
(It's almost enough to make you wonder whether dogs know calculus
21 .)
20 http://news.bbc.co.uk/2/hi/science/nature/3794079.stm
21 http://www.maa.org/features/elvisdog.pdf
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
40 CHAPTER 2. PROPOSITIONAL LOGIC
Figure 2.5: This Border Collie knows his inference rules.
There is a subtle dierence between implication (⇒) and provability (`). Both embody the idea that
the truth of the right-hand-side follows from the left-hand-side. But, ⇒ is a syntactic formula connective
combining two WFFs into a larger WFF, while ` combines a list of propositions and a WFF into a statement
about provability.
Exercise 2.4.1.4 (Solution on p. 69.)
Show that, φ`ψ is equivalent to `φ⇒ψ in that, we can show one if and only if we can show
the other.
2.4.2 using subproofs22
2.4.2.1 Subproofs
The reductio ad absurdum (RAA), Latin for reduction to absurdity, seems very strange: If we can prove
that false is true, then we can prove the negation of our premise. Huh!?! What on Earth does it mean to
prove that false is true?
This is known as proof-by-contradiction. We start by making a single unproven assumption. We then
try to prove that false is true. Clearly, that it nonsense, so we must have done something wrong. Assuming
we didn't make any mistakes in the individual inference steps, then the only thing that could be wrong is
the assumption. It must not hold. Therefore, we have just proven its negation.
This form of reasoning is often expressed via contrapositive. Consider the slogan
If you paid list price, you didn't buy it at SuperMegaMart.
(This is a contrapositive, because the real statement the advertisers want to make is that if you buy it at
SuperMegaMart, then you won't pay list price.), which we'll abbreviate payFull ⇒ ¬boughtAtSMM. You
know this slogan is true, and you just made a SuperMegaMart purchase (boughtAtSMM), and are suddenly
wanting a proof that you got a good deal. Well, suppose we didn't. That is, suppose payFull. Then by
the truth of the marketing slogan, we infer ¬boughtAtSMM. But this contradicts boughtAtSMM (that is,
from ¬boughtAtSMM and boughtAtSMM together we can prove that false is true). The problem must have
been our pessimistic assumption payFull; clearly that couldn't have been true, and we're happy to know that
¬payFull.
22 This content is available online at <http://cnx.org/content/m12077/1.12/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
41
Example 2.17
Spot the proof-by-contradiction used in The Simpsons:
Bart, ling through the school records: Hey, look at this: Skinner makes $25,000 per year!
Other kids: Ooooh!
Milhouse: And he's 40 years old; that makes him a millionaire!
Skinner, indignantly: I wasn't a principal when I was 1!
Milhouse: And, he paints houses during the summer ... he's a billionaire!
Skinner: If I were a billionaire, would I still be living with my mother? [Kids' laughter]
Skinner, to himself: The kids just aren't responding to logic anymore!
In the particular set of inference rules we have chosen to use, RAA is surprisingly important. It is the only
way to prove formulas that begin with a single ¬.
23
Example 2.18
We'll prove ` ¬ (α ∧ ¬α).
1 subproof:α ∧ ¬α ` false
1.a α ∧ ¬α Premise for subproof
1.b α ∧Elim (left), line 1.a,
whereφ = α, and
ψ = ¬α
1.c ¬α ∧Elim (right), line 1.a,
whereφ = α, and
ψ = ¬α
1.d false falseIntro, lines 1.b,1.c,
where φ = α
2 ¬ (α ∧ ¬α) RAA, line 1, where
φ = α ∧ ¬α
Table 2.14
Exercise 2.4.2.1 (Solution on p. 69.)
Here's another relatively simple example which uses RAA. Show that the modus tollens rule
holds: α ⇒ β , ¬β ` ¬α
Another use of subproofs is to organize proofs' presentations. Many proofs naturally break down into larger
subparts, each with its own intermediate conclusion. These steps between these subparts are big enough to
correspond to our intuition, but too big to correspond to individual inference rules. This gives additional
useful structure to a proof, aiding our understanding.
Example 2.19
Previously, we showed that ∧ (AND) commutes (Example 2.14). However, that conclusion is only
directly applicable when the ∧ is at the top-level, i.e., not nested inside some other connective.
Here, we'll show that ∧ commutes inside ¬, or more formally, ¬ (α ∧ β) ` ¬ (β ∧ α).
warning: When doing inference-style proofs, we will not use the Boolean algebra laws nor replace
subformulas with equivalent formulas. Conversely, when doing algebraic proofs, don't use inference
rules! While theoretically it's acceptable to mix the two methods, for homeworks we want to make
sure you can do the problems using either method alone, so keep the two approaches separate!
23 This is an example of reasoning about our logic system. It shows us that while we might have some redundant inference
rules, RAA isn't one of them. The only other rule which produces formulas starting with an initial ¬ is ¬Intro. Is this also
essential, or could we still prove all the same things even without ¬Intro?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
42 CHAPTER 2. PROPOSITIONAL LOGIC
We'll do two proofs of this to illustrate that there's always more than one way to prove something!
In our rst proof, we'll use RAA. Why? Looking at our desired conclusion, what could be the
last inference rule used in the proof to reach the conclusion? By the shape of the formula, the
last step can't use any of the introduction inference rules (∧Intro, ∨Intro, ⇒Intro, falseIntro, or
¬Intro). We could potentially use any of the elimination inference rules. But, for ∧Elim, ∨Elim,
⇒Elim, ¬Elim, or CaseElim, we would rst have to prove some more complicated formula to obtain
our desired conclusion. That seems somewhat unlikely or unnecessary. For falseElim, we'd have to
rst prove false, i.e., obtain a contradiction, but our only premise isn't self-contradictory. The only
remaining option is RAA.
1 ¬ (α ∧ β) Premise
2 subproof:β ∧ α ` false
2.a β ∧ α Premise for subproof
2.b α ∧ β Theorem: ∧ commutes
(Example 2.14), line 2a
2.c false falseIntro, lines 1,2.b
3 ¬ (α ∧ β) RAA, line 2
Table 2.15
The proof above uses a subproof because it is necessary for the use of RAA. In contrast, the
proof below uses two subproofs simply for organization.
For our second proof, let's not use RAA directly. Our plan is as follows:
• Assume the premise ¬ (α ∧ β).
• Again, use commutativity to show that β ∧ α⇒α ∧ β
• Use modus tollens (Exercise 2.4.2.1) to obtain the conclusion.
We can organize the proof into corresponding subparts:
1 ¬ (α ∧ β) Premise
2 subproof:β ∧ α⇒α ∧ β
2.a β ∧ α`α ∧ β Theorem statement: ∧
commutes
(Example 2.14)
2.b β ∧ α⇒α ∧ β ⇒Intro, line 2.a
3 subproof:¬ (β ∧ α)
3.a β ∧ α⇒ Theorem statement:
α ∧ β , ¬ (α ∧ β) ` modus tollens
¬ (β ∧ α) (Exercise 2.4.2.1)
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
43
3.b (β ∧ α ⇒ α ∧ β) ∧ ⇒Intro, line 3.a
¬ (α ∧ β) ⇒
¬ (β ∧ α)
3.c (β ∧ α ⇒ α ∧ β) ∧ ∧Intro, lines 2,1
¬ (α ∧ β)
3.d ¬ (β ∧ α) ⇒Elim, lines 3.b,3.c
Table 2.16
2.4.2.2 More examples
Now let's use these rules in a couple larger proofs, to show some more interesting results.
Example 2.20
Let's redo the rst example (Example 2.11)'s proof formally and show H − has − 2 ∧ J − safe `
G − unsafe. The inference rules we used informally above don't correspond exactly to those in our
de nition, so the formal proof is more complicated.
1 H − has − 2 ⇒ P − unsafe ∧ G − unsafe ∨ WaterWorld axiom,
J − unsafe ∧ P − unsafe ∨ G − unsafe ∧ choosing a grouping of
J − unsafe the ternary ∨, as
justi ed by ∨
commutativity
(Example 2.14)
2 H − has − 2 ∧ J − safe Premise
3 H − has − 2 ∧Elim (left), line 2
4 P − unsafe ∧ G − unsafe ∨ J − unsafe ∧ ⇒Elim, lines 1,3
P − unsafe ∨ G − unsafe ∧ J − unsafe
5 J − safe ∧Elim (right), line 2
6 J − safe ⇒ ¬J − unsafe WaterWorld axiom
7 ¬J − unsafe ⇒Elim, lines 5,6
8 subproof:G − unsafe ∧ J − unsafe ` false
8.a G − unsafe ∧ Premise for subproof
J − unsafe
8.b J − unsafe ∧Elim (right), line 8.a
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
44 CHAPTER 2. PROPOSITIONAL LOGIC
8.c false falseIntro, lines 7,8.b
9 ¬ (G − unsafe ∧ J − unsafe) RAA, line 8
10 P − unsafe ∧ G − unsafe ∨ J − unsafe ∧ CaseElim (right), lines
P − unsafe 4,9
11 subproof:J − unsafe ∧ P − unsafe ` false
11.a J − unsafe ∧ P − unsafe Premise for subproof
11.b J − unsafe ∧Elim (left), line 11.a
11.c false falseIntro, lines 7,11.b
12 ¬ (J − unsafe ∧ P − unsafe) RAA, line 11
13 P − unsafe ∧ G − unsafe CaseElim (right), lines
10,12
14 G − unsafe ∧Elim (right), line 13
Table 2.17
Wow! This formalization is a lot longer than the original informal proof. That's a result of the particular set
of inference rules we are using, that we can only make inferences in small steps. Also, here we were pickier
about the distinction between not safe and unsafe.
Example 2.21
The previous example (Example 2.20) is a perfect candidate for adding structure to the proof by
using additional subproofs. The following is more similar to the original informal proof (Exam-
ple 2.11).
Note also that subproofs can have their own subproofs.
1 H − has − 2 ⇒ P − unsafe ∧ G − unsafe ∨ J − unsafe ∧ WaterWorld
P − unsafe ∨ G − unsafe ∧ J − unsafe axiom, choosing a
grouping of the
ternary ∨, as
justi ed by ∨
commutativity
(Example 2.14)
2 subproof:` H − has − 2
2.a H − has − 2 ∧ J − safe Premise
2.b H − has − 2 ∧Elim (left), line
2.a
3 P − unsafe ∧ G − unsafe ∨ J − unsafe ∧ P − unsafe ∨ ⇒Elim, lines 1,3
G − unsafe ∧ J − unsafe
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
45
4 subproof:` ¬J − unsafe
4.a H − has − 2 ∧ J − safe Premise
4.b J − safe ∧Elim (right), line
4.a
4.c J − safe ⇒ ¬J − unsafe WaterWorld
axiom
4.d ¬J − unsafe ⇒Elim, lines
4.b,4.c
5 subproof:` P − unsafe ∧ G − unsafe
5.a subproof:G − unsafe ∧ J − unsafe `
false
5.a.i G − unsafe ∧ Premise for
J − unsafe subproof
5.a.ii J − unsafe ∧Elim (right), line
5.a.1
5.a.iii false falseIntro, lines
4,5.a.2
5.b ¬ (G − unsafe ∧ J − unsafe) RAA, line 5.a
5.c P − unsafe ∧ G − unsafe ∨ J − unsafe ∧ CaseElim (right),
P − unsafe lines 3,5.b
5.d subproof:J − unsafe ∧ P − unsafe `
false
5.d.i J − unsafe ∧ Premise for
P − unsafe subproof
5.d.ii J − unsafe ∧Elim (left), line
5.d.1
5.d.iii false falseIntro, lines
4,5.d.2
5.e ¬ (J − unsafe ∧ P − unsafe) RAA, line 5.d
5.f P − unsafe ∧ G − unsafe CaseElim (right),
lines 5.c,5.e
6 G − unsafe ∧Elim (right), line
5
Table 2.18
A standard way of presenting proofs is by using lemmas to show parts of the proofs. Lemmas are simply
formulas which we prove not as an end result, but as intermediate steps in a larger proof. So, they are simply
another way of presenting subproofs.
Example 2.22
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
46 CHAPTER 2. PROPOSITIONAL LOGIC
Figure 2.6: Example WaterWorld board
Consider the above gure (Figure 2.6). We'll show B − has − 1 ∧ G − has − 1 ∧ J − has − 1 `
K − unsafe. We'll do this through the following series of lemmas:
• Lemma A: ¬A − unsafe, G − has − 1 ` H − unsafe
• Lemma B: ¬A − unsafe, B − has − 1 ` C − unsafe
• Lemma C: H − unsafe, C − unsafe, J − has − 1 ` false
• Lemma D: A − unsafe, B − has − 1 ` C − safe
• Lemma E: A − unsafe, G − has − 1 ` H − safe
• Lemma F: C − safe, H − safe, J − has − 1 ` K − unsafe
First, we'll show the main proof, assuming each of the lemmas. Then, proofs of each of the lemmas
will follow.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
47
1 B − has − 1 ∧ G − has − 1 ∧ J − has − 1 Premise
2 B − has − 1 ∧Elim (left), line 1
3 G − has − 1 ∧ J − has − 1 ∧Elim (right), line 1
4 G − has − 1 ∧Elim (left), line 3
5 J − has − 1 ∧Elim (right), line 3
6 subproof:¬A − unsafe ` false
6.a ¬A − unsafe Premise for subproof
6.b H − unsafe Lemma A, lines 6.a,4
6.c C − unsafe Lemma B, lines 6.a,2
6.d false Lemma C, lines
6.b,6.c,5
7 A − unsafe RAA, line 6
8 C − safe Lemma D, lines 7,2
9 H − safe Lemma E, lines 7,3
10 K − unsafe Lemma F, lines 8,9,5
Table 2.19
And that's the desired proof ! Now it just remains to show each of the six lemmas.
Lemma A: ¬A − unsafe, G − has − 1 ` H − unsafe
1 ¬A − unsafe Premise
2 G − has − 1 Premise
3 subproof:A − unsafe ∧ H − safe ` false
3.a A − unsafe ∧ H − safe Premise for subproof
3.b A − unsafe ∧Elim
3.c false falseIntro, lines 1,3b
4 ¬ (A − unsafe ∧ H − safe) RAA, line 3
5 G − has − 1 ⇒ A − safe ∧ H − unsafe ∨ WaterWorld axiom
A − unsafe ∧ H − safe
6 A − safe ∧ H − unsafe ∨ A − unsafe ∧ H − safe ⇒Elim, lines 5,2
7 A − unsafe ∧ H − safe ∨ A − safe ∧ H − unsafe Theorem: ∨ commutes,
line 6
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
48 CHAPTER 2. PROPOSITIONAL LOGIC
8 A − safe ∧ H − unsafe CaseElim, lines 4,7
9 H − unsafe ∧Elim (right), line 8
Table 2.20
Lemma B: ¬A − unsafe, B − has − 1 ` C − unsafe
1 ¬A − unsafe Premise
2 B − has − 1 Premise
3 subproof:A − unsafe ∧ C − safe ` false
3.a A − unsafe ∧ C − safe Premise for subproof
3.b A − unsafe ∧Elim (left), line 3a
3.c false falseIntro, lines 1,3b
4 ¬ (A − unsafe ∧ C − safe) RAA, line 3
5 B − has − 1 ⇒ A − safe ∧ C − unsafe ∨ WaterWorld axiom
A − unsafe ∧ C − safe
6 A − safe ∧ C − unsafe ∨ A − unsafe ∧ C − safe ⇒Elim, lines 5,2
7 A − unsafe ∧ C − safe ∨ A − safe ∧ C − unsafe Theorem: ∨ commutes,
line 6
8 A − safe ∧ C − unsafe CaseElim, lines 4,7
9 C − unsafe ∧Elim (right), line 8
Table 2.21
Proving the other lemmas is left as an exercise to the reader.
Note that we took a little shortcut: we used the lemmas as if they were inference rules. According to
our previous de nition of proofs, we technically should present the lemma as a subproof and then use an
inference rule or two to show how that applies, as we've done in previous examples. This shorter form is
common practice and much easier to read.
In summary, we must state one of the following four possible reasons for each step in a proof, allowing
subproofs.
• This step's WFF is a premise.
• This step's WFF is an axiom.
• This step's WFF follows from a inference rule applied to previous steps' WFFs. The reason includes a
statement of which inference rule is used and how.
• This step's WFF follows from a subproof, where that subproof may temporarily introduces additional
premises. The reason includes the entire subproof. When that subproof has been shown elsewhere,
such as in class or another exercise, it may simply be cited, for brevity. Of course, subproofs may have
additional embedded subproofs, in turn.
Technically, when using subproofs, one must be careful to rename variables, to avoid clashes. Rather than
formalize this notion, we'll leave it as obvious.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
49
2.4.3 the soundness and completeness of inference rules24
The folly of mistaking a paradox for a discovery, a metaphor for a proof, a torrent of verbiage for a spring of
capital truths, and oneself for an oracle, is inborn in us. Paul Valery, poet and philosopher (1871-1945)
Throughout this discussion, we've implicitly assumed that if we've proven something, it must be true. But
we should be careful: What if one of those listed inference rule isn't always valid? What if we introduced
a new rule? (Sure, you'd probably balk if we proposed something silly like a ∨ b ⇒ a, or even more
degenerately false. But what about some more reasonable-sounding rule?) What if our new rule introduces
an inconsistency, when combined with the other rules in a some complicated way? In fact, are we absolutely
certain that this can't already happen with the inference rules we have?! This brings us back to the questions
of soundness and completeness (Section 2.3.3) of a proof system. Fortunately, the system presented here is
both sound and complete (though proving this is beyond our current scope). However, we can rest assured,
that for propositional logic, what we can prove really does correspond entirely to what is true.
Exercise 2.4.3.1 (Solution on p. 69.)
If we omitted the RAA inference rule, would this new system be sound? Would it be complete?
2.4.4 proofs and programming25
2.4.4.1 Proofs and programming
Proofs are organized a lot like programs. Based on some premises (inputs), we obtain some conclusion
(output) after using a series of inference rules (basic computation like addition and other operations). Using
subproofs, especially when citing previous proofs, is just like organizing our program into functions that can
be used many times.
Naturally, since using inference rules is not only how people prove things, but also computers. A clear
example is in type checking. The core idea of type checking a function application is If function f takes an
argument of type α and producing an output of type β, and expression exp is of type α, then f(exp) is of
type β . This type rule closely resembles ⇒Elim: If a proven formula is a ⇒ b and other proven formula
is a, then together, b is a proven formula. Furthermore, this similarity is highlighted by notation in many
programming languages which would write the type of f as α → β . Type rules are simply inference rules for
proving results about the types of programs, and in most typical programming languages these rules closely
correspond to those we are using for logic. This correspondence is known as the Curry-Howard Isomorphism.
As with logic, we want type checkers to be sound and complete. Soundness here means that if the
program passes type checking, when we execute the program (or single function) and get a value, that value
is of the stated type. In other words, if our program type checks, then we are guaranteed that some kinds of
errors will not happen at run-time. That also means that if our program would have a run-time type error,
the type checker will correctly report that our program is erroneous. Completeness here means that if we
execute the program (or single function) and get a value of a certain type, then our type checker indeed tells
us that type.
Note that type checking is still an area of active research, since the job is made di cult in the presence
of language features such as inheritance, multiple inheritance, dynamic class loading, etc. When people
introduce new computer languages with new features, and want to claim that their new language is type
safe (that no function ever will be applied to the wrong type at run-time), then the paper which introduces
the language will contain such a proof.
24 This content is available online at <http://cnx.org/content/m12078/1.10/>.
25 This content is available online at <http://cnx.org/content/m12079/1.7/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
50 CHAPTER 2. PROPOSITIONAL LOGIC
2.4.5 conclusions26
2.4.5.1 Are we done yet?
These inference rules may seem limited, and you may have some more general ones in mind. Soon, we'll
see additional inference rules in the context of rst-order logic, which will give us a richer set of proofs.
In general, a hard problem is nding a language that is both expressive enough to describe the domain
succinctly, but also limited enough to automate reasoning. This is a very practical issue in type checking
and other program analysis. While it can be easy to nd some program errors automatically, it is very
di cult or impossible to guarantee that you can nd all errors (of some speci c kind, like type errors).
One thing we would like to eliminate is the need (at least technically) to restate structurally identical
proofs, as discussed for commutativity (Example 2.14). We will be able to add the idea of generalizing such
proofs directly into the logic and inference rules.
Despite the desire for more exible reasoning, we'd also like to consider whether we have more inference
rules than are necessary. Are some of them redundant? This is similar to the software rule that we should
have a single point of control, or the similar idea that libraries should provide exactly one way of doing
something. In general, this is not easy to ensure. We have shown that some potential additional inference
rules, like commutativity and associativity, weren't necessary. But we haven't shown our core inference
rules to be minimal. What do you think? (See the homework exercise problems on the redundancy of not-
elimination (Exercise 2.6.16), not-introduction (Exercise 2.6.17), and case-elimination (Exercise 2.6.18).)
2.4.5.2 Distinctness of the approaches (optional)
You might be wondering can we use propositional equivalences as axioms when using inference rules?
The short answer is no. First, Boolean equivalences are pairs of formulas, whereas axioms are individual
formulas. Second, none of our inference rules mention equivalences.
However, let's reword the question could we use propositional equivalences when using inference rules?
It would make sense to add an inference rule to allow this. One possibility would be an inference rule that
turns an equivalence into an implication: if we know φ ≡ ψ, then we know φ ⇒ ψ . Another possibility
would be an inference rule that allows us to substitute equivalence subterms, as we do in equivalence proofs:
if we know φ≡ψ and θ, then we know θ[φ7→ψ ], i.e., θ, except with instances of φ replaced by ψ . With
either, we would also have to allow equivalence proofs as subproofs or lemmas in inference proofs.
Traditionally, and in our presentation, we do not combine equivalences and inference rules in any such
way. The disadvantage of combining them is that instead of two relatively simple proof systems, you would
have one more complicated proof system. It would be harder to learn all that you could do in such a system,
and for theorists, it would be harder to prove things such as soundness and completeness for the combined
system. In learning and describing proofs, it is best to keep them separate. However, the advantage would be
shorter proofs. When using the combined system, you'd have exibility to use whichever technique suited
the current step best. In practice, people commonly combine these and other proof techniques.
2.5 Exercises for Propositional Logic I27
Please write logic formulas using the syntax previously de ned, using false (or for brevity, F), true (or T),
¬, ∧, ∨, and ⇒. Except where directed, use only these connectives.
aside: You can download WaterWorld
28 if you like. At Rice University, WaterWorld is installed
on OwlNet, in /home/comp280/bin/waterworld.
26 This content is available online at <http://cnx.org/content/m12080/1.11/>.
27 This content is available online at <http://cnx.org/content/m10514/2.57/>.
28 http://www.teachlogic.org/WaterWorld/
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
51
2.5.1 Propositional Logic
Exercise 2.5.1 (Solution on p. 69.)
[Practice problem− − −solution provided.]
Your friend Tracy argues: It is bad to be depressed. Watching the news makes me feel
depressed. Thus, it's good to avoid watching the news.
Regardless of whether the premises and conclusion are true, show that the argument is not, by
showing it doesn't hold for all domains. Replace depressed and watching news with expressions
which leave the premises true, but the conclusion false (or at least, what most reasonable people
would consider false).
Exercise 2.5.2 (Solution on p. 69.)
[Practice problem− − −solution provided.]
An acquaintance says the following to you: Chris claims knowledge is more important than
grades. But she spent yesterday doing an extra-credit assignment which she already knew how to
do. Therefore, she's a hypocrite and deserves no respect.
Regardless of whether the premises and conclusion are true, show that the argument is not, by
showing it doesn't hold for all domains. Replace knowledge and grades with expressions which
give you true premises, but a false conclusion (or at least, what most reasonable people would
consider false).
note: Exaggerate knowledge to something more important, and grades to something less
important.
Exercise 2.5.3 (Solution on p. 70.)
[Practice problem− − −solution provided.]
While the following argument may sound plausible initially, give a particular situation where
the conclusion doesn't hold (even though the premises do). Then, in a sentence or two, sketch why
your counterexample may still represent rational behavior by pointing out a real-world subtlety
that the initial argument ignored.
1. If a certain out t meets a dress code, then per force all less-revealing out ts also meet that
dress code.
2. In public transportation projects, out of two alternatives, the cheaper one which gets the job
done is the better choice.
Exercise 2.5.4
Choose just one of the following informal arguments. While the argument sounds plausible
initially, give a particular situation where the conclusion doesn't hold (even though the premises
do). Then, brie y state why your counterexample may still represent rational behavior by pointing
out a real-world subtlety that the initial argument ignored.
1. [cell phone] Talking on a cell phone while driving increases the likelihood of an accident.
Interestingly, hands-free phones do not signi cantly help
29 . It's just the distraction of a
phone conversation that causes the problem.
2. [equivalent products] If two companies oer two materially equivalent products, then most
everybody will buy the cheaper one.
3. [service] In a free market, if a company doesn't oer good service, individual customers will
become fed up and take their business elsewhere.
4. [web browser] If there are two versions of a free web browser, and they run equally quickly,
users will use the one with better features/interface.
29 http://www.sciencedaily.com/releases/2003/01/030129080944.htm
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
52 CHAPTER 2. PROPOSITIONAL LOGIC
5. [door-locking] Anybody who really wants to break into your house while you're gone will be
able to. (For instance, using a towel to mu e sound, break the corner of a back window,
reach in and unlatch the window, and climb through.) So there's no point in locking your
front door.
Exercise 2.5.5 (Solution on p. 70.)
[Practice problem− − −solution provided.]
Let p, q , and r be the following propositions:
• p: You get an A on the nal exam
• q: You do every exercise in the book.
• r: You get an A in this class.
Write the following formulas using p, q , and r and logical connectives.
1. You get an A in this class, but you do not do every exercise in the book.
2. To get an A in this class, it is necessary for you to get an A on the nal.
3. Getting an A on the nal and doing every exercise in the book is su cient for getting an A
in this class.
Exercise 2.5.6
Translate the following English sentences into propositional logic. Your answers should be WFFs.
1. If the Astros win the series ( AW), then pigs will y ( PF).
2. Pigs will not y, and/or bacon will be free ( BF).
3. The Astros will win the series, or bacon will be free, but not both.
Exercise 2.5.7 (Solution on p. 70.)
[Practice problem− − −solution provided.]
It just so happens that all the web pages in Logiconia which contain the word Poppins also
contain the word Mary. Write a formula (a query) expressing this. Use the proposition Poppins
to represent the concept the web page contains 'Poppins ' (and similar for Mary).
Exercise 2.5.8
• If a Logicanian page contains the word weasel, then it also contains either words or eyed;
and
• Whenever a Logiconian page contains the word mongoose, it does not also contain the word
weasel; and
• Finally, all Logiconian pages contain the word Logiconia, rather patriotically.
Write a formula expressing all this. (Your formula will involve ve propositions: weasel, words, . . .
Try to nd a formula which mirrors the wording of the English above.)
Given the above statements, if a web page in Logiconia does not contain weasel, does it contain
mongoose?
Let's go meta for a moment: Is this web page Logiconian? (Yes, this one you're looking at now,
the one with the homework problems.) Explain why or why not.
Exercise 2.5.9
Dierent search engines on the web have their own syntax for specifying searches.
aside: Note that a formula may be true for some web pages, and false for others. The search
engine is concerned with nding all web pages which satisfy the formula. This is called a query,
in database lingo.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
53
Only a few
30 allow full Boolean queries. Some interpret a list of several words in a row as an implicit
conjunction, others as an implicit disjunctions.
1. Read about the search syntax for the search language of eBay ® 31 . Write an eBay query
for auctions which contain border, do not contain common, and contain at least one of
foreign or foriegn [sic, misspellings are a great way to nd underexposed auctions].
2. Google¿'s advanced search
32 is typical for the online search engines. In particular, you can
search for results containing all of a, b, . . ., at least one of c, d, . . ., and none of e, f , . . . .
Describe how that corresponds to a Boolean formula.
3. Give an example of a Boolean formula which cannot be rewritten to conform to Google's
advanced search interface.
Exercise 2.5.10 (Solution on p. 70.)
[Practice problem− − −solution provided.]
30 http://www.exalead.com/search/
31 http://pages.ebay.com/help/search/search-commands.html
32 http://www.google.com/advanced_search?hl=en
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
54 CHAPTER 2. PROPOSITIONAL LOGIC
Figure 2.7: A sample WaterWorld board
Consider the particular board shown in the above gure (Figure 2.7).
1. Y − safe, Y − has − 0, and ¬Y − has − 2 are among the formulas which are true for this board
but not for all boards. That is, they are neither domain axioms nor tautologies. Give two
other such formulas.
2. V − safe might or might not be true for this board. Give two other such formulas.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
55
Exercise 2.5.11
In that same board (Figure 2.7), is location W safe? What is your informal reasoning? (List all
your small steps.) Similarly for location P.
Exercise 2.5.12
Give a domain axiom of WaterWorld which was omitted in the ellipses in the WaterWorld domain
axioms (Section 6.5).
Exercise 2.5.13
Even allowing for ellision, the list of WaterWorld domain axioms (Section 6.5) is incomplete, in
a sense. The game reports how many pirates exist in total, but that global information is not
re ected in the propositions or axioms.
First, assume we only use the default WaterWorld board size and number of pirates, i.e., ve.
Give samples of the additional axioms that we need.
Next, generalize your answer to model the program's ability to play the game with a dierent
number of pirates.
Exercise 2.5.14
Give one WFF which meets all three conditions:
• true in all WaterWorld boards (A theorem of WaterWorld)
• not already listed as one of the WaterWorld domain axioms (Section 6.5), and
• not a tautology of propositional logic (can be made false in some truth assignment, though it
may not be a truth assignment which satis es the waterworld axioms).
2.5.2 Reasoning with Truth Tables
When writing truth tables, please list rows in the order used in all examples: FF, FT, TF, TT. For three-input
tables, use the above four lines preceded by F, then the above four lines preceded by T.
Exercise 2.5.15
In a truth table for two inputs, provide a column for each of the sixteen possible distinct functions.
Give a small formula for each of these functions.
note: These functions will include those for ∧, ∨, and the other connectives whose truth tables
you've already seen (Section 2.1.1.1.2: Connectives).
Exercise 2.5.16 (Solution on p. 70.)
[Practice problem− − −solution provided.]
Write the truth table for xnor, the negation of exclusive-or, What is a more common name for
this Boolean function?
Exercise 2.5.17
How many years would it take to build a truth table for a formula with 1000 propositions? Assume
it takes 1 nanosecond to evaluate each formula.
A formula with 1000 propositions clearly isn't something you would create by hand. However,
such formulas easily arise when modeling the behavior of a program with a 1000-element data
structure.
Exercise 2.5.18
Use truth tables to answer each of the following. Showing whether the connectives obey such
properties via truth tables is one way of establishing which equivalences or inference rules we
should use.
1. Show whether ⇒ is commutative.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
56 CHAPTER 2. PROPOSITIONAL LOGIC
2. Show whether ⊕ is commutative.
3. Show whether ⊕ is associative.
4. Prove that ∧ distributes over ∨: φ ∧ (ψ ∨ θ) ≡ φ ∧ ψ ∨ φ ∧ θ
note: This version is left-distributivity. Right-distributivity follows from this plus the
commutativity of ∧.
5. Prove that ∨ distributes over∧: φ ∨ ψ ∧ θ ≡ (φ ∨ ψ) ∧ (φ ∨ θ)
6. Show whether ∧ or ∨ distribute over ⇒.
7. Show whether ⇒ distributes over ∧ or ∨.
8. Show whether ∧ or ∨ distribute over ⊕.
9. Show whether ⊕ distributes over ∧ or ∨.
Exercise 2.5.19
For each of the following, nd a satisfying truth assignment, (values of the propositions which
make the formula true), if any exists.
1. (a ⇒ ¬b) ∧ a
2. (a ⇒ c ⇒ ¬b) ∧ (a ∨ b)
Exercise 2.5.20
For each of the following, nd a falsifying truth assignment, (values of the propositions which
make the formula false), if any exists.
1. (a ⇒ ¬b) ∨ a
2. (¬b ⇒ (a ⇒ c)) ∨ a ∧ b
Exercise 2.5.21
Formula φ is stronger than formula ψ if ψ is true whenever φ is true (i.e., φ is at least a strong
as ψ ), but not conversely. Equivalently, this means that φ ⇒ ψ is always true, but ψ ⇒ φ is not
always true.
As one important use of this concept, if we know that ψ ⇒ θ, and that φ is stronger than ψ,
then we also know that φ ⇒ θ. That holds simply by transitivity. Another important use, which is
outside the scope of this module, is the idea of strengthening an inductive hypothesis.
Similarly, φ is weaker than formula ψ whenever ψ is stronger than φ.
Show which of the following hold. When true, show φ⇒ψ is true by a truth table, and show
a falsifying truth assignment for ψ ⇒ φ. When false, give a truth table and truth assignment the
other way around.
1. a ∧ b is stronger than a ∨ b.
2. a ∨ b is stronger than a.
3. a is stronger than a ⇒ b.
4. b is stronger than a ⇒ b.
Exercise 2.5.22
Using truth tables, show that (a ∨ c) ∧ (b ⇒ c) ∧ (c ⇒ a) is equivalent to (b ⇒ c) ∧ a. but
not equivalent to (a ∨ c) ∧ (b ⇒ c).
Exercise 2.5.23 (Solution on p. 70.)
[Practice problem− − −solution provided.]
When writing a complicated conditional that involves multiple pieces of data, it is easy to
incorrectly oversimplify. One strategy for avoid mistakes is to write such code in a two-step process.
First, write a conditional with a case for every possible combination, as in a truth table. Second,
simplify the conditional.
Using this approach, we might obtain the following code after the rst step. Simplify this code.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
57
list merge_sorted_lists(list list1, list list2)
{
if (is_empty(list1) && is_empty(list2))
return empty_list;
else if (is_empty(list1) && !is_empty(list2))
return list2;
else if (!is_empty(list1) && is_empty(list2))
return list1;
else if (!is_empty(list1) && !is_empty(list2)) {
if (first_element(list1) < first_element(list2))
return make_list(first_element(list1),
merge_sorted_lists(rest_elements(list1),list2));
else if (first_element(list1) >= first_element(list2))
return make_list(first_element(list2),
merge_sorted_lists(list1,rest_elements(list2)));
}
}
Exercise 2.5.24
Consider the following conditional code, which returns a boolean value.
int i;
bool a,b;
...
if (a && (i > 0))
return b;
else if (a && i <= 0)
return false;
else if (a || b)
return a;
else
return (i > 0);
Simplify it by lling in the following blank with a single Boolean expression. Do not use a conditional
(such as if or ?:).
int i;
bool a,b;
...
return ________________;
Use either Java/C++ or Scheme syntax. In the former case, please fully parenthesize to make
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
58 CHAPTER 2. PROPOSITIONAL LOGIC
your formula unambiguous, rather than relying on Java's
33 or C++'s34 many levels of operator
precedence.
2.5.3 Reasoning with Equivalences
Exercise 2.5.25 (Solution on p. 71.)
[Practice problem− − −solution provided.]
Using algebraic identities (Section 6.1), and the de nition or nor (mnemonic: not or), written
↓, φ ↓ ψ ≡ ¬ (φ ∨ ψ), express the function ∧ in terms of ↓ only. That is, give a formula which
doesn't use ∧, ∨, ¬, but instead only uses ↓ and which has the same truth table as φ ∧ ψ .
Exercise 2.5.26
Similar to the previous exercise, express each of the following using nand (Example ) only, and
prove correctness using the algebraic identities (Section 6.1).
This operation is particularly interesting, since making a NAND gate in hardware requires only
two transistors.
1. ¬
2. ∧
3. ∨
Exercise 2.5.27
Using algebraic identities (Section 6.1), show that (a ∨ c) ∧ (b ⇒ c) ∧ (c ⇒ a) is equivalent
to (b ⇒ c) ∧ a.
This is an algebraic hand-evaluation: a series of formulas joined by ≡. Don't write just portions
of previous formulas and mysteriously re-introduce the dropped parts later. For each step, mention
which identity you used. It is also helpful if you underline the formula you are rewriting in the
next step. You can use commutativity and associativity without using a separate line, but mention
when you use it.
Exercise 2.5.28
In two exercises, you've shown the same equivalence by truth tables (Exercise 2.5.22) and by
algebraic identities (Exercise 2.5.27).
1. What is an advantage of using truth tables? What is an advantage of using identities?
2. In that truth table exercise (Exercise 2.5.22), you also showed two formulas φ and ψ non-
equivalent. It is also possible to do so with Boolean algebra rather than truth tables. How?
3. Describe a hybrid approach, combining truth tables and Boolean algebra, to prove the equiv-
alence and non-equivalence of formulas.
4. To ponder on your own without turning it in: Which approach appeals more to you?
Exercise 2.5.29
Using algebraic identities (Section 6.1), rewrite the formula (a ⇒ b ∨ c) ∧ ¬b to one with fewer
connectives.
33 http://java.sun.com/docs/books/tutorial/java/nutsandbolts/expressions.html
34 http://www.cppreference.com/operator_precedence.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
59
2.6 Exercises for Propositional Logic II35
2.6.1 Reasoning with Inference Rules
For proofs on this homework, remember that each step must be justi ed by one of the following:
• a premise,
• a WaterWorld axioms (Section 6.5.2: The domain axioms),
• a listed inference rule (Section 6.2) with the referenced line numbers (and, if ambiguous, substitutions
for the inference rule's meta-variables), or
• a subproof shown inline, or equivalently, a theorem/lemma shown previously.
Except where otherwise directed, you may use any theorem shown in the text or by a previous exercise, even
if that exercise was not assigned.
Exercise 2.6.1
Fill in the blank reasons in the following proof that ∨ commutes, that is, χ ∨ υ ` υ ∨ χ.
1 χ ∨ υ Premise
2 subproof:χ `υ ∨ χ
2.a χ Premise for subproof
2.b υ ∨ χ ∨Intro, line 2.a
3 subproof:υ `υ ∨ χ
3.a υ Premise for subproof
3.b υ ∨ χ ____________________
4 υ ∨ χ ____________________
Table 2.22
Exercise 2.6.2
Show that φ ∧ ψ, φ ⇒ θ, ψ ⇒ δ ` θ ∧ δ .
note: It should take around 8 steps.
Exercise 2.6.3
Show what is often called the implication chain rule: φ ⇒ ψ, ψ ⇒ θ ` φ ⇒ θ.
Exercise 2.6.4 (Solution on p. 71.)
[Practice problem− − −solution provided.]
Show what is often called negated-or-elimination (left): ¬ (φ ∨ ψ) ` ¬φ.
note: Think backwards. How can we end with ¬φ? One way is to end with RAA, under
the premise φ. Using that premise φ and the starting premise ¬ (φ ∨ ψ) can you derive the
contradiction?
Exercise 2.6.5
Using the inference rule RAA, prove ¬φ ` ¬ (φ ∧ ψ).
Exercise 2.6.6
Show that ¬W − safe ∨ ¬Y − unsafe ` W − unsafe ∨ Y − safe.
35 This content is available online at <http://cnx.org/content/m12352/1.20/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
60 CHAPTER 2. PROPOSITIONAL LOGIC
note: The proof is a bit longer than you might expect. Use the ∨Elim inference rule to get the
nal result.
Exercise 2.6.7
In our inference rules, unlike our equivalences, we chose to not include any corresponding to
distributivity.
1. Prove a left-hand version of one direction of distributivity: φ ∧ (ψ ∨ θ) ` φ ∧ ψ ∨ φ ∧ θ.
∧'s commutativity
2. Use the previous part's result, plus to prove the corresponding right-hand
version: (ψ ∨ θ) ∧ φ ` ψ ∧ φ ∨ θ ∧ φ.
Exercise 2.6.8
In our inference rules, unlike our equivalences, we chose to not include any corresponding to
DeMorgan's Law. Show that each of the following versions is still provable.
1. φ ∨ ψ ` ¬ (¬φ ∧ ¬ψ)
2. ¬ (φ ∨ ψ) ` ¬φ ∧ ¬ψ
3. φ ∧ ψ ` ¬ (¬φ ∨ ¬ψ)
4. ¬ (φ ∧ ψ) ` ¬φ ∨ ¬ψ
Exercise 2.6.9
The above exercise suggests that it would be useful to have an inference rule or theorem that says
given θ ` ¬δ , then ¬θ ` δ . Or, equivalently, because of ⇒Intro and ⇒Elim, θ ⇒ ¬δ ` ¬θ ⇒ δ .
Why don't we?
Exercise 2.6.10
In our inference rules, unlike our equivalences, we have nothing that directly equates φ⇒ψ and
¬φ ∨ ψ . Prove each of the following.
1. φ ⇒ ψ ` ¬φ ∨ ψ
2. ¬φ ∨ ψ ` φ ⇒ ψ
Exercise 2.6.11
Prove the following: φ ⇒ ψ , ψ ⇒ φ ` φ ∧ ψ ∨ ¬φ ∧ ¬ψ
Exercise 2.6.12
Prove what is commonly called the Law of Excluded Middle36 : ` χ ∨ ¬χ.
1. Give a short proof citing our previous proof (Example 2.18) of ` ¬ (χ ∧ ¬χ) and the relevant
version of DeMorgan's Law from above (Exercise 2.6.8).
2. Give a direct version without using previous theorems.
note: Use RAA two or three times.
Exercise 2.6.13
Prove the missing steps and reasons in the following WaterWorld proof of X − has − 1 `
W − unsafe ∨ Y − unsafe.
36 http://en.wikipedia.org/wiki/Law_of_excluded_middle
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
61
1 X − has − 1 ____________________
2 ____________________ WaterWorld axiom
3 ____________________ ⇒Elim, lines 1,2
4 subproof:W − safe ∧ Y − unsafe ` W − unsafe ∨
Y − unsafe
4.a W − safe ∧ Y − unsafe Premise for subproof
4.b Y − unsafe ____________________
4.c W − unsafe ∨ ____________________
Y − unsafe
5 subproof:¬ (W− safe ∧ Y − unsafe) `
W − unsafe ∨ Y − unsafe
5.a ¬ (W − safe ∧ Y − unsafe)Premise for subproof
5.b W − unsafe ∧ Y − safe CaseElim (left), lines
____________________
where φ=
____________________,
and ψ=
____________________
5.c ____________________
____________________
5.d W − unsafe ∨ ____________________
Y − unsafe
6 W − safe ∧ Y − unsafe ∨ Theorem: Excluded
¬ (W − safe ∧ Y − unsafe) Middle, where χ=
____________________
7 W − unsafe ∨ Y − unsafe ____________________
Table 2.23
Exercise 2.6.14 (Solution on p. 72.)
[Practice problem− − −solution provided.]
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
62 CHAPTER 2. PROPOSITIONAL LOGIC
Figure 2.8: A sample WaterWorld board
Given the above gure (Figure 2.8), and using any of the immediately obvious facts as premises,
prove that location P is safe by using our proof system and the WaterWorld axioms.
While this proof is longer (over two dozen steps), it's not too bad when sub-proofs are used
appropriately. To make life easier, you may use the following theorem: Q − has − 1 ⇒ P − safe ∧
R − safe ∨ P − safe ∧ W − safe ∨ R − safe ∧ W − safe, along with any proven previously. When
looking at the given board, you can use premises like Y − safe as well as ¬Y − unsafe.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
63
Exercise 2.6.15
Starting from the WaterWorld axiom Q − has − 1 ⇒ P − safe ∧ R − safe ∧ W − unsafe ∨
P − safe ∧ R − unsafe ∧ W − safe ∨ P − unsafe ∧ R − safe ∧ W − safe, we could prove the
following theorem cited in the previous problem (Exercise 2.6.14): Q − has − 1 ⇒ P − safe ∧
R − safe ∨ P − safe ∧ W − safe ∨ R − safe ∧ W − safe.
Prove the following theorem which is slightly simpler: φ ⇒ ψ ∧ θ ∨ δ ∧ ` φ ⇒ ψ ∨ δ .
note: If you have trouble, rst prove an even simpler version: φ ⇒ ψ ∧ θ ` φ ⇒ ψ.
Exercise 2.6.16 (Solution on p. 73.)
[Practice problem− − −solution provided.]
Show that the ¬Elim inference rule is redundant in our system. In other words, without using
¬Elim, prove that ¬¬φ ` φ.
Exercise 2.6.17
Show that the ¬Intro inference rule is redundant in our system. In other words, without using
¬Intro, prove that φ ` ¬¬φ. To make sure that you're not hiding any uses of ¬Intro, also do not
use any previous theorems.
Exercise 2.6.18
Show that the CaseElim inference rule is redundant in our system. For brevity, we'll just consider
the left-hand version. In other words, without using CaseElim, prove that φ ∨ ψ , ¬φ ` ψ . To make
sure that you're not hiding any uses of CaseElim, also do not use any previous theorems.
Exercise 2.6.19
• State where on a board pirates could be positioned, so that: P − has − 1 ∧ U − has − 1 ∧
W − has − 1, but X − safe.
• Compare this with a previous theorem (Example 2.22), B − has − 1 ∧ G − has − 1 ∧
J − has − 1 ⇒ K − unsafe, the same idea shifted down a couple of rows. Suppose we try
to translate this theorem's proof so as to conclude ¬X − safe (clearly untrue, by the above).
What is the rst step of the modi ed proof which doesn't hold when B ,G,J ,K are mindlessly
replaced with P ,U ,W ,X , respectively? (Just give a line number; no explanation needed. Your
answer will be of the form Lemma A line 1 or main proof line 2.)
• We've just seen that the mindless changing of location-names introduces false steps. But we
can be a little smarter, and modify the false step to get a formula which is true, and is also
still in the spirit of the original proof. We can thus patch the problem from the previous part,
and continue on modifying the original proof for several more steps. But clearly we can't
translate the entire original proof; we eventually hit a more fundamental snag: a formula
which isn't true, yet can't be patched up, either. What is the rst line that can't be patched?
(Again, just give a line number; no explanation needed. Your answer will be of the form
Lemma A line 1 or main proof line 2.)
Exercise 2.6.20
Which is worse, having an unsound (but complete) inference system or an incomplete (but sound)
one? Why?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
64 CHAPTER 2. PROPOSITIONAL LOGIC
Solutions to Exercises in Chapter 2
Solution to Exercise 2.1.1.1 (p. 19)
Truth table for ∨ (OR)
a b (a ∨ b)
false false false
false true true
true false true
true true true
Table 2.24
Solution to Exercise 2.1.1.2 (p. 19)
Truth table for ¬ (NOT)
a ¬a
false true
true false
Table 2.25
Solution to Exercise 2.1.1.3 (p. 19)
Truth table for ⇒ (IMPLIES)
a b (a ⇒ b)
false false true
false true true
true false false
true true true
Table 2.26
Solution to Exercise 2.1.1.4 (p. 20)
Exactly one is true if either (a is true, and b is false) or (a is false, and b is true). So, one way to de ne it
is a ⊕ b ≡ a ∧ ¬b ∨ ¬a ∧ b.
The two halves of that formula also correspond to the two true rows of xor's truth table:
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
65
Truth table for xor
a b (a ⊕ b)
false false false
false true true
true false true
true true false
Table 2.27
Solution to Exercise 2.1.1.5 (p. 20)
1. Inclusive.
2. Exclusive.
3. Inclusive.
4. Exclusive (hopefully).
Solution to Exercise 2.1.2.1 (p. 24)
Unsatis able.
Solution to Exercise 2.1.2.2 (p. 24)
Tautology, arguably.
Solution to Exercise 2.1.2.3 (p. 24)
Unsatis able, unless of course you interpret nobody as nobody of note.
Solution to Exercise 2.1.2.4 (p. 24)
Neither. If you interpret gets late as a social issue but early as a clock issue, then the statement might
be true, depending on where here is.
Solution to Exercise 2.1.2.5 (p. 24)
Unsatis able, except perhaps in a karmic
37 sense.
Solution to Exercise 2.2.1.1 (p. 25)
Truth table to check associativity of implication
a b c (a ⇒ (b ⇒ c)) ((a ⇒ b) ⇒ c)
false false false true false
false false true true true
false true false true false
false true true true true
true false false true true
true false true true true
true true false false false
true true true true true
Table 2.28
37 http://www.cs.rice.edu/∼ian/Rants/karmaIsReal.shtml
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
66 CHAPTER 2. PROPOSITIONAL LOGIC
By inspecting the two right-most columns, we see that the formulas are indeed not equivalent. They have
dierent values for two truth-settings, those with a = false and c = false.
Solution to Exercise 2.2.1.2 (p. 26)
In the original code, we return value2 when the rst case is false, but the second case is true. Using a
WFF, when ¬ (a ∧ b) ∧ (a ∨ b). Is this equivalent to the WFF a ∧ ¬b ∨ ¬a ∧ b? Here is a truth table:
Truth table for comparing conditionals' equivalence
a b ¬ (a ∧ b) (a ∨ b) (¬ (a ∧ b) ∧ (a ∨ b)) (a ∧ ¬b) (¬a ∧ b) ((a ∧ ¬b) ∨ (¬a ∧ b))
false false true false false false false false
false true true true true false true true
true false true true true true false true
true true false true false false false false
Table 2.29
Yes, looking at the appropriate two columns we see they are equivalent.
Solution to Exercise 2.2.2.1 (p. 27)
• 2 variables: As we're seen, 4 rows.
• 3 variables: 8 rows.
• 5 variables: 32 rows.
• 10 variables: 1024 rows.
• n variables: 2n rows.
Solution to Exercise 2.2.2.2 (p. 27)
• With 2 variables, we have 4 rows. How many dierent ways can we assign true and false to those 4
positions? If you write them all out, you should get 16 combinations.
• With 3 variables, we have 8 rows and a total of 256 dierent functions.
n
• With n variables, we have 2n rows and a total of 22 dierent functions. That's a lot!
Solution to Exercise 2.3.1.1 (p. 31)
1 (a ∨ b) ∧ b
2 ≡ (a ∨ b) ∧ (b ∨ false) Identity of ∨
3 ≡ (b ∨ a) ∧ (b ∨ false) Commutativity of ∨
4 ≡ b ∨ a ∧ false Distributivity of ∨ over ∧
5 ≡ b ∨ false Dominance of ∧
6 ≡b Identity of ∨
Table 2.30
Solution to Exercise 2.3.1.2 (p. 31)
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
67
1 a ∧ (a ⇒ b) ⇒ b
2 ≡ a ∧ (¬a ∨ b) ⇒ b De nition of ⇒
3 ≡ a ∧ ¬a ∨ a ∧ b ⇒ b Distributivity of ∨ over ∧
4 ≡ false ∨ a ∧ b ⇒ b Complement
5 ≡ a ∧ b ∨ false ⇒ b Commutativity of ∨
6 ≡a ∧ b⇒b Identity of ∨
7 ≡ ¬ (a ∧ b) ∨ b De nition of ⇒
8 ≡ ¬a ∨ ¬b ∨ b DeMorgan's law
9 ≡ ¬a ∨ ¬b ∨ b Associativity of ∨
10 ≡ ¬a ∨ b ∨ ¬b Commutativity of ∨
11 ≡ ¬a ∨ true Complement
12 ≡ true Dominance of ∨
Table 2.31
Solution to Exercise 2.3.2.1 (p. 33)
• CNF: (a ∨ b) ∧ (¬a ∨ b ∨ c) ∧ (¬a ∨ ¬b)
• DNF: (¬a ∧ b) ∨ (a ∧ ¬b ∧ c)
aside: Karnaugh maps
38 are a general technique for nding minimal CNF and DNF formulas.
They are most easily used when only a small number of variables are involved. We won't worry
about minimizing formulas ourselves, though.
Solution to Exercise 2.3.3.1 (p. 34)
We can indeed reduce the question of Tautology to the question of Equivalence: if somebody asks you whether
φ is true, you can just turn around and ask your friend whether the following two formulas are equivalent:
φ, and true. Your friend's answer for this variant question will be your answer to your customer's question
about φ. Thus, the Tautology problem isn't particularly harder than the Equivalence problem.
But also, Equivalence can be reduced to Tautology: if somebody asks you whether φ is equivalent to
ψ , you can construct a new formula (φ ⇒ ψ) ∧ (ψ ⇒ φ). This formula is true exactly when φ and ψ are
equivalent. So, you ask your friend whether this bigger formula is a tautology, and you then have your answer
to whether the two original formulas were equivalent. Thus, the Equivalence problem isn't particularly harder
than the Tautology problem!
Given these two facts (that each problem reduces to the other), we realize that really they are essentially
the same problem, in disguise.
Solution to Exercise 2.3.3.2 (p. 34)
Compare the last two columns in the following:
Truth table to prove validity of conjunctive Redundancy
a b ¬a ∨ b a ∧ (¬a ∨ b) a ∧ b
false false true false false
false true true false false
true false false false false
true true true true true
38 http://www.ee.surrey.ac.uk/Projects/Labview/minimisation/karnaugh.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
68 CHAPTER 2. PROPOSITIONAL LOGIC
Table 2.32
Solution to Exercise 2.4.1.1 (p. 37)
Intuitively, this is straightforward. Since A − has − 2, then both of its two neighbors, including B, must be
unsafe. For this problem, let's be a bit more formal and use WFFs instead of prose in the steps.
1 A − has − 2 Premise
2 A − has − 2 ⇒ WaterWorld domain axiom, i.e.,
B − unsafe ∧ F − unsafe de nition of A − has − 2
3 B − unsafe ∧ F − unsafe lines 1,2
4 B − unsafe line 3
Table 2.33
Solution to Exercise 2.4.1.2 (p. 37)
Again a similar idea, if A − has − 1, then at least one of A's two neighbors must be unsafe. But, since we
know that one of these, G isn't unsafe, then the other, B , must be unsafe.
1 A − has − 1 ⇒ WaterWorld domain axiom
B − safe ∧ G − unsafe ∨
B − unsafe ∧ G − safe
2 A − has − 1 Premise
3 B − safe ∧ G − unsafe ∨ lines 1,2
B − unsafe ∧ G − safe
4 G − safe Premise
5 B − unsafe ∧ G − safe lines 3,4
6 B − unsafe line 5
Table 2.34
Solution to Exercise 2.4.1.3 (p. 39)
Here, we'll show only χ ∧ υ ∧ ω`χ ∧ υ ∧ ω and leave the other direction (and ∨'s associativity) to the
reader. These are all very similar to the previous commutativity example (Example 2.14).
1 χ ∧ υ ∧ ω Premise
2 χ ∧ υ ∧Elim (left), line 1
3 χ ∧Elim (left), line 2
4 υ ∧Elim (right), line 2
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
69
5 ω ∧Elim (right), line 1
6 υ ∧ ω ∧Intro, lines 4,5
7 χ ∧ υ ∧ ω ∧Intro, lines 3,6
Table 2.35
Note that we omitted the detailed explanation of how each rule applies, since this should be clear in each
of these steps.
Solution to Exercise 2.4.1.4 (p. 40)
First, if we know φ ` ψ, then that means there is some written proof. . . we know ` φ ⇒ ψ, simply by
⇒Intro.
If we know ` φ ⇒ ψ, then if we add a premise φ, then ψ follows by ⇒Elim.
Note how this proof is about other proofs! (However, while we reason about this particular inference
system, we're not using this system while proving things about it this proof is necessarily outside the
inference system.
Solution to Exercise 2.4.2.1 (p. 41)
1 α⇒β Premise
2 ¬β Premise
3 subproof:α ` false
3.a α Premise for subproof
3.b β ⇒Elim, lines 1,3.a
3.c false falseIntro, lines 2,3.b
4 ¬α RAA, line 3
Table 2.36
Solution to Exercise 2.4.3.1 (p. 49)
It would be sound: Look at all the possible proofs that can be made in the original system; all those proofs
lead to true conclusions (since that original system is sound, as we're claiming). If we just discard all those
that include RAA, the remaining proofs are still all true, so the smaller system is sound.
It would not be complete, though: As pointed out, RAA is our only way to prove negations without
premises. There are negated formulas that are true (and have no premises) for example ¬false. Without
RAA, we cannot provide a proof of ¬false, so the smaller system is incomplete.
Solution to Exercise 2.5.1 (p. 51)
Lots of possible counterexamples. It is bad to be depressed. Doing homework makes me depressed; so it's
good to not do my homework. Or, It is bad for people to be in physical pain. Childbirth causes pain.
Therefore childbirth needs be avoided by all people. If the original conclusion is really correct, Tracy needs
to elucidate some of his unspoken assumptions.
The aw seems to be along the lines of, avoiding bad in the short run may not always be good in the
long run (or equivalently, sometimes you have to choose the lesser of two evils). No, you weren't asked to
name a speci c aw, and reasonable people can dier on precisely what the aw is. (And, formal logic is not
particularly helpful here.) Nonetheless, uncovering hidden assumptions in arguments often helps understand
the real issues involved.
aside: For fun, pick up the front page of the daily newspaper, and see how many arguments use
faulty rules of inference and/or rely on unspoken premises (which not all might agree with). In
particular, political issues as spun to the mainstream press are often riddled with error, even though
there are usually reasonable arguments on both sides which policy-makers and courts debate.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
70 CHAPTER 2. PROPOSITIONAL LOGIC
Solution to Exercise 2.5.2 (p. 51)
Terry claims that encouraging human-rights is more important than playing Tetris. But Terry played
Tetris yesterday rather than volunteering with Amnesty International
39 . Most people wouldn't condemn
Terry as a hypocrite just because of this; even the most dedicated of people are entitled to some free time.
If your friend wants to prove Terry hypocritical, they'll have to provide further evidence or arguments.
Or similarly, Politician X claims to support science funding, but voted against a proposal to shift all
Medicare funds to NASA.
Solution to Exercise 2.5.3 (p. 51)
1. It can be socially acceptable to wear my swimsuit into a fast-food restaurant. My underwear is less
revealing than my swimsuit, and yet it would still raise many more eyebrows to go to that restaurant
in my underwear, than my swimsuit.
Clothes (and style in general) somehow encompass a form of communication, and people may object
to an out t's mood or message without actually objecting to how much the out t reveals. (Other
examples of communication-through-style include team logos, t-shirts with humorous slogans, and arm
bands.)
2. Buses are a lot cheaper than light rail. Yet, the light-rail here in Houston demonstrates that many
people who wouldn't routinely take a bus are willing to take light rail. (Only after we recognize this,
can we try to gure out what why the dierence exists, and then brainstorm to nd a better overall
solution.)
Solution to Exercise 2.5.5 (p. 52)
1. r ∧ ¬q
2. r⇒p
Think of the English being reworded to If you got an A in this class, you must have gotten an A on
the nal.
3. p ∧ q⇒r
Solution to Exercise 2.5.7 (p. 52)
Poppins ⇒ Mary
Solution to Exercise 2.5.10 (p. 53)
1. There are many simple answers, such as Y − has − 1, ¬W − has − 1, . . .
2. There are many simple answers, such as a, N − has − 1, J − has − 3, . . .
For each, there are also many such formulas composed with connectives such as ∧ and ∨.
Solution to Exercise 2.5.16 (p. 55)
Truth table for xnor
φ ψ φxnorψ
false false true
false true false
true false false
true true true
Table 2.37
39 http://www.amnesty.org/
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
71
This is the equals for Booleans. It is also represented by the connective if-and-only-if (Example 2.1).
If you said something like the both-or-neither function, that's not quite good enough, as it's a round-
about way of expressing the simple idea of equivalence. Granted, it takes some practice to internalize
Booleans as values, and that equality is as valid for them as for any other value.
Solution to Exercise 2.5.23 (p. 56)
list merge_sorted_lists(list list1, list list2)
{
if (is_empty(list1))
return list2;
else if (is_empty(list2))
return list1;
else {
if (first_element(list1) < first_element(list2))
return make_list(first_element(list1),
merge_sorted_lists(rest_elements(list1),list2));
else
return make_list(first_element(list2),
merge_sorted_lists(list1,rest_elements(list2)));
}
}
Alternatively, we could test the emptiness of the lists in the other order.
Solution to Exercise 2.5.25 (p. 58)
First we show that we can write negation in terms of ↓, or more speci cally, ¬θ ≡ θ ↓ θ. Checking this on a
truth table is pretty easy (there are only two rows to check). But for this question we need to use algebraic
manipulation. This can be derived in a couple of simple steps:
1 ¬θ
2 ≡ ¬θ ∧ ¬θ Idempotency of ∧
3 ≡ ¬ (θ ∨ θ) DeMorgan's law
4 ≡θ↓θ De nition of nor
Table 2.38
We use this lemma to show our ultimate goal:
1 δ ∧
2 ≡ ¬¬ (δ ∧ ) Double Complementation
3 ≡ ¬ (¬δ ∨ ¬) DeMorgan's law
4 ≡ ¬ ((δ ↓ δ) ∨ ¬) Lemma, with [θ7→δ ]
5 ≡ ¬ ((δ ↓ δ) ∨ ( ↓ )) Lemma, with θ=
6 ≡δ↓δ↓↓ De nition of nor, where
φ = δ ↓ δ, and ψ=↓
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
72 CHAPTER 2. PROPOSITIONAL LOGIC
Table 2.39
Note that we judiciously used new meta-variables δ and rather than re-using φ and ψ (which would still
be correct, but would make the graders need to pay much closer attention to the scope of those variables).
Solution to Exercise 2.6.4 (p. 59)
1 ¬ (φ ∨ ψ) Premise
2 subproof:φ ` false
2.a φ Premise for subproof
2.b φ ∨ ψ ∨Intro, line 2a
2.c false falseIntro, lines 1,2b
3 ¬φ RAA, line 2
Table 2.40
Solution to Exercise 2.6.14 (p. 61)
1 Q − has − 1 Premise
2 X − has − 1 Premise
3 ¬Y − unsafe Premise
4 W − unsafe ∨ Y − unsafe Theorem: above
problem
(Exercise 2.6.13), line 2
5 Y − unsafe ∨ W − unsafe Theorem: ∨ commutes,
line 4
6 W − unsafe CaseElim, lines 3,5
7 subproof:¬¬ (P − safe ∧ W − safe) ` false
7.a ¬¬ (P − safe ∧ W − safe) Premise for subproof
7.b P − safe ∧ W − safe ¬Elim, line 7.a
7.c W − safe ∧Elim, line 7.b
7.d W − safe ⇒ WaterWorld axiom
¬W − unsafe
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
73
7.e ¬W − unsafe ⇒Elim, lines 7.c,7.d
7.f false falseIntro, lines 6,7.e
8 ¬ (P − safe ∧ W − safe) RAA, line 7
9 subproof:¬¬ (R − safe ∧ W − safe) ` false
9.a ¬¬ (R − safe ∧ W − safe) Premise for subproof
9.b R − safe ∧ W − safe ¬Elim, line 9.a
9.c W − safe ∧Elim, line 9.b
9.d W − safe ⇒ WaterWorld axiom
¬W − unsafe
9.e ¬W − unsafe ⇒Elim, lines 9.c,9.d
9.f false falseIntro, lines 6,9.e
10 ¬ (R − safe ∧ W − safe) RAA, line 9
11 Q − has − 1 ⇒ P − safe ∧ R − safe ∨ P − safe ∧ Theorem: Allowed by
W − safe ∨ R − safe ∧ W − safe problem statement
12 P − safe ∧ R − safe ∨ P − safe ∧ W − safe ∨ ⇒Elim, lines 1,11
R − safe ∧ W − safe
13 R − safe ∧ W − safe ∨ P − safe ∧ R − safe ∨ Theorem: ∨ commutes,
P − safe ∧ W − safe line 12
14 P − safe ∧ R − safe ∨ P − safe ∧ W − safe CaseElim, lines 8,13
15 P − safe ∧ W − safe ∨ P − safe ∧ R − safe Theorem: ∨ commutes,
line 14
16 P − safe ∧ R − safe CaseElim, lines 10,15
17 P − safe ∧Elim, line 16
Table 2.41
Alternatively, the subproofs could easily have been pulled out into lemmas. Just like using subroutines
in a program, that would make the proof somewhat clearer, even though in this case each lemma would be
used only once.
Observe how the two subproofs have some identical lines (7.c-7.f and 9.c-9.f ). It would be incorrect to
replace those lines in the second subproof with a citation of the results of the rst subproof. First, because
the previous subproof had been completed, and moreover, the two subproofs have dierent premises. This
is analogous to two subroutines that happen to have some identical code lines, even through they are called
separately and have dierent parameters.
note: Interestingly, we didn't need to use R − safe as a premise. (In fact, we nearly proved that
¬R − safe would have been inconsistent with the other premises.)
Solution to Exercise 2.6.16 (p. 63)
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
74 CHAPTER 2. PROPOSITIONAL LOGIC
1 ¬¬φ Premise
2 subproof:¬φ ` false
2.a ¬φ Premise for subproof
2.b false falseIntro, lines 1,2.a
3 φ RAA, line 2
Table 2.42
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
Chapter 3
Relations and Models
3.1 relations1
3.1.1 Relations: Building a better (representation of) WaterWorld
So far, we have represented WaterWorld boards using propositions like A − has − 2 and B − unsafe. You've
probably already felt that this is unwieldy, having hundreds propositional variables running around, with only
our naming convention implying any relation between them. Worse, this zoo of propositions doesn't re ect
how we actually think about WaterWorld. For instance, the only way the rules recognize that locations
A and B are near each other is because of several axioms which simultaneously involve A − has − 2 and
B − unsafe, etc., in just the right way to result in our idea of the concept neighbor. In fact, there is no way
of talking about the location A directly; we only had propositions which dealt with its properties, such as
whether or not it neighbored exactly two pirates.
If writing a program about WaterWorld, our program should re ect our conception of the problem. How-
ever, as it stands, our conception corresponds to having many many Boolean variables named A − has − 2,
B − unsafe, etc. Even worse, the rules would be encodings of the hundreds of axioms. A long enumeration
of the axioms is probably not how you think of the rules. In other words, when explaining the game to
your friend, you probably say if a location contains a 2, then two of its neighbors are pirates, rather than
droning on for half an hour about how if location A contains a 2, then either location B is unsafe or . . ..
Moreover, the original rules only pertained to a xed-size board; inventing a new game played on a 50×50
grid would require a whole new set of rules! That is clearly not how we humans conceptualize the game!
What we want, when discussing the rules, is a generic way to discussing neighboring locations, so that we
can have one single rule, saying that if a (generic) location has a zero, then any neighboring location is
safe. Thus, we allow the exact details of neighboring location to change from game to game as we play on
dierent boards (just as which locations contain pirates changes from game to game).
In a program, you'd probably represent the board as a collection (matrix, list, whatever) of Booleans. In
our logic, to correspond to this data structure, we'll introduce binary relations.
aside: By including relations (rather than sticking entirely with propositions), we are leaving the
realm of propositional logic; we'll soon reach rst-order logic once we also introduce quanti ers
(Section 4.1.1) corresponding to aspects of program control- ow (loops).
We'll start by adding a way to express whether any two locations are adjacent: a relation nhbr, which
will encode the board's geography as follows: nhbr (A, B) and nhbr (Z, Y ) are true, while nhbr (A, D) and
nhbr (M, Z) are false.
What, exactly, do we mean by relation? We'll see momentarily (Section 3.2), that we can represent
nhbr as a set of pairs-of-locations (or equivalently, a function which takes in two locations, and returns either
true or false.)
1 This content is available online at <http://cnx.org/content/m10724/2.25/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
75
76 CHAPTER 3. RELATIONS AND MODELS
This relation "nhbr" entirely encodes the board's geography. Giving somebody the relation is every bit as
good as to showing them a picture of the board (in some ways, better the relation makes it perfectly clear
whether two locations which just barely touch at a single point, like B and G, are meant to be considered
neighbors.)
Exercise 3.1.1 (Solution on p. 86.)
We used a binary (two-input) relation to describe neighboring locations. How can we use a relation
to capture the notion location A is safe?
After de ning relations and discussing their properties, we'll talk about interpreting logic formulas (Sec-
tion 3.3) relative to particular relations.
Using relations gives us additional exibility in modeling our domain, so that our formal logical model
more closely corresponds to our intuition. Relations help separate the WaterWorld domain axioms (code)
from the data, i.e., the particular board we're playing on.
3.2 properties of relations2
When using relations in logic formulas, there are two things going on:
• relations themselves, as mathematical entities, and
• formulas involving symbols, which must be interpreted as speci c relations.
First things rst: we'll just discuss relations for now, and later tackle using relations in logic formulas.
We'll start with a couple of equivalent ways of de ning relations, and then discuss a common subclass of
relations: binary relations.
3.2.1 Relations as subsets
Consider the set of WaterWorld locations Loc = {A, B, . . ., Z}. For this domain (also known as a universe),
we'll say a binary relation is a set of (ordered) pairs of the domain.
Example 3.1
For instance, the nhbr relation of the previous section is the set
{(A, B) , (A, G) , (B, A) , (B, C) , . . ., (Y, X) , (Y, Z) , (Z, Y )}.
That is, x is related to y if (x, y) is in the set nhbr.
Example 3.2
For the domain D = {Object, String, MutableString}, the relation subclass − of might be
{(String, Object) , (MutableString, Object) , (MutableString, String)}.
In general, a binary relation over the domain D is a subset of D×D . Note that these are ordered
pairs; just because x is related to y doesn't mean y has the same relation to x. For example,
while (MutableString, Object) is in the relation subclass − of , the pair (Object, MutableString)
most certainly is not.
Example 3.3
You can consider the relation hasStarredWith, over the domain of Hollywood actors. We won't
list all the elements of the relation, but some related pairs are:
• hasStarredWith (Ewan McGregor, Cameron Diaz), as witnessed by the movie A Life Less Or-
dinary, 1997.
• hasStarredWith (Cameron Diaz, John Cusack), as witnessed by the movie Being John
Malkovich, 1999.
2 This content is available online at <http://cnx.org/content/m10725/2.28/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
77
If binary relations are subsets of pairs of the domain, what might a unary relation be? Simply, subsets of
the domain.
Example 3.4
For the domain of vegetables, Ian de nes the relation yummy? as
{tomatoes, okra, cucumbers, carrots, potatoes} and nothing else.
Example 3.5
In one particular game of WaterWorld, the relation hasPirate turned out to be {K, T, R, U, E}.
If unary and binary relations make sense, what about ternary, etc., relations? Sure! In general, a k -ary
relation (or, relation of arity k ) over the domain D is a subset of Dk . However, any given relation has a
xed arity. That is, a relation may be binary or ternary, but not both.
As with propositions, rather than writing R (x, y) is true , we'll simply write R (x, y) . In fact, notice
that once you choose some particular pair of x and y , then R (x, y) can be treated as a single true/false
proposition. (We'll soon extend the idea of propositions to include such relation symbols, and then allow
formulas to include these terms.)
Example 3.6
prime (18) is a proposition that's false, assuming the standard interpretation (Section 3.3) of
prime.
Example 3.7
safe (A) is a proposition that is true on some boards and not others.
3.2.2 Relations as functions
The relation nhbr, which we're de ning as a set (of pairs), could also be thought of being a function. We
say that the indicator function of a set is a Boolean function indicating whether its input is in the set
or not. So instead of being given the set nhbr, you would have been equally happy with its indicator
function fnhbr , where (for example) fnhbr (B, C) = true and fnhbr (B, Q) = false. Similarly, if you know
that fhasPirate (K) = true and that fhasPirate (L) = false, then this is enough information to conclude that
K ∈ hasPirate and L ∈ / hasPirate. The set and the function are equivalent ways of modeling the
same underlying relation.
The next two exercises aren't meant to be di cult, but rather to illustrate that, while we've sketched
these two approaches and suggested they are equivalent, we still need an exact de nition.
Exercise 3.2.1 (Solution on p. 86.)
true 2
if y=x
For the indicator function f (x, y) = on the domain of (pairs of ) natural
false otherwise
numbers, write down the set-of-pairs representation for the corresponding binary relation. It's
insightful to give the answer both by listing the elements, possibly with ellipses, and also by using
set-builder notation.
In general, for a binary indicator function f, what, exactly, is the corresponding set?
Exercise 3.2.2 (Solution on p. 86.)
For the relation hasPirate = {K, T, R, U, E} on the set of (individual) WaterWorld locations, write
down the indicator-function representation for the corresponding unary relation. In general, how
would you write down this translation?
Since these two formulations of a relation, sets and indicator functions, are so close, we'll often switch
between them (a very slight abuse of terminology).
Think about when you write a program that uses the abstract data type Set. Its main operation is
elementOf. When might you use an explicit enumeration to encode a set, and when an indicator function?
Which would you use for the set of WaterWorld locations? Which for the set of prime numbers?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
78 CHAPTER 3. RELATIONS AND MODELS
3.2.3 Functions as Relations
Some binary relations have a special property: each element of the domain occurs as the rst item in exactly
one tuple. For example, isPlanet = {(Earth, true) , (Venus, true) , (Sol, false) , (Ceres, false) , (Mars, true)}
is actually a (unary) function. On the other hand, isTheSquareOf =
{(0, 0) , (1, 1) , (1, −1) , (4, 2) , (4, −2) , (9, 3) , (9, −3) , . . .} is not a function, for two reasons. First, some
numbers occur as the rst element of multiple pairs. Second, some numbers, like 3, occurs as the rst
element of no pairs.
We can generalize this to relations of higher arity, also. This is explored in this exercise (Exercise 4.4.2)
and this one (Exercise 4.4.3).
3.2.4 Binary Relations
One subclass of relations are common enough to merit some special discussion: binary relations. These are
relations on pairs, like nhbr.
3.2.4.1 Binary Relation Notation
Although we introduced relations with pre x notation, e.g., < (i, j), we'll use the more common in x nota-
tion, i < j, for well-known arithmetic binary relations.
3.2.4.2 Binary Relations as Graphs
In fact, binary relations are common enough that sometimes people use some entirely new vocabulary: A
domain with a binary relation can be called vertices with edges between them. Together this is known as
a graph. We won't stress these terms right now, as we're not studying graph theory.
Binary relations (graphs) can be depicted visually, by drawing the domain elements (vertices) as dots,
and drawing arrows (edges) between related elements.
A binary relation with a whole website devoted to it is has starred in a movie with. We'll call this
relation hasStarredWith over the domain of actors. Some sample points in this relation:
• hasStarredWith (Ewan McGregor, Cameron Diaz), as witnessed by the movie A Life Less Ordinary,
1997.
• hasStarredWith (Cameron Diaz, John Cusack), as witnessed by the movie Being John Malkovich, 1999.
You can think of each actor being a location, and two actors being adjacent to each other if they have
ever starred in a movie together; two of these locations, even if not adjacent might have a multi-step path
between them. (There is also a shorter path; can you think of it? The (in)famous Kevin Bacon game asks to
nd a shortest path from one location to the location Kevin Bacon. Make a guess, as to the longest shortest
path leading from (some obscure) location to Kevin Bacon.)
Some other graphs:
• Vertices can be tasks, with edges meaning dependencies of what must be done rst.
• In parallel processing, Vertices can be lines of code; there is an edge between two lines if they involve
common variables. Finding subsets of vertices with no lines between them represent sets of instructions
that can be executed in parallel (and thus assigned to dierent processors.)
• Word ladders seek to transform one word to another by changing one letter at a time, while always
remaining a word. For example, a ladder leading from WHITE to SPINE in three steps is:
· WHITE
· WHINE
· SHINE
· SPINE
If a solution to such a puzzle corresponds to a path, what do vertices represent? What are edges? Do
you think there is a path from any 5-letter word to another?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
79
3.3 interpretations3
3.3.1 Needing Interpretations to Evaluate Formulas
You might have noticed something funny: we said safe (a) depended on the board, but that prime (18) was
false. Why are some some relations dierent than others? To add to the puzzling, there was a caveat in some
ne-print from the previous section: prime (18) is false under the standard interpretation of prime
. Why these weasel-words? Everybody knows what prime is, don't they? Well, if our domain is matrices of
integers (instead of just integers), we might suddenly want a dierent idea prime.
Consider the formula E (x, x) true for all x in a domain? Well, it depends not only on the domain, but
also on the speci c binary relation E actually stands for:
• for the domain of integers where E is interpreted as both are even numbers, E (x, x) is false for some
x.
• for the domain {2, 4, 6, 8} where E is interpreted as sum to an even number, E (x, x) is true for every
x.
• for the domain of integers where E is interpreted as greater than, E (x, x) is false for some x (indeed,
it's false for every x).
• for the domain of people where E is interpreted as is at least as tall as, E (x, x) is true for every x.
Thus a formula's truth depends on the interpretation of the (syntactic, meaning-free) relation symbols
in the formula.
De nition 3.1: Interpretation
The interpretation of a formula is a domain, together with a mapping from the formula's relation
symbols to speci c relations on the domain.
One analogy is Programs are to data, as formulas are to interpretations . (In particular, the formula
is a like a boolean function: it takes its input (interpretation), and returns true or false.)
3.3.1.1 Using Truth Tables to Summarize Interpretations (Optional)
Consider the formula ϕ = R (x, y) ⇒ S (x, y) ∧ ¬T (x, y). As yet, we haven't said anything about the
interpretations of these three relations. But, we do know that each of R (x, y), S (x, y), and T (x, y) can
either be true or false. Thus, treating each of those as a proposition, we can describe the formula's truth
under dierent interpretations.
R (x, y) S (x, y) R (x, y) [U+03D5]
false false false true
false false true true
false true false true
false true true true
true false false false
true false true false
true true false true
true true true false
Table 3.1
3 This content is available online at <http://cnx.org/content/m10726/2.23/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
80 CHAPTER 3. RELATIONS AND MODELS
3.3.1.2 Using Formulas to Classify Interpretations (Optional)
In the previous section, having a formula was rather useless until we had a particular interpretation for it.
But we can view that same idea backwards: Given a formula, what are all the interpretations for which the
formula is true?
For instance, consider a formula expressing that an array is sorted ascendingly: For all numbers i,j ,
(i < j) ⇒ (element (i) ≤ element (j)). But if we now broaden our mind about what relations/functions the
symbols element, < , and ≤ represent and then wonder about the set of all structures/interpretations which
make this formula true, we might nd that our notion of sorting is broader than we rst thought. Or
equivalently, we might decide that the notion ascending applies to more structures than we rst suspected.
Similarly, mathematicians create some formulas about functions being associative, having an identity
element, and such, and then look at all structures which have those properties; this is how they de ne
notions such as groups, rings, elds, and algebras.
3.3.1.3 Encoding Functions as Relations
What about adding functions, to our language, in addition to relations? Well, functions are just a way of
relating input(s) to an output. For example, 3 and 9 are related by the square function, as are 9 and 81, and
0,0. Is any binary relation a function? No, for instance {(9, 81) , (9, 17)} is not a function, because there is
no unique output related to the input 9.
How can we enforce uniqueness? The following sentence asserts that for each element x of the domain,
R associates at most one value with x: For all x, y and z of the domain,
R (x, y) ∧ R (x, z) ⇒ (y = z) (3.1)
This is a common trick, for to describe uniqueness: if y and z each have some property, then they must
be equal. (We have not yet speci ed that for every element of the domain, there is at least one element
associated with it; we'll get to that later.)
Exercise 3.3.1 (Solution on p. 86.)
We just used a binary relation to model a unary function. Carry on this idea, by using a ternary
relation to start to model a binary function. In particular, write a formula stating that for every
pair of elements w, x in the domain, the relation S associates at most one value with that pair.
3.4 Nonstandard Interpretations (optional)4
3.4.1 Prime factorization
Note that there are other possible interpretations of prime. For example, since one can multiply integer
matrices, there might be a useful concept of prime matrices.
For example: Consider only the numbers F = {1, 5, 9, 13, . . . } that is, F = { k, 4k + 1 | k ∈ N }. It's
easy to verify that multiplying two of these numbers still results in a number of the form 4k + 1. Thus it
makes sense to talk of factoring such numbers: We'd say that 45 factors into 59, but 9 is considered prime
since it doesn't factor into smaller elements of F.
Interestingly, within F , we lose unique factorization: 441 = 9 × 49 = 21 × 21, where each of 9, 21, and 49
are prime, relative to F! (Mathematicians will then go and look for exactly what property of a multiplication
function are needed, to guarantee unique factorization.)
The point is, that all relations in logical formula need to be interpreted. Usually, for numbers, we use a
standard interpretation, but one can consider those formulas in dierent, non-standard interpretations!
4 This content is available online at <http://cnx.org/content/m12741/1.3/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
81
3.4.2 The Poincaré Disc
A long outstanding problem was that of Euclid's parallel postulate: Given a line and a point not on the
line, how many lines parallel to the rst go through that point? Euclid took this as an axiom (unable to
prove that it followed from his other axioms). Non-Euclidean geometries of Lobachevsky and Riemann took
dierent postulates, and got dierent geometries. However, it was not clear whether these geometries were
sound whether one could derive two dierent results that were inconsistent with each other.
Henri Poincaré developed an ingenious method for showing that certain non-Euclidean geometries are
consistent or at least, as consistent as Euclidean geometry. Remember that in Euclidean geometry, the
concepts point and line are left unde ned, and axioms are built on top of them (e.g., two dierent lines
have at most one point in common). While it's usually left to common sense to interpret point, line, and
a point is on a line, any interpretation which satis es the axioms means that all theorems of geometry will
hold.
The Poincaré disc is one such interpretation: point is taken to mean a point in the interior of the unit
disc, and line is taken to mean a circular arc which meets the unit disc at right angles. So a statement
like two points determine a line can be interpreted as [*] For any two points inside the disc, there is exactly
one circular arc which meets the disc at right angles. Indeed, this interpretation
5 preserves all of Euclid's
postulates except for the parallel postulate. You can see that for a given line and a point not on it, there
are an in nite number of parallel (that is, non-intersecting) lines.
Figure 3.1: Some lines in the Poincaré disc, including several lines parallel to a line L through a point
p.
5 http://mcs.open.ac.uk/tcl2/nonE/CABRI2001/PDiscMod.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
82 CHAPTER 3. RELATIONS AND MODELS
(Note that the distance function is very dierent within the Poincaré disc; in fact the perimeter of the
disc is o at in nity. Angles, however, do happen to be preserved.)
The critical point of his interpretation of a non-Euclidean geometry is this: it is embedded in Eu-
clidean geometry! So we are able to prove (within the embedding Euclidean geometry) that the disc-
postulates hold (e.g., we can prove the statement [*] above as a theorem about circular arcs in Euclidean
geometry). Therefore, if there is any inconsistency in non-Euclidean geometry, then that could be parlayed
into some inconsistency of Euclidean geometry. Thus, his interpretation gives a proof that the strange
non-Euclidean geometry is as sound as our familiar Euclidean geometry.
3.4.3 P vs. NP and Oracles
A well-known problem in computer science P vs. NP asks whether (for a given problem) it is truly
more di cult to nd a short solution (when one exists) (NP), than it is to verify a short purported solution
handed to you (P). For example, Given a set of people and how strong person is, can you partition them
into two tug-of-war teams which are exactly evenly matched? Certainly it seems easier to check that a pair
of proposed rosters has equal strength (and, verify that everybody really is on one team or the other) than to
have to come up with two perfectly-matched teams. But conceivably, the two tasks might be equally-di cult
up to some acceptable (polynomial time) overhead. While every assumes that P is easier than NP, nobody
has been able to prove it.
An interesting variant of the problem lets both the problem-solver and the purported-answer-veri er
each have access to a particular oracle a program that will gives instant yes/no answers to some other
problem (say, given any set of numbers, yes or no: is there an even-sized subset whose total is exactly the
same as some odd sized subset?).
It has been shown that there is some oracle which makes the problem-solver's job provably tougher than
the proof-veri er's job, and also there is some other oracle problem-solver's job provably no-tougher than
the proof-veri er's job.
This means that any proof of P being dierent from NP has to be subtle enough so that when P and
NP are re-interpreted as P and NP with respect to a particular oracle, the proof will no longer go through.
Unfortunately, this eliminates all the routine methods of proof; we know that solving this problem will take
some new attack.
3.4.4 Löwenheim-Skolem and the real numbers
The Löwenheim-Skolem theorem of logic states that if a set of (countable) domain axioms has a model at
all, then it has a countable model. This is a bit surprising when applied to the axioms of arithmetic for the
real numbers: even though the real numbers are uncountable, there is some countable model which meets
all our ( nite) axioms of the real numbers!
3.4.5 Object-oriented programming
Note that object-oriented programming is founded on the possibility for nonstandard interpretations: perhaps
you have some code which is given a list of Objects, and you proceed to call the method toString on each
of them. Certainly there is a standard interpretation for the function Object.toString, but your code is
built to work even when you call this function and some nonstandard, custom, overridden method is called
instead.
It can become very di cult to reason about programs when the run-time method invoked might be
dierent from the one being called. We're used to specifying type constratins which any interpretation must
satisfy; wouldn't it be nice to specify more complicated constraints, e.g. this function returns an int which
is a valid index into [some array]? And if we can describe the constraint formally (rather than in English
comments, which is how most code works), then we could have the computer enforce that contract! (for
every interpretation which gets executed, including non-static ones).
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
83
An obvious formal speci cation language is code itself have code which veri es pre-conditions before
calling a function, and then runs code verifying the post-condition before leaving the function. Indeed, there
are several such tools about (Java
6 , Scheme7 ). In the presence of inheritance, it's harder than you might
initially think to do this correctly
8 .
It is still a research goal to be able to (sometimes) optimize away such run-time veri cations; this requires
proving that some code is correct (at least, with respect to its post-condition). The fact that the code might
call a function which will be later overridden (our non-standard interpretations) exacerbates this di culty.
(And proving correctness in the presence of concurrency
9 is even tougher!)
Even if not proving programs correct, being able to specify contracts in a formal language (code or logic)
is a valuable skill.
3.4.6 Real-World Arguments
Finally, it is worth noting that many rebuttles of real world arguments (Exercise 1.3.1) (see also some
exercises (Exercise 2.5.1)) amount to showing that the argument's form can't be valid since it doesn't hold
under other interpretations, and thus there must be some unstated assumptions in the original.
3.5 modeling with relations10
3.5.1 Modeling with Relations
aside: Note that the nhbr relation can actually represent an arbitrarily weird board, such as
locations that look adjacent on the map but actually aren't, boards which wrap around a cylinder
11
or toroid
12 , or a location with a tunnel connecting it to a location far across the board (like the
secret passages in the game Clue, or the harrowing sub trip through Naboo in Star Wars: The
Phantom Menace
13 .) One-way passages can be encoded as well (meaning the nhbr relation need
not be symmetric). Actually, any graph can be represented!
Exercise 3.5.1 (Solution on p. 86.)
How shall we encode concepts such as location A has 3 dangerous neighbors , using relations?
Proofs otherwise unchanged. Note that we might express our rules as for any locations x and y, we have
the following axiom: has − 3 (x) ∧ nhbr (x, y) ⇒ ¬safe (y) . Really, note that there's something else going
on here: x and y are symbols which can represent any location: they are variables, whose value can be any
element of the domain.
For the domain of types-of-vegetables, the relation yummy is a useful one to know, when cooking. In
yummy (Brussels sprouts) = false, and yummy (carrots) = true.
case you weren't sure,
Suppose we had a second relation, yucky. Is it conceivable that we could model a vegetable that's neither
yucky nor yummy, using these relations? Sure! (Iceberg lettuce, perhaps.) In fact, we could even have a
vegetable which is both yummy and yucky radishes!
aside: A quick digression on a philosophical nuance: the domain for the above problem is not
vegetables; it's types-of-vegetables. That is, we talk about whether or not carrots are yummy;
this is dierent than talking the yumminess of the carrot I dropped under the couch yesterday, or
the carrot underneath the chocolate sauce. In computer science, this often manifests itself as the
6 http://www.javaworld.com/javaworld/jw-02-2001/jw-0216-cooltools.html
7 http://download.plt-scheme.org/doc/209/html/mzlib/mzlib-Z-H-13.html#node_chap_13
8 http://people.cs.uchicago.edu/∼robby/pubs/index-abstracts.html#9
9 "Concurrent Processes: Basic Issues" <http://cnx.org/content/m12312/latest/>
10 This content is available online at <http://cnx.org/content/m10727/2.24/>.
11 http://hades.ph.tn.tudelft.nl/Internal/PHServices/Documentation/MathWorld/math/math/c/c904.htm
12 http://hades.ph.tn.tudelft.nl/Internal/PHServices/Documentation/MathWorld/math/math/t/t188.htm
13 http://www.starwars.com/episode-i/
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
84 CHAPTER 3. RELATIONS AND MODELS
dierence between values, and types of values. As examples, we distinguish between 3 and the set of
all integers, and we distinguish between particular carrots and the abstract idea of carrots. (Some
languages even include types as values.) Philosophers enjoy debating how particular instances
de ne the abstract generalization, but for our purposes we'll take each both vegetables and types-
of-vegetables as given.
Exercise 3.5.2 (Solution on p. 86.)
You might have objected to the idea of the unary relation yummy, since dierent people have
dierent tastes. How could you model individuals' tastes? (Hint: Use a binary relation.) )
Modeling actors and the has-starred-with relation didn't include information about speci c movies. For
instance, it was impossible to write any formula which could capture the notion of three actors all being in
the same movie.
Exercise 3.5.3 (Solution on p. 86.)
Why doesn't hasStarredWith (a, b) ∧ hasStarredWith (b, c) ∧ hasStarredWith (c, a) capture the
notion of a, b, and c all being in the same movie? Prove your answer by giving a counterexample.
Exercise 3.5.4 (Solution on p. 87.)
How might we make a model which does capture this? What is the domain? What relations do
you want?
Of course, the notion of interpretations are still with us, though usually everybody wants to be thinking
of one standard interpretation. Consider a relation with elements such as isChildOf (Bart, Homer, Marge).
Would the triple (Bart, Marge, Homer) be in the relation as well as (Bart, Homer, Marge)?
As long as all the writers and users of formulas involving isChildOf all agree on what the intended
interpretation is, either convention can be used.
3.5.2 A Case Study: iTunes
Consider iTunes' smart playlists: you can create a playlist consisting of (say)
All songs I've rated 3-stars or better, and whose genre is not Classical
. This is smart because its a program which is re-run every time your music library changes: For example,
if you change a song's genre, it may be immediately added or deleted from the playlist. We realize actually
have a simple formula (which we can express in propositional logic with relations). The structure (instance)
for a single is the interpretation. This formula is true when interpreted on (my library's representation of )
Brian Eno's Here Come the Warm Jets, but false for Bonnie Tyler's '80's epic Holding Out for a Hero
and for Bach's Little Fugue in Gm . We now have one formula, and want to determine its truth-value
in many dierent particular interpretations. In fact, we want to return all interpretations which make the
formula (playlist) true.
Exercise 3.5.5 (Solution on p. 87.)
Look at the GUI box for de ning these queries. Compared to propositional logic, what sort of
formulas can you de ne?
Exercise 3.5.6 (Solution on p. 87.)
Are there queries which can't be directly transliterated into the GUI box?
14
Exercise 3.5.7 (Solution on p. 87.)
Can you nd formulas equivalent to each of the preceding two, which can be expressed?
The upshot is that iTunes came up with a query language which is as expressive as propositional
15 logic.
For some queries, it can be awkward to use, but the GUI designers who came up with smart playlists might
have gured that few users would want such queries.
14 Transliterate meaning a word-for-word substitution, while translate preserves meanings and idioms. So while the German
Übung macht den Meister transliterates to Drill makes the master, it translates to Practice makes perfect.
15 Technically, this is a relational calculus formula, since we are using relations instead of at propositions.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
85
note: How might you create a GUI widget which can specify any propositional formula, and yet
still look nice and be intuitive enough for my mother to use? Is there a better usability/expressibility
trade-o than what iTunes has done, or are they optimal?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
86 CHAPTER 3. RELATIONS AND MODELS
Solutions to Exercises in Chapter 3
Solution to Exercise 3.1.1 (p. 76)
We'll use a unary (one-input) relation: safe(A) is true if and only if (i ) location A is safe.
Solution
to Exercise 3.2.1 (p. 77)
(0, 0) , (1, 1) , (2, 4) , (3, 9) , . . ., i, i2 , . . . In set-builder notation, this is (x, y) | y = x2
In general, for an indicator function f , the corresponding set would be { (x, y) | f (x, y) } (Note that we
don't need to write . . . f (x, y) = true ; as computer scientists comfortable with Booleans as values, we
see this is redundant.)
Solution to Exercise 3.2.2 (p. 77)
true if x=K
true if x=T
true if x=R
fhasPirate (x) = .
true
if x=U
true if x=E
false
otherwise
true if x∈R
In general, for a (unary) relation R, fR (x) = .
false if x∈
/R
Solution to Exercise 3.3.1 (p. 80)
For all w , x, y , and z of the domain,
S (w, x, y) ∧ S (w, x, z) ⇒ (y = z) (3.2)
Solution to Exercise 3.5.1 (p. 83)
A good rst guess might be to say we have a function which returns the number of pirates next to a given
location. That is, piratesNear (A) = 3 . However, piratesNear doesn't qualify as a relation. Why not?
To work around this, we could propose a binary relation along the lines of piratesNear (A, 3) = true .
This is better, but it requires our domain to be not only board locations, but also numbers. And to be able
to talk about numbers, we'd need more axioms, as well as numeric relations such as >.
aside: Hmmm, what is the arity of >?
While this approach is feasible, and ultimately might be what we want, for now, let's stick with relations
involving only locations, not numbers.
Okay, the third time's the charm: we'll implement the concept A neighbors three pirates as a relation
has − 3 (A) being true. To cover the cases when there are exactly two neighboring pirates, we'll use a whole
new separate relation, has − 2; has − 2 (A) would be false on any board where has − 3 (A) is true (at least,
in our standard interpretation).
Solution to Exercise 3.5.2 (p. 84)
We can use the binary relation thinksIsYummy: In particular, thinksIsYummy (Ian, anchovies) = false but
thinksIsYummy (Phokion, anchovies) = true What set are we using, as the domain for this? Really, the
domain is the union of people and pizza-toppings. So thinksIsYummy (radishes, brusselsSprouts) is a valid
thing to write down; it would be false. Note that if working with such a domain, having unary predicates
isVegetable and isPerson would be useful.
Solution to Exercise 3.5.3 (p. 84)
The proposed formula asserts that each pair has been in some movie together, but they each
could have been dierent movies without being in the same one simultaneously. As a counterexam-
ple, it is true that hasStarredWith (Charlie Chaplin, Norman Lloyd) (as witnessed by Limelight, 1952),
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
87
hasStarredWith (Norman Lloyd, Janeane Garofolo) (as witnessed by The Adventures of Rocky and Bullwin-
kle, 2000), and if we generously include archive footage, hasStarredWith (Charlie Chaplin, Janeane Garofolo)
(as witnessed by Outlaw Comic: The Censoring of Bill Hicks, 2003); however, they have not all been in a
movie together. Might the counterexample you chose become nulli ed, in the future?
Solution to Exercise 3.5.4 (p. 84)
As always, there are several ways of modeling this problem. We'll outline three.
First, we could augment the hasStarredWith to be a ternary (3-input) relation to include the movie. Like
in the yummy extension (Exercise 3.5.2), the domain would then include both actors and movies, and we'd
also want relations to know which is which.
Second, we could use a bunch of relations. Starting with the familiar binary hasStarredWith, we'd
add the ternary hasStarredWith3, the quaternary hasStarredWith4, . . .. Our domain would just be actors.
However, we'd either need an in nite number of such relations, which we normally don't allow, or we'd need
an arbitrary cap on the number of people we're interested in at a time.
Third, we could use sets of actors, instead of individuals. We'd need only one relation,
haveStarredtogether, that states a set of actors have starred together in a single movie.
Solution to Exercise 3.5.5 (p. 84)
This is eectively Disjunctive or Conjunctive Normal Form, limited to clauses of one term each.
Solution to Exercise 3.5.6 (p. 84)
Yes. Two examples are ¬ ((genre = Classical) ∨ (genre = Holiday)), and (genre = Rock) ∧
((Rating ≥ 4) ∨ (genre = Classical)).
Solution to Exercise 3.5.7 (p. 84)
For the rst example, ¬ ((genre = Classical) ∨ (genre = Holiday)), we can clearly use DeMorgan's law and
make the query ¬ (genre = Classical) ∧ ¬ (genre = Holiday).
However, for (genre = Rock) ∧ ((Rating ≥ 4) ∨ (genre = Classical)) there is no equivalent one-term-
per-clause DNF or CNF formula!
16
Fortunately, iTunes has a way around this. Playlist membership or non-membership is itself an available
predicate, allowing you to nest playlists. Thus, you can build a playlistGoodOrClassical for (Rating ≥ 4) ∨
(genre = Classical), then another (genre = Rock) ∧ GoodOrClassical for the desired result.
16 Budding logicians might wonder how you actually prove this claim!
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
88 CHAPTER 3. RELATIONS AND MODELS
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
Chapter 4
First-Order Logic
4.1 A formal vocabulary
4.1.1 syntax and semantics of quanti ers1
4.1.1.1 Talking about unnamed items
Suppose we want to express a statement like there is a location which has two neighbors (which is true,
at least for the domain of WaterWorld board locations), or all actors have co-starred with Kevin Bacon
2
(which isn't true, at least for the domain of all Hollywood actors). As it stands, we can formulate these
only awkwardly, by talking about speci c (constant) locations like A and G, or speci c actors like Ewan
McGregor
3 and Cameron Diaz4 . To talk about all locations, or actors, we're forced to make huge formulas
such as nhbr (Z, Y ) ∧ ¬nhbr (Z, A) ∧ ¬nhbr (Z, B) ∧ . . . ∧ ¬nhbr (Z, X), just to express there is a
location which has only one neighbor.
We'll redress this by introducing two quanti ers, ∃ (there exists) and ∀ (for all). For example, all
actors have co-starred with Kevin Bacon will be written ∀a : (coStarredWith (a, Kevin Bacon)). For there
is a location which has (at least) two neighbors , we'll start with there exists a location x . . . , written
∃x : (. . . ).
For all is really just an abbreviation for a large conjunction, while exists is a disjunction (it could also
be called for some, though it's not). How large a conjunction/disjunction? As big as your domain, which
actually could be very small, or it could be in nitely large. Even aside from the fact that we can't write
down an in nitely large conjunction or disjunction, quanti ers let us form the conjunction without having
to select a domain in advance.
To continue with our WaterWorld example, how can we express the concept x has (at least) two neigh-
y and z , which each of which is a neighbor
bors? Well, we'll rephrase this as, there exist distinct locations,
of x ∃x : (∃y : (∃z : ((y 6= z) ∧ nhbr (x, y) ∧ nhbr (x, z)))). We need the condition ¬ (y = z) in
, written
that formula to ensure that we have distinct locations. Compare to the algebraic equation x + y = 4 in which
one possible solution is x = y = 2. Variables act the same way in both logic and algebra: dierent variables
can happen to take on the same value.
We use quanti ers all the time in natural language. Consider the following examples, where we provide
a natural English wording together with an equivalent phrasing that makes the quanti cation more explicit.
We'll take the translations with a grain of salt, since sometimes people can disagree on the exact details of
the intended English meaning. Such ambiguity can sometimes be a rich source of creativity, but it's not
tolerable when documenting safety properties of software. While some of these examples are a bit frivolous,
1 This content is available online at <http://cnx.org/content/m10728/2.35/>.
2 http://us.imdb.com/Name?Bacon,+Kevin
3 http://us.imdb.com/Name?McGregor,+Ewan
4 http://us.imdb.com/Name?Diaz,+Cameron
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
89
90 CHAPTER 4. FIRST-ORDER LOGIC
in general quanti ers let us precisely capture more interesting concepts in type-checking, data structures
such as trees and hash tables, circuit speci cations, etc.
Quanti cation in English
Natural English Formalized English
If you don't love yourself, you can't love anybody If you don't love you, there does not exists a per-
else. son y, such that you love y.
N*Sync is the best band ever! For all bands x, N*Sync is better than band x (or,
x = N ∗ Sync). A quick listen can easily show this
statement false.
A casually subtle line from Something About Mary: For all days x, x is better than next(x).
Every day is better than the next.
A buggy line from a song (Everybody Loves My For all persons x, x loves my baby. For all persons
Baby, Jack Palmer and Spencer Willson, 1924): y, if my baby loves y , then y is me. If true, one
Everybody loves my baby; My baby don't love can conclude the speaker is his own baby, and is
[anybody] but me. narcissistic.
Every neighbor of x is unsafe. For all locations y, if y is a neighbor of x, then y
is unsafe.
There is a safe location that is a neighbor of x, if If num(x)<3, then there exists a location y, such
num(x)<3. that y is safe, and y is a neighbor of x.
If you've seen one episode, you've seen 'em all. If there exists one episode x such that you've seen
x, then for all episodes z, you've seen z .
Somebody loves everybody. There exists some person y, such that for all per-
sons x, y loves x.
There is someone for everybody. For all persons x, there exists a person y , such that
y is for x.
All's well that ends well. For all events x, if x ends well then x is well.
Table 4.1
4.1.1.1.1 Warning: The Ambiguous Any
The ambiguous any: I was playing a game with some friends, and we came across the rule: If you have
more cards than any other player, then discard a card. Does this mean than all other players, or than
some other player? Our group's opinion was divided (incl. across many native English speakers).
In our class terms, it's not always clear whether any means for-all, or for-some (there-exists). Or maybe
more accurately, in the phrase for any x, does x necessarily mean an arbitrary (p. 99) player?
aside: Linguistics students, or those who are so sure the rule clearly intended than all other
players: Switching x>y to x<y changes from an active voice to a passive voice but may
also reverse your interpretation of the English quanti er any: If any player has fewer points than
you, . . .
In your proof-writing (and your English writing, and your informal writing), think about replacing any
with either every or with some, to make your meaning clear.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
91
4.1.1.2 First-order logic: WFFs revisited
We originally de ned a well-formed formula (WFF) for propositional logic; we'll extend this to WFFs for
rst-order logic, also known as predicate logic. At the same time, we'll more precisely de ne the binding
of variables.
This logic allows use of both functions and relations. Since these functions' outputs are not Booleans
(otherwise, we'd call them relations), but rather data than can be used as a relation's input, we separate the
syntax into that of terms and formulas. Terms are all the possible inputs for a relation.
De nition 4.1: term
1. A variable.
Example
a , b, . . .
2. A constant.
Example
WaterWorld location F, Kevin Bacon, or the number 3.
3. A function applied to one or more terms.
Example
successor (3)
De nition 4.2: Well-Formed Formula (WFF) for rst-order logic
1. A constant: true or false.
atomic formula: a relation symbol applied to one or more terms.
2. An
Example
nhbr (x, F )
3. A negation of a WFF, ¬φ.
4. A conjunction of WFFs, φ ∧ ψ.
5. A disjunction of WFFs, φ ∨ ψ .
6. An implication of WFFs, φ ⇒ ψ .
universal quanti cation of a WFF,
7. A ∀x : (φ).
Example
∀x : (nhbr (x, F ))
existential quanti cation of a WFF,
8. An ∃x : (φ).
Example
∃x : (nhbr (x, F ))
While a formula is just a piece of syntax, the meaning of its connectives, including the quanti ers, is part
of the de nition of a WFF. However, as previously discussed, the meaning of a WFF also depends on the
interpretation (Section 3.3) we give to its relations.
4.1.1.2.1 Examples
Example 4.1
Everybody likes John Cusack: ∀x : (likes (x, John Cusack)).
Example 4.2
Somebody likes Joan Cusack: ∃x : (likes (x, Joan Cusack)).
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
92 CHAPTER 4. FIRST-ORDER LOGIC
Example 4.3
Somebody likes everybody: ∃x : (∀y : (likes (x, y))). (We use n for needy?)
Example 4.4
Everybody likes somebody: ∀y : (∃x : (likes (y, x))). Careful; this formula looks similar to the
preceding one, but it has a very dierent meaning!
Exercise 4.1.1.1 (Solution on p. 110.)
How would you express Somebody is liked by everybody?
Exercise 4.1.1.2 (Solution on p. 110.)
How would you express Everybody is liked by somebody?
Example 4.5
The following formula is a simple application of symmetry. ∀x : (∀y : (near (x, y) ⇒ near (y, x))) ∧
near (Sue, Joe) ⇒ near (Joe, Sue).
While it is certainly true under the intended interpretation, it is also true under any inter-
pretation. Such formulas are called valid. Valid rst-order formulas are the natural analog of
tautological propositional formulas.
Example 4.6
∀x : (even (x) ∧ prime (x) ⇒ (x = 2)) is a mathematical fact, in the standard interpretation of
arithmetic.
While technically not allowed by our term (De nition: "term", p. 91) and formula (De nition: "Well-
Formed Formula (WFF) for rst-order logic", p. 91) syntax, we'll continue using in x notation for common
mathematical functions and relations, as in the previous example (Example 4.6).
Exercise 4.1.1.3 (Solution on p. 110.)
The previous example (Example 4.6) used the relations even and prime. Of course, to use such
relations, they must either be de ned directly by the interpretation, or be de ned in terms of
functions and relations provided by the interpretation.
How would you de ne these two relations in terms of the basic numerical functions (addition,
multiplication, . . .) and relations (= , <, >)?
Exercise 4.1.1.4 (Solution on p. 110.)
One hypothesis about natural numbers is known as Goldbach's Conjecture
5 . It states that all
even integers greater than two can be expressed as the sum of two primes. It is one of the oldest
still-unsolved problems about numbers. How would you write this conjecture as a WFF?
Enough about number theory. Let's look at some examples about common data structures and some about
our favorite problem, WaterWorld.
Example 4.7
If your program uses binary search trees and your domain is tree nodes, you need to
know ∀node : ((data (left (node)) ≤ data (node)) ∧ (data (right (node)) > data (node))).
If these trees are also balanced, you need to know ∀node :
((height (left (node)) = height (right (node))) ∨ (height (left (node)) + 1 = height (right (node))) ∨ (height (left (node))
Again, these assume the implied interpretations.
Example 4.8
We would like to be able to state that the output of a sorting routine is, in fact, sorted. Let's
assume we're sorting arrays into ascending order.
To talk about the elements of an array in a typical programming language, we would write
something like a [i]. For this example, we'll use that notation, even though it doesn't quite t the
logic's syntax.
5 http://www.wikipedia.org/wiki/Goldbachs_conjecture
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
93
To describe sortedness (in non-decreasing order), we want to state that each element is greater
than or equal to the previous one. However, just like in a program, we need to ensure our formula
doesn't index outside the bounds of the array. For this example, we'll assume that an array's indices
are zero to (but not including) size (a).
sorted (a) ≡ ∀i : ((1 ≤ i) ∧ (i < size (a)) ⇒ (a [i − 1] < a [i]))
When proving things about programs, it's often useful to realize there are alternate ways of
de ning things. So, let's see a couple more de nitions.
We could change our indexing slightly: sorted (a) ≡ ∀i :
((0 ≤ i) ∧ (i < size (a) − 1) ⇒ (a [i] < a [i + 1])).
Or we could state that the ordering holds on every pair of elements: sorted (a) ≡ ∀i :
(∀j : ((0 ≤ i) ∧ (i < size (a)) ∧ (0 ≤ j) ∧ (j < size (a)) ∧ (i < j) ⇒ (a [i] ≤ a [j]))). This
de nition isn't any stronger, but it makes an additional property explicit. Generally, you'd nd it
harder to prove that this formula was true, but once you did, you'd nd it easier to use this formula
to prove other statements.
Exercise 4.1.1.5 (Solution on p. 110.)
The two preceding examples used functions like left, size, and subtraction, although our logic syn-
tax doesn't include such functions. However, we can rewrite any use of functions with appropriate
new relations.
As an example, rewrite i < size (a) in proper rst-order syntax.
Exercise 4.1.1.6 (Solution on p. 110.)
One simple WaterWorld fact is that if a location has no unsafe neighbors, then its number of
adjacent pirates is zero. Furthermore, the implication goes both ways. How would you state that
as a WFF?
Exercise 4.1.1.7 (Solution on p. 110.)
How would you make a similar statement about the number of adjacent pirates being one?
These statements are very similar to, and provable from, the rst-order WaterWorld domain axioms (Sec-
tion 6.6).
4.1.1.2.2 A hint on deciphering formulas' meanings
Some formulas can get pretty hairy: ∀x : (∃y : (∀z : (likes (x, y) ∧ ¬likes (y, z)))). The zeroth step is to take
a breath, and read this in English: for every x, there's some y such that for every z , x likes y but y doesn't
like z. Even so, how do we approach getting a handle on what this means? Given an interpretation, how do
we know it's true?
The top-down way would be to read this formula left-to-right. Is the whole formula true? Well, it's only
true if, for every possible value of x, some smaller formula is true (namely, there exists a y such that forall
z , likes (x, y) and ¬likes (y, z). ). (This is a formula with x free, that is, it's a statement about x.) And is
that formula true? Well, precisely when we can nd some y such that . . . (and so on). This direct approach
is hard to keep inside your head all at once.
Most people prefer approaching the problems in a bottom-up manner (or if you prefer, right-to-left or
inside-out): First consider at the small inner bits alone, gure out what they mean, and only then gure out
how they relate.
• What does the innermost formula likes (x, y) ∧ ¬likes (y, z) mean, in English? That's not so bad: x
likes y, and y dislikes z. A statement about three people called x, y , z .
• Working outward, what does ∀z : (likes (x, y) ∧ ¬likes (y, z)) mean? Ah, not so bad either: x likes y,
and y dislikes everybody.
6
• Keep on going: ∃y : (∀z : (likes (x, y) ∧ ¬likes (y, z))) becomes x likes some misanthrope.
• Now it's clear: ∀x : (∃y : (∀z : (likes (x, y) ∧ ¬likes (y, z)))) is just everybody likes some misanthrope.
Phew!
6 Or if you prefer, x likes y, who is a misanthrope. A self-loathing misanthrope, at that!
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
94 CHAPTER 4. FIRST-ORDER LOGIC
4.1.1.2.3 Forall 's friend if
We have already seen quite a few formulas of the general form ∀x : (P (x) ⇒ . . . ). Indeed, this is a very
useful idiom: If our domain is natural numbers but we want to say something about all primes, we simply
write ∀n : (prime (n) ⇒ . . . ). Don't be fooled; this formula is in no way suggesting that all numbers are
prime!
warning: This same construct using ∃ is usually a mistake. Consider ∃x : (P (x) ⇒ . . . ). By
choosing x to be any non-P element, this entire formula is true, without even glancing at what is
inside the . . .!
note: If you have to demonstrate that all ravens are black, ∀i : (isRaven (i) ⇒ isBlack (i)), there
are two ways to do so: You can go out and nd every raven and verify that it's black. Alternately,
you can go and nd every non-black item, and verify that it's a non-raven. Epistemologists, philoso-
phers dealing with how we humans come to learn and know things (about, say, raven colors), go on
to ponder about real-world degrees-of-belief: If we have only looked at some ravens, and we nd
another raven and con rm it is black, does this increase our degree of belief about all ravens being
black? If so, then whenever we nd a non-black item which is a non-raven, this must also increase
our degree of belief that all ravens are black. This leads to Hempel's (so-called) Paradox: if we are
looking for evidence to choose between two competing hypotheses (say, all non-black items are not
ravens versus all non-orange items are not ravens), then nding a purple cow increases our belief
in both of these hypotheses, simultaneously!
4.1.2 bound variables, free variables7
In the previous examples we often re-used variable names, even within the same formula. This shouldn't
be surprising or confusing, since we do the same thing in programs (another formal language). In fact, the
same notions of bound and free variables occur in both situations. An occurrence of variable
x is bound if
it is in the body of a quanti er ∀x . . . or ∃x . . .. Otherwise, the occurrence is free.
For example, in ∀x : (likes (x, y)), the variable y is free but x is not. So this is a statement about y ; we
can't evaluate this to true/false until we get some context for y . It's useful as a subpart for some bigger
formula.
note: The concept x free in φ does not talk about the context of φ. So don't confuse it
with well, over on this part of the page, φ happens to occur as the sub-part of another formula
containing ∀x : (. . . ), so x really is bound. (Just as 7 is prime, even though people sometimes use
7 in the context of 7+1.) Whether x is free in a φ can be determined by a function isFree (x, φ),
needing no other information to produce an answer.
Looking back at our previous examples, we can see that many of the formulas we made had no free variables
all variables were bound by some quanti er in the formula. The truth of such formulas depends only on
the interpretation and not on any additional knowledge about what any free variables refer to. Thus, these
formulas are common and important enough that we give them a special name, sentences.
A given variable name can actually have both bound and free occurrences within the same formula, as
in R (x) ∧ ∃x : (¬R (x)). (This formula about x is satis able: it says that R is true about x, but isn't true
about everything.) In essence, there are two dierent underlying variables going on, but they each happen
to have the same name; from scope it can be decided which one each occurence refers to. In programming
language terms, we'd say that the inner x (the local variable) shadows the outer x (the enclosing variable).
In these terms, free variables in logic correspond to global variables in programs.
Clearly ∀x : (R (x)) is always equivalent to ∀y : (R (y)); variable names are entirely arbitrary (except
maybe for their mnemonic value). So the previous formula might be more clearly re-written as R (x) ∧ ∃y :
7 This content is available online at <http://cnx.org/content/m12081/1.7/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
95
(¬R (y)). (This careful re-writing while respecting a variable's scope is called α-renaming.) Even if 17
quanti ers each used the same variable (name) x, we could carefully α-renaming 17 times, and end up
with an equivalent formula where all quanti ers use distinct variables. This will be useful to avoid potential
confusion, especially in the upcoming inference rules (Note, p. 99), where we'll be introducing and eliminating
quanti ers.
Example 4.9
The formula ∀x : (A (x)) ∧ ∃x : (B (x)) ∧ ∀x : (C (x)) is equivalent to the more readable
∀x : (A (x)) ∧ ∃y : (B (y)) ∧ ∀z : (C (z)).
4.1.3 normal forms revisited8
4.1.3.1 CNF and DNF revisited (Optional)
In rst-order logic, normal forms are still useful for providing a notion of a canonical form. However, their
other bene t of corresponding closely to truth tables does not apply here, since truth tables aren't useful for
rst-order logic.
A formula in Prenex Conjunctive Normal Form, or Prenex CNF, has a body in CNF preceded by
Prenex Disjunctive Normal Form, or Prenex DNF, has
a series of quanti ers. Similarly, a formula in
a body in DNF preceded by a series of quanti ers.
Example 4.10
Assuming φ is in CNF, then the following are each in prenex CNF. On the other hand, if φ is in
DNF, these are in prenex DNF.
• φ
• ∀x.φ
• ∃x.∀y .∃z .φ
Every formula has an equivalent prenex CNF formula and equivalent prenex CNF formula. For brevity,
we'll skip proving this.
4.2 Reasoning with equivalences
4.2.1 rst-order equivalences9
Now that we can express interesting concepts using the quanti ers ∃ (there exists) and ∀ (for all),
how can we use them for the problem of determining whether a formula is true? Back in lowly propositional
logic, we had three methods:
• truth tables,
• equivalences, and
• formal proofs with inference rules.
How can we adapt these approaches, for rst-order logic?
Well, truth tables have no analog approach. With quanti ers, we don't have a nite set of propositions.
Furthermore, variables can't refer to speci c items in the domain until we try to interpret them. And when
we do, the domain may be of any size possibly even in nite. Using a truth table on an in nite domain
is clearly infeasible, but the real problem stems from how we want to be able to discuss reasoning without
respect to a particular domain.
8 This content is available online at <http://cnx.org/content/m12082/1.5/>.
9 This content is available online at <http://cnx.org/content/m10729/2.26/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
96 CHAPTER 4. FIRST-ORDER LOGIC
However, we can add equivalences and inference rules to cope with quanti ers. After showing how to
work with quanti ers, we'll come back to examine our newly-augmented systems for those desirable traits,
soundness and completeness.
4.2.1.1 First-order Equivalences
When we upgrade from propositional logic to rst-order logic, what changes do we need to make to the
laws of boolean algebra? Well rst o, we can keep all the existing propositional equivalences (Section 6.1).
For example, ∀x : (¬ (φ ∧ ψ)) ≡ ∀x : (¬φ ∨ ¬ψ). (Technically, we're even making those equivalences
stronger, since those meta-variablesφ, ψ , θ can now stand for any rst-order formula, rather than merely
propositional formulas.)
But, we also need additional identities to deal with our new-fangled quanti ers. What should these be?
The most interesting are those that relate the two kinds of quanti ers. Universal quanti cation (∀) says that
something holds for all members of the domain, and existential quanti cation (∃) says that something holds
for at least one member. Clearly, ∀x : (φ) implies ∃x : (φ), but the other direction doesn't hold, so that is
not an equivalence.
aside: Wait just a minute! That implication holds only if the domain is non-empty, so that there
is at least one member in it. We'll see this restriction appear a few times.
What about ∀x : (¬φ)? In English, for all items x, φ(x) does not hold. A more natural way to say this
is that there is no item x such that φ(x) does hold that is, ¬∃x : (φ). Indeed, this will be one of our new
boolean algebra rules.
See a list of equivalences with quanti ers (Section 6.3). As before, we can use these to show other pairs
of formulas equivalent, as in the following examples.
Example 4.11
Using these identities, we can simplify formulas such as the following: ∀y :
(∀x : (R (x) ∧ Q (x, y))) ∧ ¬∃z : (¬R (z)).
1 ∀y : (∀x : (R (x) ∧ Q (x, y))) ∧
¬∃z : (¬R (z))
2 ≡ ∀y : Complementation of ∃
(∀x : (R (x) ∧ Q (x, y))) ∧ ∀z :
(¬¬R (z))
3 ≡ ∀y : Double Complementation
(∀x : (R (x) ∧ Q (x, y))) ∧ ∀z :
(R (z))
4 ≡ ∀x : Reordering ∀s
(∀y : (R (x) ∧ Q (x, y))) ∧ ∀z :
(R (z))
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
97
5 ≡ ∀x : Distribution of ∀ over ∧
(∀y : (R (x)) ∧ ∀y : (Q (x, y))) ∧
∀z : (R (z))
6 ≡ ∀x : Simpli cation of ∀ (y not free in
(R (x) ∧ ∀y : (Q (x, y))) ∧ ∀z : R (x))
(R (z))
7 ≡ ∀x : renaming
(R (x) ∧ ∀y : (Q (x, y))) ∧ ∀x :
(R (x))
8 ≡ ∀x : Distribution of ∀ over ∧
(R (x) ∧ ∀y : (Q (x, y)) ∧ R (x))
9 ≡ ∀x : Commutativity of ∧
(∀y : (Q (x, y)) ∧ R (x) ∧ R (x))
10 ≡ ∀x : Associativity of ∧
(∀y : (Q (x, y)) ∧ R (x) ∧ R (x))
11 ≡ ∀x : (∀y : (Q (x, y)) ∧ R (x)) Idempotency of ∧
Table 4.2
Admittedly, some of these steps are rather small and obvious (e.g., our use of commutativity
and associativity); we include them to illustrate how the identities of propositional logic are also
used in rst-order logic.
Example 4.12
An example of ∀x : (ψ) ≡ ψ where ψ doesn't contain x occurring free: Let ψ be the formula
we've seen before (Exercise 4.1.1.3), asserting that a positive integer n ∀j :
was noncomposite:
(∀k : ((jk = n) ⇒ (j = 1) ∨ (k = 1))). Since n occurs free, the truth of this formula depends on
the value of n. The formula ∀x : (ψ) really is equivalent to ψ : It's true for exactly the same values
of n. The use of x is essentially a bit of a rus, since x plays no part of the meat of the ψ .
However, the following formula is certainly not equivalent: ∀n : (ψ). This formula asserts
that all elements of the domain are non-composite (and it doesn't depend on choosing a particular
interpretation for n). Because n occurred free, we can't use the simpli cation of quanti ers
identity on it.
Finally, one more variant: ∀j : (ψ). This is equivalent to the original, just like ∀x : (ψ) was.
Why? The j that occurs inside ψ is a local variable, and is dierent from any enclosing bindings'
j. As we saw, local variables shadow less-local ones, just as in most programming languages.
Exercise 4.2.1.1 (Solution on p. 110.)
The equivalences for distributing implication over equivalences seem counterintuitive at rst glance.
Show that the following one holds, given all the identities which don't involve both implication and
quanti ers.
Assuming that ψ does not have any free occurrences of variable x, ∀x : (φ ⇒ ψ) ≡ ∃x : (φ) ⇒ ψ .
Are the following two sentences true?
• All ying pigs wear top hats. ∀p : (wears_top_hat (p)) (over the domain of ying pigs).
• All numbers in the empty set are even. ∀x : (even (x)) (over the empty domain).
• Every Pulitzer prize winner I've met thinks I'm smart, and cute, too! ∀x :
(thinksImSmartAndCute (x)) (over the empty, since I haven't met any Pulitzer prize winners).
Each sentence states that some property holds for every member of some set ( ying pigs or the empty set),
but there are no such members. Such sentences are considered vacuously true.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
98 CHAPTER 4. FIRST-ORDER LOGIC
Okay, maybe you believe that the sentences aren't false, but you still want some reason to consider them
true. Well, think of their negations:
• There exists a ying pig not wearing a top hat. ∃p : (¬wears_top_hat), over the (empty) domain
of ying pigs. You can't go o and nd a ying pig which contradicts this, since you can't nd any
ying pig at all. (Note that the negation isn't No ying pigs wear top hats.)
• There exists a number in the empty set that is even. ∃x : (¬even), over the empty domain. (The
negation isn't No numbers in the empty set are even.)
Since these negations are false, the original sentences must be true. This is also similar to the fact that a
simple propositional implication, a⇒b is true, if a is in fact false, regardless of the truth of b; in this crude
analogy, a corresponds to in the domain.
aside: In boolean algebra, we only allow the values false and true, with no third option. This
is sometimes called the law of the excluded middle. Philosophers have developed trimodal
logics which have a third option, but everything in those logics can be translated into something
¶
in traditional logic; such logics might be more convenient in some cases, but they aren't more
expressive. Fuzzy Logic, on the other hand, is a variant where every proposition has a degree
of truth (from zero to one). While this is dierent than propositional logic (and, it is the right
way to model many real-world problems), as a logic it hasn't yielded interesting new mathematical
results.
Even more silliness can ensue when the domain is empty: For example, not only is every member of the
empty set even, but every member is simultaneously odd! That is, ∀x : (R (x) ∧ ¬R (x)) is true (only)
when the domain is the empty set. Even more degnerately, ∀x : (false) is a true (only) on the emtpy domain.
4.2.1.2 Are we done yet?
While equivalences are very useful, we are often interested in implications such as the one mentioned previ-
ously: ∀x : (φ) ⇒ ∃x : (φ). We could rephrase that as an equivalence, ∀x : (φ) ⇒ ∃x : (φ) ≡ true. Informally,
it should be clear that that is rather awkward, and formally it is as well.
But such implications are exactly what inference rules are good for. So, let's continue and consider what
rst-order inference rules (Section 4.3.1) should be.
4.3 Reasoning with inference rules
4.3.1 rst-order inference rules10
4.3.1.1 Inference with quanti ers
Proving rst-order sentences with inference rules is not too dierent than for propositional ones. We have
two slight twists to add: upgrading propositions to relations, and quanti ers. We still keep all our original
propositional inference rules (Section 6.2), but declare they can now be used on rst-order WFFs. For our
quanti ers, we introduce new rst-order inference rules (Section 6.4) for adding and eliminating quanti ers
from formulas. These four new rules look surprisingly simple, but they do have a couple of subtleties we
have to keep track of.
4.3.1.1.1 Exists-intro
What is the most natural way to prove an existential sentence, like there exists a prime num-
ber greater than 5? That's easy you just mention such a number, like 11, and show
that it is indeed prime and greater than 5. In other words, once we prove (11 > 5) ∧ ∀j :
10 This content is available online at <http://cnx.org/content/m10774/2.27/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
99
(∀k : ((11 = jk) ⇒ (j = 1) ∨ (k = 1))) we can then conclude using the inference rule ∃Intro that
the formula ∃p : ((p > 5) ∧ ∀j : (∀k : ((p = jk) ⇒ (j = 1) ∨ (k = 1)))) is true. In general, to prove a for-
mula of the form ∃x : (φ), we show that φ holds when x is replaced with some particular witness. (The
witness was 11 in this example.) The inference rule is φ[p7→c] ` ∃p : (φ). The notation φ[v7→w ] means the
formula φ but with every occurrence of v replaced by w . For example, we earlier wrote down the formula
φ[p7→11], and then decided that this was su cient to conclude ∃p : (φ).
note: Observe that you'll never use the substitution-notation φ[. . .7→. . .] as part of
a literal formula it is only used in the inference rule, as a shorthand to describe the actual
formula. (It's a pattern-matching metalanguage!)
note: While it seems like substitution should be a simple textual search-and-replace, it is some-
times more complicated. In the formula φ = (x > 5) ∧ ∃x : (R (x)), we don't want φ[x7→6] to
try to mention R (6), much less generate something nonsensical like ∀6 : (. . . ). In programming
languages, we say we want hygienic macros, to respect our the language's notions of variables and
scope. E.g., the C pre-processor's #define and #include notably does not respect hygiene, and
can inadvertently lead to hard-to- nd bugs. Solution: For simplicity, we will always consistently
rename variables (p. 94) so that each quanti er binds a distinct variable.
How do you nd a witness? That's the di cult part. You, the person creating the proof, must grab
a suitable example out of thin air, based on your knowledge of what you want to prove about it. In our
previous example, we used our knowledge about prime numbers and about the greater-than relation to pick
a witness that would work. In essence, we gured out what facts needed to be true about the witness for the
formula to hold, and used that to guide our choice of witness. Of course, this can easily be more di cult, as
when proving that there exists a prime greater than 6971 of the form 4x − 1. (It turns out that 796751 will
su ce as a witness here.) Another approach is trial-and-error: Pick some candidate value, and see if it does
indeed witness what you're trying to prove. If you succeed, you're done. If not, pick another candidate.
4.3.1.1.2 Exists-Elim
The complementary ∃Elim rule corresponds to giving a (new) name to a witness. Thus if you know there
exists some prime bigger than 5, then by ∃Elim we can think of giving some witness the name (say) c, and
end up concluding c is a prime bigger than 5. The caveats are that c must be a new name not already
used in the proof, and dierent from any variables free in the conclusion we're aiming for. However, we will
be able to use that variable c along with universal formulas to get useful statements.
Thus the general form of the rule is that ∃p : (φ) ` φ[p7→c]. That is, we can rewrite the body of the exists,
replacing the quanti ed variable p with any new variable name c, subject to the restrictions just mentioned.
4.3.1.1.3 Forall-Intro
Can we extend that idea to proving a universal sentence? One witness is certainly not enough. We'd need to
work with lots of witnesses, in fact, every single member of our domain. That's not very practical, especially
with in nitely large domains. We need to show that no matter what domain element you choose, the formula
holds.
Consider the statements If n is prime, then we know that . . . and A person X who runs a business
should always . . .. Which n is being talked about, and which person? Well, any number or person,
respectively. After learning about quanti ers, you may want to preface these sentences with For all n or
For all [any] persons X . But a linguist might point out that while yes for all is related to the speaker's
thought, they are actually using a subtly dierent mode that of referring to a single person or number,
albeit an anonymous, arbitrary one. If an arbitrary element really is a natural mode of thought, should
our proof system re ect that?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
100 CHAPTER 4. FIRST-ORDER LOGIC
If we choose an arbitrary member of the domain, and show that the sentence holds for it, that is
su cient. But, what do we mean by arbitrary? In short, it means that we have no control over what
element is picked, or equivalently, that the proof must hold regardless of what element is picked. More
precisely, a variable is arbitrary unless:
• A variable is not arbitrary if it is free in (an enclosing) premise.
• A variable is not arbitrary if it is free after applying ∃Elim either as the introduced witness c, or
free anywhere else in the formula.
The usual way to introduce arbitrary variables is during ∀Elim (w/o later using it in ∃Elim). The formal
inference rule for introduction of universal quanti cation will use these cases as restrictions.
4.3.1.1.4 Forall-Elim
∀x : (φ) (where φ is a formula presumably involving
Getting rid of universal quanti ers is easy: if you know
x), well then you can replace x with anything you want, and the resulting formula will be true. We say
∀x : (φ) ` φ[x7→t] where t is any term. Any variables in t are arbitrary, unless it is an already-existing
non-arbitrary variable.
For example, suppose we know that ∀n : (prime (n) ∧ (n > 2) ⇒ odd (n)). We can replace n with some
term like m+4 to conclude prime (m + 4) ∧ (m + 4 > 2) ⇒ odd (m + 4). The variable m is arbitrary, unless
it already occurred in non-arbitrary in a previous line of the proof (perhaps introduced via ∃Elim). A more
usual step is to use a term which is just a single variable, and (by coincidence) happens to have the same
name as the quanti ed variable we are eliminating. Thus we often conclude prime (n) ∧ (n > 2) ⇒ odd (n)
(note the absence of the initial ∀); n is arbitrary (unless it had already been confusingly in use as a non-
arbitrary variable earlier). This is helpful when we'll be later re-introducing the ∀ in a later step; see the
example below.
4.3.1.2 Formal inference rules and proofs
Recall the syllogisms from a previous lecture. The general form of a syllogism is
1. ∀x : (P (x) ⇒ Q (x)) [major premise]
2. P (c) [minor premise]
3. Q (c) [conclusion]
In our system, we don't have syllogism as a separate rule of inference, but it's easy to see how to translate
any syllogism into our system: (for speci c relations P and Q, and a speci c constant c).
1 ∀x : (P (x) ⇒ Q (x)) Premise
2 P (c) Premise
3 P (c) ⇒ Q (c) ∀Elim, by line 1, with x=c
4 Q (c) ⇒Elim, by lines 2,3, with
φ = P = c and ψ = Q = c
Table 4.3
Eliminating a quanti er via ∀Elim and ∃Elim is often merely an intermediate step, where the quanti er
will be reintroduced later. This moves the quanti cation from being explicit to implicit, so that we can use
other inference rules on the body of the formula. When this is done, it is very important to pay attention
to the restrictions on ∀Intro, so that we don't accidentally prove anything too strong.
Example 4.13
∃x : (∀y : (φ)) ` ∀y : (∃x : (φ)) , for the particular case of φ = R (x, y) (other cases all similar).
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
101
1 ∃x : (∀y : (R (x, y))) Premise
2 ∀y : (R (p, y)) ∃Elim, line 1
3 R (p, q) ∀Elim, line 2
4 ∃x : (R (x, q)) ∃Intro, line 3
5 ∀y : (∃x : (R (x, y))) ∀Intro, line 4
Table 4.4
Remember that in line 5, for ∀Intro, we must verify that q is arbitrary. It is, since it was
introduced in line 3 by ∀Elim, and there hasn't been an intervening ∃Elim between lines 3 and 5.
We cannot instead conclude in line 4 that ∀x : (R (x, q)) by ∀Intro, since variable p was
introduced by ∃Elim in line 2, and therefore not arbitrary.
Exercise 4.3.1.1 (Solution on p. 111.)
Let's reverse the previous proof goal: ∀y : (∃x : (φ)) ` ∃x : (∀y : (φ)) , for the particular case
of φ = R (x, y) (other cases all similar). This statement does not hold in general. So, what's the
problem with the following proof ?
1 ∀y : (∃x : (R (x, y))) Premise
2 ∃x : (R (x, q)) ∀Elim, line 1
3 R (p, q) ∃Elim, line 2
4 ∀y : (R (p, y)) ∀Intro, line 3
5 ∃x : (∀y : (R (x, y))) ∃Intro, line 4
Table 4.5
The ∀Intro principle is actually very familiar. For instance, after having shown ¬ (a ∧ b) ` ¬a ∨ ¬b, we
then claimed this was really true for arbitrary propositions instead of just a,b. (We actually went a bit
further, generalizing individual propositions to entire (arbitrary) WFFs φ,ψ . This could only be done because
in any particular interpretation, a formula φ will either be true or false, so replacing it by a proposition still
preserves the important part of the proof-of-equivalence.)
The ∀Intro is also used in many informal proofs. Consider: If a number n is prime, then . . .. This
translates to prime (n) ⇒ . . ., where n is arbitrary. We are entirely used to thinking of this as ∀n :
(prime (n) ⇒ . . .) even though n was introduced as if it were a particular number.
4.3.1.3 Proofs and programming
We previously saw (Section 2.4.4.1: Proofs and programming) that the inference rules of propositional logic
are closely related to the process of type checking. The same holds here. For example, in many programming
languages, we can write a sorting function that works on any type of data. It takes two arguments, a
comparison function for the type and a collection (array, list, . . .) of data of that type. The type of the
sorting function can then be described as for all types T, given a function of type (T and T) → T, and data
of type (collection T ), it returns data of type (collection T ). This polymorphic type-rule uses universal
quanti cation.
Note that the details about substitutions and capture noted here arise in any kind of program that
manipulates expressions with bound variables. That includes not only automated theorem provers, but
compilers. To avoid such issues, many systems essentially rename all variables by using pointers or some
similar system of each variable referring to its binding-site.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
102 CHAPTER 4. FIRST-ORDER LOGIC
When people speak of proofs written by computer
11 , they're talking about this style of inference rule
proofs.
4.4 Exercises for First-Order Logic12
Throughout these exercises, a 6= b is simply a shorthand for ¬ (a = b).
4.4.1 Relations and Interpretations
Exercise 4.4.1
Consider the binary relation is − a − factor − of on the domain {1, 2, 3, 4, 5, 6}.
1. List all the ordered pairs in the relation.
2. Display the relation as a directed graph.
3. Display the relation in tabular form.
4. Is the relation re exive? symmetric? transitive?
Exercise 4.4.2
How would you de ne addsTo as a ternary relation?
1. Give a prose de nition of addsTo (x, y, z) in terms of the addition function.
2. List the set of triples in the relation on the domain {1, 2, 3, 4}.
Exercise 4.4.3
Generalize the previous problem (Exercise 4.4.2) to describe how you can represent any k -ary
function as a (k + 1)-ary relation.
Exercise 4.4.4
Are each of the following formulas valid, i.e., true for all interpretations? (Remember that the
relation names are just names in the formula; don't assume the name has to have any bearing on
their interpretation.)
• For arbitrary a and b in the domain, atLeastAsWiseAs (a, b) ∨ atLeastAsWiseAs (b, a)
• For arbitrary a in the domain, prime (a) ⇒ (odd (a) ⇒ prime (a))
• For arbitrary a and b in the domain, betterThan (a, b) ⇒ ¬betterThan (b, a)
For each, if it is true or false under all interpretations, prove that. For these small examples, a
truth table like this one (Table 3.1) will probably be easier than using Boolean algebra or inference
rules. Otherwise, give an interpretation in which it is true, and one in which it is false.
note: As always, look at trivial and small test cases rst. Here, try domains with zero, one, or
two elements, and small relations.
Exercise 4.4.5 (Solution on p. 111.)
[Practice problem− − −solution provided.]
Suppose we wanted to represent the count of neighboring pirates with a binary relation, such
that when location A has two neighboring pirates, piratesNextTo (A, 2) will be true. Of course,
piratesNextTo (A, 1) would not be true in this situation. These would be analogous with the
propositional WaterWorld propositions A − has − 2 and A − has − 1, respectively.
1. If we only allow binary relations to be subsets of a domain crossed with itself, then what must
the domain be for this new relation piratesNextTo?
11 http://www.economist.com/science/displayStory.cfm?story_id=3809661
12 This content is available online at <http://cnx.org/content/m12353/1.28/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
103
2. If we further introduced another relation, isNumber?, what is a formula that would help
distinguish intended interpretations from unintended interpretations? That is, give a formula
that is true under all our intended interpretations of piratesNextTo but is not true for some
nonsense interpretations we want to exclude. (This will be a formula without an analog in
the WaterWorld domain axioms (Section 6.5).)
Exercise 4.4.6
Determine whether the relation R on the set of all people is re exive, antire exive, symmetric,
antisymmetric, and/or transitive, where (a, b) ∈ R if and only if ...
1. a is older than b.
2. a is at least as old as b.
3. a and b are exactly the same age.
4. a and b have a common grandparent.
5. a and b have a common grandchild.
Exercise 4.4.7
For each of the following, if the statement is true, explain why, and if the statement is false, give
a counter-example relation.
1. If R is re exive, then R is symmetric.
2. If R is re exive, then R is antisymmetric.
3. If R is re exive, then R is not symmetric.
4. If R is re exive, then R is not antisymmetric.
5. If R is symmetric, then R is re exive.
6. If R is symmetric, then R is antire exive.
7. If R is symmetric, then R is not antire exive.
4.4.2 Quanti ers
Exercise 4.4.8
Let P (x) be the statement has been to Prague, where the domain consists of your classmates.
1. Express each of these quanti cations in English.
• ∃x : (P (x))
• ∀x : (P (x))
• ¬∃x : (P (x))
• ¬∀x : (P (x))
• ∃x : (¬P (x))
• ∀x : (¬P (x))
• ¬∃x : (¬P (x))
• ¬∀x : (¬P (x))
2. Which of these mean the same thing?
Exercise 4.4.9
Let C (x) be the statement x has a cat, let D (x) be the statement x has a dog, and let F (x)
be the statement x has a ferret. Express each of these statements in rst-order logic using these
relations. Let the domain be your classmates.
1. A classmate has a cat, a dog, and a ferret.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
104 CHAPTER 4. FIRST-ORDER LOGIC
2. All your classmates have a cat, a dog, or a ferret.
3. At least one of your classmates has a cat and a ferret, but not a dog.
4. None of your classmates has a cat, a dog, and a ferret.
5. For each of the three animals, there is a classmate of yours that has one.
Exercise 4.4.10
Determine the truth value of each of these statements if the domain is all real numbers. Where
appropriate, give a witness.
1. ∃x : x2 = 2
2. ∃x : x2 = −1
3. ∀x : x2 + 2 ≥ 1
4. ∀x : x2 6= x
4.4.3 Interpreting First-order Formulas
Exercise 4.4.11
Let P (x), Q (x), R (x), and S (x) be the statements x is a duck, x is one of my poultry, x is an
o cer, and x is willing to waltz, respectively. Express each of these statements using quanti ers,
logical connectives, and the relations P (x), Q (x), R (x), and S (x).
1. No ducks are willing to waltz.
2. No o cers ever decline to waltz.
3. All my poultry are ducks.
4. My poultry are not o cers.
5. Does the fourth item follow from the rst three taken together? Argue informally; you don't
need to use the algebra or inference rules for rst-order logic here.
Exercise 4.4.12
You come home one evening to nd your roommate exuberant because they have managed to
prove that there is an even prime number bigger than two. More precisely, they have a correct
proof of ∃y : (P (y) ∧ (y > 2) ⇒ E (y)), for the domain of natural numbers, with P interpreted
as is prime? and E interpreted as is even?. While they are celebrating their imminent fame at
this amazing mathematical discovery, you ponder. . .
1. . . .and realize the formula is indeed true for that interpretation. Brie y explain why. You
don't need to give a formal proof using Boolean algebra or inference rules; just give a particular
value for y and explain why it satis es the body of ∃y : (y) .
2. Is the formula still true when restricted to the domain of natural numbers two or less? Brie y
explain why or why not.
3. Is the formula still true when restricted to the empty domain? Brie y explain why or why
not.
4. Give a formula that correctly captures the notion there is an even prime number bigger than
2 .
Exercise 4.4.13
For the sentence ∀x : (∀y : (A (x) ∧ B (x, y) ⇒ A (y))) state whether it is true or false, relative
to the following interpretations. If false, give values for x and y witnessing that.
1. The domain of the natural numbers, where A is interpreted as even?, and B is interpreted
as equals
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
105
2. The domain of the natural numbers, where A is interpreted as even?, and B is interpreted
as is an integer divisor of
3. The domain of the natural numbers, where A is interpreted as even?, and B is interpreted
as is an integer multiple of
4. The domain of the Booleans, {true, false}, where A is interpreted as false?, and B is inter-
preted as equals
5. The domain of WaterWorld locations in the particular board where locations Y and Z contain
pirates, but all other locations are safe, the relation symbol A is interpreted as unsafe?, and
B is interpreted as neighbors
6. All WaterWorld boards, where A is interpreted as safe?, and B is interpreted as neighbors.
(That is, is the formula valid for WaterWorld?)
Exercise 4.4.14
Translate the following conversational English statements into rst-order logic, using the suggested
predicates, or inventing appropriately-named ones if none provided. (You may also freely use =
which we'll choose to always interpret as the standard equality relation.)
1. All books rare and used. This is claimed by a local bookstore; what is the intended domain?
Do you believe they mean to claim all books rare or used?
2. Everybody who knows that UFOs have kidnapped people knows that Agent Mulder has
been kidnapped. (Is this true, presuming that no UFOs have actually visited Earth. . .yet?)
Exercise 4.4.15
Write a formula for each of the following. Use the two binary relations isFor and isAgainst and
domain of all people.
• All for one, and one for all! We'll take one to mean one particular person, and moreover,
that both ones are referring the same particular person, resulting in There is one whom
everybody is for, and that one person is for everybody.
13
• If you're not for us, you're against us. In aphorisms, you is meant to be an arbitrary
person; consider using the word one instead. Furthermore, we'll interpret us as applying
to everybody. That is, One always believes that `if one is not for me, then one is against
me' .
• The enemy of your enemy is your friend. By your enemy we mean somebody you are
against, and similarly, your friend will mean somebody you are for. (Be carefule! This
may be dierent than somebody who is against/for you).
• Somebody has an enemy. (We don't know of an aphorism expressing this.
14 )
Two interpretations are considered fundamentally the same (or isomorphic) if you can map one interpre-
tation to the other simply by a consistent renaming of domain elements.
Exercise 4.4.16 (Solution on p. 111.)
[Practice problem− − −solution provided.]
Find two fundamentally dierent interpretations that satisfy the statement There exists one
person who is liked by two people.
Exercise 4.4.17
For the four Musketeer formulas from a previous exercise (Exercise 4.4.15), nd three fun-
damentally dierent interpretations of isFor which satisfy all the formulas on a domain of three
people.
13 Dumas' original musketeers presumably meant something dierent: that each one of them was for each (other) one of the
them, making the vice-versa clause redundant. But this is boring for our situation, so we'll leave that interpretation to Athos,
Porthos, and Aramis alone.)
14 None of the following quite capture it: Life's not a bed of roses; It's a dog-eat-dog world; Everyone for themselves;
You can't please all the people all the time.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
106 CHAPTER 4. FIRST-ORDER LOGIC
Depict each of these interpretations as a graph. Draw three circles (nodes) representing the
three people, and an arrow (edge) from a person to each person they like. (You can glance at
Rosen Section 9.1, Figure 8 for an example.)
note: One of the interpretations is unintuitive in that isFor and isAgainst don't correspond to
what we probably mean in English.
Exercise 4.4.18
Translate the following statements into rst-order logic. The domain is the set of natural numbers,
and the binary relation kth (k, n) indicates whether or not the k th number of the sequence is n.
For example, the sequence (5, 7, 5), is represented by the relation kth = {(0, 5) , (1, 7) , (2, 5)}. You
can also use the binary relations =, <, and ≤, but no others.
You may assume that kth models a sequence. No index k is occurs multiple times, thus excluding
kth = {(0, 5) , (1, 7) , (0, 9)}. Thus, kth is a function, as in a previous example representing an array
as a function (Example 4.8). Also, no higher index k occurs without all lower-numbered indices
being present, thus excluding kth = {(0, 5) , (1, 7) , (3, 9)}.
1. The sequence is nite.
2. The sequence contains at least three distinct numbers , e.g., (5, 6, 5, 6, 7, 8), but not (5, 6, 5, 6).
3. The sequence is sorted in non-decreasing order, e.g., (3, 5, 5, 6, 8, 10, 10, 12).
4. The sequence is sorted in non-decreasing order, except for exactly one out-of-order element,
e.g., (20, 30, 4, 50, 60).
Exercise 4.4.19
Some binary relations can be viewed as the encoding of a unary function, where the rst element of
the ordered pair represents the function's value. For instance, in a previous exercise (Exercise 4.4.2)
we encoded the binary function addition as a ternary relation addsTo.
1. Give one example of a binary relation which does not correspond to the encoding of a function.
2. Write a rst-order formula describing the properties that a binary relation R must have to
correspond to a unary function.
Exercise 4.4.20
Alternation of quanti ers: Determine the truth of each of the following sentences in each of the
indicated domains.
note: To help yourself, you might want to develop an English version of what the logic sentences
say. Start with the inner formula (talking about people x,y ,z ), then add the quanti er for z to get
a statement about people x,y , and repeat for the other two quanti ers.
Four sentences:
1. ∀x : (∀y : (∃z : (likes (x, y) ∧ ((z 6= y) ⇒ ¬likes (y, z)))))
2. ∃x : (∀y : (∀z : (likes (x, y) ∧ ((z 6= y) ⇒ ¬likes (y, z)))))
3. ∃x : (∃y : (∀z : (likes (x, y) ∧ ((z 6= y) ⇒ ¬likes (y, z)))))
4. ∀x : (∃y : (∀z : (likes (x, y) ∧ ((z 6= y) ⇒ ¬likes (y, z)))))
Four domains:
1. The empty domain.
2. A world with one person, who likes herself.
3. A world with Yorick and Zelda, where Yorick likes Zelda, Zelda likes herself, and that's all.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
107
4. A world with many people, including CJ (Catherine Zeta-Jones), JC (John Cusack), and JR
(Julia Roberts). Everybody likes themselves; everybody likes JC; everybody likes CJ except
JR; everybody likes JR except CJ and IB. Any others may or may not like each other, as
you choose, subject to the preceding. (You may wish to sketch a graph of this likes relation,
similar to Rosen Section 9.1 Figure 8.)
Determine the truth of all sixteen combinations of the four statements and four domains.
4.4.4 Modeling
Exercise 4.4.21
Translate the following into rst-order logic: Raspberry sherbet with hot fudge (rshf ) is the
tastiest dessert. Use tastier as your only relation.
What is the intended domain for your formula? What is a relation which makes this statement
true? One which makes it false?
Exercise 4.4.22
Even allowing for ellision, the list of WaterWorld domain axioms (Section 6.6) is incomplete, in a
sense. The game reports how many pirates exist in total, but that global information is not re ected
in the propositions or axioms. We had the same problem (Exercise 2.5.13) with the propositional
logic domain axioms
1. First, assume we only use the default WaterWorld board size and number of pirates, i.e., ve.
What additional axiom or axioms do we need?
2. Next, generalize your answer to model the program's ability to play the game with a dierent
number of pirates. What problem do you encounter?
Exercise 4.4.23
The puzzle game of Sudoku is played on a 9×9 grid, where each square holds a number between
1 and 9. The positions of the numbers must obey constraints. Each row and each column has each
of the 9 numbers. Each of the 9 non-overlapping 3×3 square sub-grids has each of the 9 numbers.
Like WaterWorld, throughout the game, some of the values have not been discovered, although
they are determined. You start with some numbers revealed, enough to guarantee that the rest of
the board is uniquely determined by the constraints. Thus, like in WaterWorld, when deducing the
value of another location, what has been revealed so far would serve as premises in a proof.
Fortunately, there are the same number of rows, columns, subgrids, and values. So, our domain
is {1, 2, 3, 4, 5, 6, 7, 8, 9}.
To model the game, we will use the following relations:
• value (r, c, v) indicates that at row r, column c is the value v .
• v = w is the standard equality relation.
• subgrid (g, r, c) indicates that subgrid g includes the location at row r, column c.
Provide domain axioms for Sudoku, and brie y explain them. These will model the row, column,
and subgrid constraints. In addition, you should include constraints on our above relations, such
as that each location holds one value.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
108 CHAPTER 4. FIRST-ORDER LOGIC
4.4.5 Reasoning with Equivalences
Exercise 4.4.24
Some of the rst-order equivalences (Section 6.3) are redundant. For each of the following, prove
the equivalence using the other equivalences.
1. ∀x : (ϕ ⇒ θ) ≡ ∃x : (ϕ) ⇒ θ
2. Assuming a non-empty domain, ∃x : (θ ⇒ ϕ) ≡ θ ⇒ ∃x : (ϕ).
Exercise 4.4.25
We can characterize a prime number as a number n satisfying ∀q :
(∀r : ((qr = n) ⇒ (q = 1) ∨ (r = 1))). Using the equivalences for rst-order logic, show
step-by-step that this is equivalent to the formula ¬∃q : (∃r : ((qr = n) ∧ (q 6= 1) ∧ (r 6= 1))).
Do not use any arithmetic equivalences.
Exercise 4.4.26
A student claims that ∀x : (A (x) ∧ B (x) ⇒ C (z)) ≡ ∀x : (A (x)) ∧ ∀x : (B (x)) ⇒ C (z) by
the distribution of quanti ers. This is actually trying to do two steps at once. Rewrite this as the
two separate intended steps, determine which is wrong, and describe why that step is wrong.
Exercise 4.4.27
Simplify the formula ∀x : (∀y : (∃z : (A (x) ∧ B (y) ⇒ C (z)))), so that the body of each quanti-
er contains only a single atomic formula (De nition: "Well-Formed Formula (WFF) for rst-order
logic", p. 91) involving that quanti ed variable. Provide reasoning for each step of your simpli ca-
tion.
4.4.6 Reasoning with Inference Rules
Exercise 4.4.28 (Solution on p. 111.)
[Practice problem− − −solution provided.]
Prove that syllogisms are valid inferences. In other words, show that ∀x :
(R (x) ⇒ S (x)) , R (c) ` S (c).
Exercise 4.4.29
What is wrong with the following proof of ∃x : (E (x)) ⇒ E (c)?
1 subproof:∃x : (E (x)) ` E (c)
1.a ∃x : (E (x)) Premise for subproof
1.b E (c) ∃Elim, line 1.a
2 ∃x : (E (x)) ⇒ E (c) ⇒Intro, line 1
Table 4.6
Exercise 4.4.30
Using the inference rules, formally prove the last part of the previous problem about ducks and
such (Exercise 4.4.11).
Exercise 4.4.31
Give an inference rule proof of ∀x : (Fruit (x) ⇒ hasMethod (tasty, x)) , ∀y :
(Apple (y) ⇒ Fruit (y)) ` ∀z : (Apple (z) ⇒ hasMethod (tasty, z)).
Exercise 4.4.32
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
109
1. Prove the following: ∃x : (P (x)) , ∀y : (P (y) ⇒ Q (y)) ` ∃z : (Q (z))
2. Your proof above used ∃Intro. Why can't we replace that step with the formula ∀z : (Q (z))
with the justi cation ∀Intro?
3. Describe an interpretation which satis es the proof 's premises, but does not satisfy ∀z :
(Q (z)).
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
110 CHAPTER 4. FIRST-ORDER LOGIC
Solutions to Exercises in Chapter 4
Solution to Exercise 4.1.1.1 (p. 92)
The cue Somebody . . . suggests one person who exists; we'll call them p for popular: ∃p : (. . . ). Now we
need to ll in the dots with everybody likes p, to get: ∃p : (∀x : (likes (x, p))).
Solution to Exercise 4.1.1.2 (p. 92)
The cue Everybody . . . suggests a universal; we'll call them j for J. Doe: ∀j : (. . . ). Now we need to
ll in the dots with somebody likes j , to get: ∀j : (∃x : (likes (x, j))). Note that this formula is just like
the preceding Somebody likes everybody example (Example 4.3), except that the quanti ers have been
swapped (and dierent variable names were used, a super cial dierence).
Solution to Exercise 4.1.1.3 (p. 92)
Evenness is a straightforward translation from An integer n is even, i it is twice some other integer k
: even (n) ≡ ∃k : (n = 2k). Note that by this standard de nition, zero is even.
There are many equivalent ways to de ne primality, just as there many algorithms for checking primality.
One straightforward solution is noncomposite (n) ≡ ∀j : (∀k : ((jk = n) ⇒ (j = 1) ∨ (k = 1))). Well, this
is almost expresses prime, except that n = 1 satis es this formula. A mathematician points out that just as
0 is neither positive nor negative, 1 is neither prime nor composite; as stated this formula actually captures
noncomposite, oops. There are several ways to upgrade this to exactly capture prime.
aside: 1 is called a unit. If we consider the domain of all integers (not just natural numbers),
the idea of primality still makes sense; -17 is also prime; and -1 is also another unit. Similarly,
considering the domain of complex integers { a, b, a + bi | a ∈ Z ∧ b ∈ Z } (could be written
Z + Zi ), then i and −i are also units. How might we generalize our de nition of prime, to work
in these further interpretations?
A similar, equivalent formula to the above is noncomposite (n) ≡ ¬∃j :
(∃k : ((jk = n) ∧ (j 6= 1) ∧ (k 6= 1))).
Solution to Exercise 4.1.1.4 (p. 92)
∀n : (even (n) ∧ (n > 2) ⇒ ∃p : (∃q : (prime (p) ∧ prime (q) ∧ (p + q = n))))
Solution to Exercise 4.1.1.5 (p. 93)
We need a new relation that combines the syntax of < and size. The result would look like
less − than − size (i, a). This assumes the new relation has the obvious intended de nition.
Solution to Exercise 4.1.1.6 (p. 93)
∀x : (∀y : (nhbr (x, y) ⇒ safe (y)) ⇔ has − 0 (x))
Solution to Exercise 4.1.1.7 (p. 93)
There are various solutions, but they all must capture the same idea: there exists exactly one unsafe
neighbor. This solution states that in two parts:
• There exists an unsafe neighbor, u.
• Every unsafe neighbor is u.
Together, these two parts imply there is only one such u.
∀x : (∃u : (nhbr (x, u) ∧ ¬safe (u) ∧ ∀y : (nhbr (x, y) ⇒ (¬safe (y) ⇔ y = u))) ⇔ has − 1 (x))
Solution to Exercise 4.2.1.1 (p. 97)
1 ∀x : (φ ⇒ ψ)
continued on next page
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
111
2 ≡ ∀x : (¬φ ∨ ψ) De nition of ⇒
3 ≡ ∀x : (¬φ) ∨ ψ Distribution of ∀ over ∨
4 ≡ ¬∃x : (φ) ∨ ψ Complementation of ∃
5 ≡ ∃x : (φ) ⇒ ψ De nition of ⇒
Table 4.7
Solution to Exercise 4.3.1.1 (p. 101)
In line 4, ∀Intro requires that variable being generalized, q, be arbitrary. It was introduced in line 2 by
∀Elim, so that's OK. (E.g., we could've used ∀Intro on line 3 to reintroduce the quanti er just eliminated.)
But, q was free when we used ∃Elim on line 3, and this makes the variable no longer arbitrary. Line 3's
choice of p may depend on q, and a variable is only arbitrary if it is free of any such constraints.
Solution to Exercise 4.4.5 (p. 102)
1. The relation needs to accept locations as well as numbers, so the domain is L ∪ N, where L is the set
of WaterWorld locations. Alternatively, you could use {0, 1, 2, 3} instead of N, the set of all natural
numbers.
piratesNextTo (17, 2)
2. The di culty is that it's possible to ask about nonsensical combinations like
and piratesNextTo (W, B). Adding isNumber?, any interpretation would be expected to satisfy, for
arbitrary a and b, piratesNextTo (a, b) ⇒ isNumber? (b) ∧ ¬isNumber? (a, b).
aside: More interestingly though, imagine we did interpret piratesNextTo over the domain
N only. We could then pretend that the locations, instead of being named A,. . .,Z , were just
numbered 1,. . .,24. While this representation doesn't re ect how we model the problem, it is
legal. Exercise for the reader: Write a formula which excludes relation piratesNextTo which
can't match this convention!
Solution to Exercise 4.4.16 (p. 105)
One interpretation that satis es this is a domain of three people Alice, Bob, Charlie, with the likes relation:
{(Alice, Bob) , (Bob, Bob)}. Bob is liked by two people, so it satis es the statement.
Here's another interpretation that is the same except for renaming, and thus not fundamentally dierent:
a domain of three people Alyssa, Bobby, Chuck, with the likes relation: {(Chuck, Alyssa) , (Alyssa, Alyssa)}.
With the substitutions [Chuck7→Alice] and [Alyssa7→Bob], we see that the underlying structure is the same
as before.
Here's an interpretation that is fundamentally dierent: a domain of three people Alice, Bob, Charlie,
with the likes relation: {(Charlie, Bob) , (Alice, Bob)}. No matter how you rename, you don't get somebody
liking themself, so you can see its underlying structure is truly dierent than the preceding interpretations.
English is fuzzy enough that it is unclear whether one and two are meant as exact counts. The above
two examples each assumed they are.
aside: If we change the statement slightly to add a comma: There exists one person, who is
liked by two people , we arguably change the meaning signi cantly. The now-independent rst
clause arguably means there is only one person existent in total, so the overall statement must be
false! There's a quick lesson in the dierence between English dependent and independent clauses.
Solution to Exercise 4.4.28 (p. 108)
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
112 CHAPTER 4. FIRST-ORDER LOGIC
1 ∀x : (R (x) ⇒ S (x)) Premise
2 R (c) Premise
3 R (c) ⇒ S (c) ∀Elim, line 1
4 X (c) ⇒Elim, lines 2,3
Table 4.8
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
Chapter 5
Conclusion, Acknowledgements
5.1 Logic: Looking Back1
5.1.1 Why didn't we begin with quanti ers all along?
We saw three stages of logics:
• Propositional logic, with formulas like DickLikesJane ⇒ ¬JaneLikesDick. While the propositions are
named suggestively, nothing in the logic enforces a relation among these; it is equivalent to A ⇒ ¬B .
• Predicate logic, where variables (and constants) can express a connection between dierent parts of
the formula: likes (y, x) ⇒ ¬likes (x, y) Predicate logic introduced the idea of variables, and required
domains and interpretations to determine truth. But it can't bind variables, and thus requires an
interpretation of x and y to evaluate.
• First-order logic, which included two quanti ers to bind variables: ∀y :
(∃x : (likes (y, x) ⇒ ¬likes (x, y)))
So why, you might ask, didn't we just start out with rst-order logic in the rst lecture? One reason, clearly,
is to introduce concepts one at a time: everything you needed to know about one level was needed in the
next, and then some. But there's more: by restricting our formalisms, we can't express all the concepts of
the bigger formalism, but we can have automated ways of checking statements or nding proofs.
In general, this is a common theme in the theory of any subject: determining when and where you can
(or, need to) trade o expressibility for predictive value. For example, ...
• Linguistics: Having a set of precise rules for (say) Tagalog grammar allows you to determine what is
and isn't a valid sentence; details of the formal grammar can reveal relations to other languages which
aren't otherwise so apparent. On the other hand, a grammar for any natural language is unlikely
to exactly capture all things which native speakers say and understand. If working with a formal
grammar, one needs to know what is being lost and what is being gained.
· Dismissing a grammar as irrelevant because it doesn't entirely re ect usage is missing the point
of the grammar;
· Conversely, condemning some real-life utterances as ungrammatical (and ignoring them) forgets
that the grammar is a model which captures many (if not all) important properties.
Of course, any reasonable debate on this topic respects these two poles and is actually about where
the best trade-o between them lies.
• Psychology: Say, Piaget
2 might propose four stages of learning in children. It may not trade o total
accuracy, for (say) clues of what to look for in brain development.
1 This content is available online at <http://cnx.org/content/m10775/2.20/>.
2 http://www.piaget.org/biography/biog.html
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
113
114 CHAPTER 5. CONCLUSION, ACKNOWLEDGEMENTS
• Physics: Modern pedagogy must trade o quantum accuracy for Newtonian approximations. Re-
searchers exploring elds like particle physics must trade o exact simulations for statistical (stochas-
tic) approximations.
Understanding the theoretical foundations of a eld is often critical for knowing how to apply various
techniques in practice.
5.1.2 Logic and everyday reasoning
We've looked at the impreciseness and ambiguity of natural language statements, but these are not the
only problems hidden in natural language arguments. The following illustrates a common form of hidden
assumption: saying the tenth reindeer of Santa Claus is . . . implies the existence some tenth reindeer. More
subtly, humans use much more information than what is spoken in a conversation. Even aside from body
language, consider a friend asking you Hey, are you hungry? While as a formal statement this doesn't have
any information, in real life it highly suggests that your friend is hungry.
A much more blatant form of missing information is when the speaker simply chooses to omit it. When
arguing for a cause it is standard practice to simply describe its advantages, without any of its disadvantages
or alternatives.
aside: Economists measure things not in cost, but opportunity cost, the price of something
minus the bene ts of what you'd get using the price for something else. E.g., for $117 million the
university can build a new research center. But what else could you do on campus with $117m?
Historically, logic and rhetoric, the art of persuasion through language, are closely linked.
5.1.3 Other logics
You've now been introduced to two logics: propositional and rst-order. But, the story does not have to end
here. There are many other logics, each with their uses.
5.1.3.1 Limitations of rst-order logic's expressiveness
We can make rst-order sentences to express concepts as vertices a and b are connected by a path of length
2 , as well as . . .by a path of length 3, length ≤ 4 , etc.
note: Write a couple of these sentences!
But trying to write vertices a and b are connected [by a path of any length] isn't obvious ... in fact, it
can be proven that no rst-order sentence can express this property! Nor can it express the closely-related
property the graph is connected (without reference to two named vertices a and b).
Hmm, what about second-order logic? It has a bigger name; whatever it means, perhaps it can express
more properties?
What exactly is second-order logic? In rst-order logic, quanti ers range over elements of the domain:
there exist numbers x and y , . . . . In second-order logic, you can additionally quantify over sets of elements
of the domain: there is a set of numbers, such that ... .
Example 5.1
For instance, for all vertices x and y, there exists a set of vertices (call the set Red), the red
vertices include a path from x to y . More precisely, every Red vertex has exactly two Red
neighbors, or it is x or y (which each have exactly 1 red neighbor) . Is this sentence true exactly
when the graph is connected? Why does this description of red vertices not quite correspond to
just the vertices on a path from x to y ?
An interesting phenomenon: There are some relations between how di cult it is to write down a property,
and how di cult to compute it! How might you try to formalize the statement there is a winning strategy
for chess?
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
115
A shortcoming of rst-order logic is that it is impossible to express the concept path. (This can be
proven, though we won't do so here.)
Thus, some other logics used to formalize certain systems include:
• As mentioned, second-order logic is like rst-order logic, but it also allows quanti cation over entire
relations. Thus, you can make formulas that state things like For all relations R, if R is symmetric
and transitive, then ... . While less common, we could continue with third-order, fourth-order, etc.
• Temporal logic is based on quanti cation over time. This is useful to describe how a program's
state changes over time. In particular, it is used for describing concurrent program speci cations and
communication protocols, sequences of communications steps used in security or networking. See, for
example, TeachLogic's Model-Checking module
3 .
• Linear logic is a resource-aware logic. Every premise must be used, but it may be used only once.
This models, for example, how keyboard input is usually handled: reading an input also removes it
from the input stream, so that it can't be read again.
5.1.4 Logic in computer science
Logics provide us with a formal language useful for
• specifying properties unambiguously,
• proving that programs and systems do (or don't) have the claimed properties, and
• gaining greater insight into other languages such as database queries
4 .
Programming language type systems are a great example of these rst two points. The connectives allow
us to talk about pairs and structures (x and y ), unions (x or y ), and functions (if you give the program a
x, it produces a y ). The generics in Java, C++, and C# are based upon universal quanti cation, while
wildcards in Java are based upon existential quanti cation. One formalization of this strong link between
logic and types is called the Curry-Howard isomorphism.
Compilers have very speci c logics built into them. In order to optimize your code, analyses check what
properties your code has e.g., are variables b and c needed at the same time, or can they be stored in the
same hardware register?
More generally, it would be great to be able to verify that our hardware and software designs were correct.
First, specifying what correct means requires providing the appropriate logical formulas. With hardware,
automated veri cation is now part of the regular practice. However, it is so computationally expensive that
it can only be done on pieces of a design, but not, say, a whole microprocessor. With software, we also
frequently work with smaller pieces of code, proving individual functions or algorithms correct. However,
there are two big inter-related problems. Many of the properties we'd like to prove about our software are
undecidable −−− it is impossible to check the property accurately for every input. Also, specifying full
correctness typically requires extensions to rst-order logic, most of which are incomplete.
5 As we've seen,
that means that we cannot prove everything we want. While proving hardware and software correct has its
limitations, logic provides us with tools that are still quite useful. For an introduction to one approach used
in veri cation, see TeachLogic's Model-Checking module
6 .
3 http://cnx.org/content/col10294/latest
4 http://www.cs.rice.edu/∼tlogic/Database/all-lectures.pdf
5 Even something as simple as rst-order logic using the integers as our domain and addition and multiplication as relations
is undecidable. Kurt Gödel, 1931
6 http://cnx.org/content/col10294/latest
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
116 CHAPTER 5. CONCLUSION, ACKNOWLEDGEMENTS
5.2 Acknowledgements7
The TeachLogic Project
8 is the work of many contributors, and was made possible through an NSF CISE9
grant. Major contributors and grant Principle Investigators are
• Moshe Vardi, Rice University
• Matthias Felleisen, Northeastern University
• Ian Barland, Rice University
• Phokion Kolaitis, University of California at Santa Cruz
• John Greiner, Rice University
In addition, Paul Steckler implemented the Base module's Waterworld game
10 .
Students who helped contribute to various TeachLogic modules include (chronologically)
• Peggy Fidelman
• Justin Garcia
• Brian Cohen
• Sarah Trowbridge
• Bryan Cash
• Fuching Jack Chi
• Ben McMahan
TeachLogic has also been in uenced by the Beseme project
11 , headed by Rex Page of Oklahoma Uni-
versity; in particular the Base module owes both some overall structure and speci c details to Beseme.
Janice Bordeaux, from the Engineering Dean's o ce at Rice University, assisted with developing class-
room assessment tools.
7 This content is available online at <http://cnx.org/content/m13613/1.1/>.
8 http://www.teachlogic.org/
9 http://www.nsf.gov/dir/index.jsp?org=CISE
10 http://www.teachlogic.org/WaterWorld
11 http://www.cs.ou.edu/∼beseme/
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
Chapter 6
Appendices and Reference Sheets
6.1 propositional equivalences1
The following lists some propositional formula equivalences. Remember that we use the symbol ≡ as a
relation between two WFFs, not as a connective inside a WFF. In these, φ, ψ , and θ are meta-variables
standing for any WFF.
Propositional Logic Equivalences
Double Complementation ¬¬φ ≡ φ
Complement φ ∨ ¬φ ≡ true φ ∧ ¬φ ≡ false
Identity φ ∨ false ≡ φ φ ∧ true ≡ φ
Dominance φ ∨ true ≡ true φ ∧ false ≡ false
Idempotency φ ∨ φ≡φ φ ∧ φ≡φ
Absorption φ ∧ (φ ∨ ψ) ≡ φ φ ∨ φ ∧ ψ≡φ
Redundancy φ ∧ (¬φ ∨ ψ) ≡ φ ∧ ψ φ ∨ ¬φ ∧ ψ ≡ φ ∨ ψ
DeMorgan's Laws ¬ (φ ∧ ψ) ≡ ¬φ ∨ ¬ψ ¬ (φ ∨ ψ) ≡ ¬φ ∧ ¬ψ
Associativity φ ∧ (ψ ∧ θ) ≡ (φ ∧ ψ) ∧ θ φ ∨ (ψ ∨ θ) ≡ (φ ∨ ψ) ∨ θ
Commutativity φ ∧ ψ≡ψ ∧ φ φ ∨ ψ≡ψ ∨ φ
Distributivity φ ∧ (ψ ∨ θ) ≡ φ ∧ ψ ∨ φ ∧ θ φ ∨ ψ ∧ θ ≡ (φ ∨ ψ) ∧ (φ ∨ θ)
Table 6.1
Equivalences for implication are omitted above for brevity and for tradition. They can be derived, using
the de nition a ⇒ b ≡ ¬a ∨ b.
Example 6.1
For example, using Identity and Commutativity, we have true ⇒ b ≡ ¬true ∨ b ≡ false ∨ b ≡
b ∨ false ≡ b.
1 This content is available online at <http://cnx.org/content/m10540/2.25/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
117
118 CHAPTER 6. APPENDICES AND REFERENCE SHEETS
6.2 propositional inference rules2
Our propositional inference rules
2 This content is available online at <http://cnx.org/content/m10529/2.28/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
119
Abbreviation Name If you know all of. . . . . .then you can infer
∧Intro and-introduction φ φ ∧ ψ
ψ
and-elimination (left) φ ∧ ψ φ
∧Elim
and-elimination (right) φ ∧ ψ ψ
or-introduction (left) φ φ ∨ ψ
∨Intro
or-introduction (right) ψ φ ∨ ψ
∨Elim or-elimination φ`θ θ
ψ`θ
φ ∨ ψ
⇒Intro if-introduction φ, ψ , . . . , θ ` ω φ ∧ ψ ∧ ... ∧ θ ⇒ ω
⇒Elim if-elimination (modus φ⇒ψ ψ
ponens)
φ
falseIntro false-introduction φ false
¬φ
falseElim false-elimination false φ
reductio ad absurdum ¬φ ` false φ
RAA
(v. 1)
reductio ad absurdum φ ` false ¬φ
(v. 2)
¬Intro negation-introduction φ ¬¬φ
¬Elim negation-elimination ¬¬φ φ
case-elimination (left) φ ∨ ψ ψ
CaseElim
¬φ
case-elimination (right) φ ∨ ψ φ
¬ψ
Table 6.2
As usual, φ, ψ , θ , ω are meta-variables standing for any WFF.
This is by no means the only possible inference system for propositional logic.
aside: This set of inference rules is based upon Discrete Mathematics with a Computer by Hall
and O'Donnell (Springer, 2000) and The Beseme Project
3 .
6.3 rst-order equivalences4
The following equivalences are in addition to those of propositional logic (Section 6.1). In these, φ and ψ
each stand for any WFF, but θ stands for any WFF with no free occurrences of x .
3 http://www.cs.ou.edu/∼beseme/
4 This content is available online at <http://cnx.org/content/m11045/2.18/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
120 CHAPTER 6. APPENDICES AND REFERENCE SHEETS
First-order Logic Equivalences
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
121
Equivalence ∀ Variant ∃ Variant
Complementation of Quanti ers ∀x : (¬φ) ≡ ¬∃x : (φ) ∃x : (¬φ) ≡ ¬∀x : (φ)
Interchanging Quanti ers ∀x : (∀y : (φ)) ≡ ∀y : (∀x : (φ)) ∃x : (∃y : (φ)) ≡ ∃y : (∃x : (φ))
∀x : (φ ∧ ψ) ≡ ∀x : (φ) ∧ ∀x : ∃x : (φ ∨ ψ) ≡ ∃x : (φ) ∨ ∃x :
(ψ) (ψ)
Distribution of Quanti ers
∀x : (φ ∨ θ) ≡ ∀x : (φ) ∨ θ ∃x : (φ ∧ θ) ≡ ∃x : (φ) ∧ θ
∀x : (φ ⇒ θ) ≡ ∃x : (φ) ⇒ θ
∀x : (θ ⇒ φ) ≡ θ ⇒ ∀x : (φ)
∀x : (φ ∧ θ) ≡ ∀x : (φ) ∧ θ ∃x : (φ ∨ θ) ≡ ∃x : (φ) ∨ θ
Distribution of Quanti ers
∃x : (φ ⇒ θ) ≡ ∀x : (φ) ⇒ θ
−−− with non-empty domain
∃x : (θ ⇒ φ) ≡ θ ⇒ ∃x : (φ)
Renaming ∀x : (φ) ≡ ∀y : (φ [x 7→ y]) ∃x : (φ) ≡ ∃y : (φ [x 7→ y])
Simpli cation of Quanti ers −− ∀x : (θ) ≡ θ ∃x : (θ) ≡ θ
− with non-empty domain
Simpli cation of Quanti ers −− ∀x : (φ) ≡ true ∃x : (φ) ≡ false
− with empty domain
Table 6.3
When citing Distribution of Quanti ers, say what you're distributing over what: e.g., distribute ∀ over
∨ (with θ being x-free) .
In renaming (Note, p. 99), the notation φ [x 7→ y] means φ with each free occurrence of x replaced by y
. It is a meta-formula; when writing any particular formula you don't write any brackets, and instead just
do the replacement.
This set of equivalences isn't actually quite complete. For instance, ∃x : (∀y : (R (x, y))) ⇒ ∀y :
(∃x : (R (x, y))) is equivalent to true, but we can't show it using only the rules above. It does become
complete
5 if we add some analogs of the rst-order inference rules (Section 6.4), replacing ` with ⇒ (and
carrying along their baggage of arbitrary and free-to-substitute-in).
6.4 rst-order inference rules6
The following are in addition to those of propositional logic (Section 6.2).
Our rst-order inference rules
Abbreviation Name If you know all of. . . . . .then you can infer
∀Intro ∀-introduction φ ∀x.φ[y7→x]
y arbitrary (p. 99).
continued on next page
5 It's not obvious when this system is complete; that's Gödel's completeness theorem
(<http://wikipedia.org/wiki/G%F6del's_completeness_theorem>), his 1929 Ph.D. thesis. Don't confuse it with his
more celebrated Incompleteness Theorem, on the other hand, which talks about the ability to prove formulas which are true
in all interpretations which include arithmetic (as opposed to all interpretations everywhere.)
6 This content is available online at <http://cnx.org/content/m11046/2.18/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
122 CHAPTER 6. APPENDICES AND REFERENCE SHEETS
∀Elim ∀-elimination ∀x.φ φ[x7→t]
t is any term that is free
to be replaced in φ.
Domain non-empty.
∃Intro ∃-introduction φ ∃x.φ[t7→x] , where t is
arbitrary
t is any term in φ that is
free to be replaced.
Domain non-empty.
∃Elim ∃-elimination ∃x.φ φ[x7→c]
c is a new constant in
the proof.
c does not occur in the
proof 's conclusion.
Table 6.4
As usual, we use φ as a meta-variable to range over rst-order WFFs. Similarly, t is a meta-variable for
rst-order terms, and c is a meta-variable for domain constants. The notation φ[v7→w] means the formula φ
but with every appropriate (Note, p. 99) occurrence of v replaced by w .
As discussed in the lecture notes (p. 99), a variable is arbitrary unless:
• A variable is not arbitrary if it is free in (an enclosing) premise.
• A variable is not arbitrary if it is free after applying ∃Elim either as the introduced witness c, or
free anywhere else in the formula.
The usual way to introduce arbitrary variables is during ∀Elim (w/o later using it in ∃Elim).
As a detail in ∀Elim and ∃Intro, the term t must be free to replace the variable x in φ. This means
that it is not the case that both t contains a variable quanti ed in φ, and that x occurs free within that
quanti er. In short, the bound variable names should be kept distinct from the free variable names. Also,
only free occurrences x get replaced. The restriction in ∃Elim on c being new is similar.
6.5 propositional axioms for WaterWorld7
We summarize the details of how we choose to model WaterWorld boards in propositional logic: exactly
what propositions we make up, and the formal domain axioms which capture the game's rules.
The board is xed at 6×4, named A,. . .,Z (with I and O omitted).
7 This content is available online at <http://cnx.org/content/m10528/2.34/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
123
Figure 6.1: A Sample WaterWorld board
6.5.1 Propositions
There are a myriad of propositions for WaterWorld, which can be grouped:
• Whether or not a location contains a pirate: A − unsafe, B − unsafe, . . ., Z − unsafe.
• Whether or not a location contains no pirate: A − safe, B − safe, . . ., Z − safe.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
124 CHAPTER 6. APPENDICES AND REFERENCE SHEETS
aside: Yes, using the intended interpretation, these are redundant with the previous ones.
Some domain axioms below will formalize this.
• Propositions indicating the number of neighboring pirates, to a location: A − has − 0, A − has − 1,
A − has − 2, B − has − 0, B − has − 1, B − has − 2, . . ., H − has − 0, H − has − 1, H − has − 2,
H − has − 3, . . ., Z − has − 0, Z − has − 1. These are all true/false propositions; there are no ex-
plicit numbers in the logic. A domain axiom below will assert that whenever (say) B − has − 1 is true,
then B − has − 0 and B − has − 2 are both false.
aside: There is no proposition A − has − 3 since location A has only two neighbors.
Similarly, there is no proposition B − has − 3. We could have chosen to include those, but
under the intended interpretation they'd always be false.
These propositions describe the state of the underlying board the model and not our
particular view of it. Our particular view will be re ected in which formulas we'll accept as premises.
So we'll accept A − has − 2 as a premise only when A has been exposed and shows a 2.
6.5.2 The domain axioms
Axioms asserting that the neighbor counts are correct:
• Count of 0:
· A0: A − has − 0 ⇒ B − safe ∧ G − safe
· ...
· H0: H − has − 0 ⇒ G − safe ∧ J − safe ∧ P − safe
· ...
· Z0: Z − has − 0 ⇒ Y − safe
• Count of 1:
· A1: A − has − 1 ⇒ B − safe ∧ G − unsafe ∨ B − unsafe ∧ G − safe
· ...
· H1: H − has − 1 ⇒ G − safe ∧ J − safe ∧ P − unsafe ∨ G − safe ∧ J − unsafe ∧ P − safe ∨
G − unsafe ∧ J − safe ∧ P − safe
· ...
· Z1: Z − has − 1 ⇒ Y − unsafe
• Count of 2:
· A2: A − has − 2 ⇒ B − unsafe ∧ G − unsafe
· ...
· H2: H − has − 2 ⇒ G − safe ∧ J − unsafe ∧ P − unsafe ∨ G − unsafe ∧ J − safe ∧ P − unsafe ∨
G − unsafe ∧ J − unsafe ∧ P − safe
· ...
There aren't any such axioms for locations with only one neighbor.
• Count of 3:
· H3: H − has − 3 ⇒ G − unsafe ∧ J − unsafe ∧ P − unsafe
· ...
There aren't any such axioms for locations with only one or two neighbors.
Axioms asserting that the propositions for counting neighbors are consistent:
• A − has − 0 ∨ A − has − 1
• A − has − 0 ⇒ ¬A − has − 1
• A − has − 1 ⇒ ¬A − has − 0
• B − has − 0 ∨ B − has − 1 ∨ B − has − 2
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
125
• B − has − 0 ⇒ ¬B − has − 1 ∧ ¬B − has − 2
• B − has − 1 ⇒ ¬B − has − 0 ∧ ¬B − has − 2
• B − has − 2 ⇒ ¬B − has − 0 ∧ ¬B − has − 1
• ...
• H − has − 0 ∨ H − has − 1 ∨ H − has − 2 ∨ H − has − 3
• H − has − 0 ⇒ ¬H − has − 1 ∧ ¬H − has − 2 ∧ ¬H − has − 3
• H − has − 1 ⇒ ¬H − has − 0 ∧ ¬H − has − 2 ∧ ¬H − has − 3
• H − has − 2 ⇒ ¬H − has − 0 ∧ ¬H − has − 1 ∧ ¬H − has − 3
• H − has − 3 ⇒ ¬H − has − 0 ∧ ¬H − has − 1 ∧ ¬H − has − 2
• ...
Axioms asserting that the safety propositions are consistent:
• A − safe ⇒ ¬A − unsafe,
• ¬A − safe ⇒ A − unsafe,
• ...
• Z − safe ⇒ ¬Z − unsafe,
• ¬Z − safe ⇒ Z − unsafe.
This set of axioms is not quite complete, as explored in an exercise (Exercise 2.5.13).
As mentioned, it is redundant to have both A − safe and A − unsafe as propositions. Furthermore, having
both allows us to express inconsistent states (ones that would contradict the safety axioms). If implementing
this in a program, you might use both as variables, but have a safety-check function to make sure that a given
board representation is consistent. Even better, you could implement WaterWorld so that these propositions
wouldn't be variables, but instead be calls to a lookup (accessor) functions. These would examine the same
internal state, to eliminate the chance of inconsistent data.
Using only true/false propositions; without recourse to numbers makes these domain axioms unwieldy.
Later, we'll see how relations (Section 3.1) and quanti ers (Section 4.1.1) help us model the game of Water-
World more concisely.
6.6 rst-order axioms for WaterWorld8
We summarize the details of how we choose to model WaterWorld boards in rst-order logic: exactly what
relations we make up, and the formal domain axioms which capture the game's rules.
This will follow almost exactly the same pattern as our WaterWorld model in propositional logic (Sec-
tion 6.5). However, we will take advantage of the additional exibility provided by rst-order logic.
Rather than modeling only the default 6×4 WaterWorld board;, we will be able to model any board
representable by our relations. This will allow boards of any size and con guration, with one major constraint
each location can have at most three neighboring pirates.
6.6.1 Domain and Relations
Our domain is simply the set of all board locations. This set can be arbitrarily large even in nite!
The board con guration is given by the binary neighbor relation nhbr.
The next relations correspond directly to the propositions (Section 6.5.1: Propositions) in the proposi-
tional logic model.
• Whether or not a location contains a pirate: safe. This is a unary relation.
aside: We choose not to include a redundant relation unsafe.
• Unary relations indicating the number of neighboring pirates: has0, has1, has2, and has3.
8 This content is available online at <http://cnx.org/content/m11072/2.12/>.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
126 CHAPTER 6. APPENDICES AND REFERENCE SHEETS
aside: Thus, we have our restriction to three unsafe neighbors. This will also be re ected
in our domain axioms below. See also this problem (Exercise 3.5.1) for a discussion of how to
avoid this restriction.
In addition, to have encode the domain axioms for an arbitrary domain, we also need an equality relation
over our domain of locations. As is traditional, we will use in x notation for this relation, for example,
x = y. Furthermore, we will allow ourselves to write x 6= y as shorthand for ¬ (x = y). Thus, we do not
need a distinct inequality relation.
Note that these relations describe the state of the underlying board the model and not our particular
view of it. Our particular view will be re ected in which formulas we'll accept as premises. So we'll accept
has2 (A) as a premise only when A has been exposed and shows a 2.
6.6.2 The domain axioms
Many of our axioms correspond directly, albeit much more succinctly, with those (Section 6.5.2: The domain
axioms) of the propositional model. In addition, we have axioms that specify that our neighbor and equality
relations are self-consistent.
Axioms asserting that the neighbor relation is anti-re exive and symmetric:
• ∀x : (¬nhbr (x, x))
• ∀x : (∀y : (nhbr (x, y) ⇒ nhbr (y, x)))
Axioms asserting that = truly is an equality relation, i.e., it is re exive, symmetric, and transitive.
• ∀x : (x = x)
• ∀x : (∀y : ((x = y) ⇒ (y = x)))
• ∀x : (∀y : (∀z : ((x = y) ∧ (y = z) ⇒ (x = z))))
Axioms asserting that the neighbor counts are correct. Each of these is of the form if location x has n
neighboring pirates, then there are n distinct unsafe neighbors of x, and any other distinct neighbor x is
safe. We use the equality relation to specify the distinctness of each neighbor.
• ∀x : (has0 (x) ⇒ ∀y : (nhbr (x, y) ⇒ safe (y)))
• ∀x : (has1 (x) ⇒ ∃a : (nhbr (x, a) ∧ ¬safe (a) ∧ ∀y : (nhbr (x, y) ∧ (a 6= y) ⇒ safe (y))))
• ∀x : (has2 (x) ⇒ ∃a : (∃b : (nhbr (x, a) ∧ nhbr (x, b) ∧ (a 6= b) ∧ ¬safe (a) ∧ ¬safe (b) ∧ ∀y : (nhbr (x, y) ∧ (a 6=
• ∀x : (has3 (x) ⇒ ∃a : (∃b : (∃c : (nhbr (x, a) ∧ nhbr (x, b) ∧ nhbr (x, c) ∧ (a 6= b) ∧ (a 6= c) ∧ (b 6= c) ∧ ¬safe (a
In addition, we want the implications to go the opposite way. Otherwise, each of has0, has1, has2, and
has3 could always be false, while still satisfying the above! For brevity, we elide the details in the following
list:
• ∀x : (∀y : (nhbr (x, y) ⇒ safe (y)) ⇒ has0 (x))
• ∀x : (. . . ⇒ has1 (x))
• ∀x : (. . . ⇒ has2 (x))
• ∀x : (. . . ⇒ has3 (x))
Axioms asserting that the neighbor counts are consistent. While redundant, including axioms like the
following can be convenient.
• ∀x : (has0 (x) ⇒ ¬ (has1 (x) ∨ has2 (x) ∨ has3 (x)))
• ∀x : (has1 (x) ⇒ ¬ (has0 (x) ∨ has2 (x) ∨ has3 (x)))
• ∀x : (has2 (x) ⇒ ¬ (has0 (x) ∨ has1 (x) ∨ has3 (x)))
• ∀x : (has3 (x) ⇒ ¬ (has0 (x) ∨ has1 (x) ∨ has2 (x)))
Note that this set of axioms is not quite complete, as explored in an exercise (Exercise 4.4.22).
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
127
6.7 browser support9
warning: The information in this module is outdated. Please see my course
10 for a table of
contents.
note: This page meant to be viewed with a MathML-enabled browser. If you see (∀x. (P(x) →
(∃y. (P(y) ∨ φ)))) as a nice version of (forall x . (P(x) -> (exists y . (P(y) v phi)))) you're doing
okay; If you further see A `B as a nice version of (scriptA |- scriptB) you're set! If not, see our
description of browser support (Section 6.7) .
At Rice on the CSNet, use mozilla. Preferably, use version 1.1, as currently available on Solaris 8
machines. On frosty.cs, version 1.1 is the default. On other Solaris 8 machines, version 1.1 is not yet the
default, but available via /opt1/mozilla-1.1/sunos5/bin/mozilla.
In general, to view TeachLogic web pages, you'll need a browser that supports the following features:
• Cascading Style Sheets (CSS)
11 Most recent browsers support CSS su ciently well.
• MathML
12 Some browsers support MathML su ciently well. However, most (all?) do not fully
support Unicode Plane 1 numerical entity references, which includes most mathematical alphanumeric
characters.
• Math-oriented fonts
13
Which browsers support these features? The above links provide more details, but here's a summary of some
browsers.
• Mozilla
14 and Netscape15 (version 7.0) Both work, except some characters (Unicode Plane 1) don't
appear correctly.
• Internet Explorer
16 is not yet an option, even with the MathPlayer plug-in17 to view MathML. IE
won't display pages with some characters (Unicode Plane 1).
Alternatively, PDF versions of the web pages are also provided via the Base module's index
18 .
9 This content is available online at <http://cnx.org/content/m10845/2.3/>.
10 http://cnx.rice.edu/content/col10154/latest/
11 http://www.w3.org/Style/CSS/
12 http://www.w3.org/Math/
13 http://www.mozilla.org/projects/mathml/fonts/
14 http://www.mozilla.org/
15 http://channels.netscape.com/ns/browsers/download_other.jsp/
16 http://www.microsoft.com/windows/ie/
17 http://www.dessci.com/webmath/mathplayer/
18 "Introduction: outline" <http://cnx.org/content/m10773/latest/>
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
128 GLOSSARY
Glossary
C completeness
If something really is true, the system is capable of proving it.
connective
1. The syntactic operator combining one or more logical expressions into a larger expression.
Example: Two operators are ∧ and ∨.
2. A function with one or more Boolean inputs and a Boolean result. I.e., the meaning of a
syntactic operator.
Example: The meaning of ∧ and ∨, e.g., as described by their truth tables.
Example: nand (mnemonic: not and), written ↑, takes in two Boolean values a and b, and
returns true exactly when a ∧ b is not true that is, a ↑ b ≡ ¬ (a ∧ b).
I Interpretation
The interpretation of a formula is a domain, together with a mapping from the formula's relation
symbols to speci c relations on the domain.
P proposition
A statement which can be either true or false.
Example: Your meal will include hashbrowns.
propositional variable
A variable that can either be true or false, representing whether a certain proposition is true or
not.
Example: HB
S soundness
If the system (claims to) prove something is true, it really is true.
T tautology
A WFF which is true under any truth assignment (any way of assigning true/false to the
propositions).
Example: A − unsafe ⇒ A − unsafe
Example: a⇒a ∨ b
term
1. A variable.
Example: a, b, . . .
2. A constant.
Example: WaterWorld location F, Kevin Bacon, or the number 3.
3. A function applied to one or more terms.
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
GLOSSARY 129
Example: successor (3)
truth assignment
An assignment of a value true or false to each proposition being used.
Example: For the formula a ⇒ a ∧ b, one possible truth assignment is a = true and b = false.
With that truth assignment, the formula is false.
truth table
A truth table for an expression has a column for each of its propositional variables. It has a row
for each dierent true/false combination of its propositional variables. It has one more column
for the expression itself, showing the truth of the entire expression for that row.
U unsatis able
A WFF which is false under any truth assignment.
Example: ¬ (A − unsafe ⇒ A − unsafe)
Example: a ⇒ ¬a
W Well-Formed formula (WFF)
1. A constant: true or false. (If you prefer brevity, you can write T or F.)
2. A propositional variable.
Example: a
3. A negation ¬φ, where φ is a WFF.
Example: ¬c
4. A conjunction φ ∧ ψ, where φ and ψ are WFFs.
Example: a ∧ ¬c
5. A disjunction φ ∨ ψ, where φ and ψ are WFFs.
Example: ¬c ∨ a ∧ ¬c, or equivalently, (¬c) ∨ (a ∧ ¬c)
6. An implication φ ⇒ ψ, where φ and ψ are WFFs.
Example: ¬c ∨ a ∧ ¬c ⇒ b, or equivalently, ((¬c) ∨ (a ∧ ¬c)) ⇒ b
Well-Formed Formula (WFF) for rst-order logic
1. A constant: true or false.
2. An atomic formula: a relation symbol applied to one or more terms.
Example: nhbr (x, F )
3. A negation of a WFF, ¬φ.
4. A conjunction of WFFs, φ ∧ ψ.
5. A disjunction of WFFs, φ ∨ ψ.
6. An implication of WFFs, φ ⇒ ψ.
7. A universal quanti cation of a WFF, ∀x : (φ).
Example: ∀x : (nhbr (x, F ))
8. An existential quanti cation of a WFF, ∃x : (φ).
Example: ∃x : (nhbr (x, F ))
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
130 INDEX
Index of Keywords and Terms
Keywords are listed by the section with that keyword (page numbers are in parentheses). Keywords
do not necessarily appear in the text of the page. They are merely associated with that section. Ex.
apples, 1.1 (1) Terms are referenced by the page they appear on. Ex. apples, 1
0 0 = 1, 1.1(1) existential quanti cation, 129
exists, 4.1.1(89)
9 90 = 100, 1.1(1)
F false, 1.1(1)
A Acknowledgements, 5.2(116)
rst-order, 4.1.1(89)
arbitrary, 100, 100, 122
rst-order inference rules, 4.3.1(98)
arity, 77
rst-order logic, 75, 91
at least a strong, 56
for all, 4.1.1(89)
atomic formula, 129
formal proof, 37
axiom, 37
free, 94
B Beseme, 5.2(116)
free to replace, 122
Fuzzy Logic, 98
binary relation, 76
binary relations, 75
G geometry, 1.1(1)
Boolean algebra, 28, 4.2.1(95)
graph, 78, 106
boolean identity, 4.2.1(95)
bound, 94 I if-and-only-if, 20
i, 20
C CNF, 32
inclusive-or, 20
CNF clauses, 32
incomplete, 69
completeness, 35, 35
indicator function, 77
Conjunctive Normal Form, 32
inference rule, 9, 38, 6.2(118)
connective, 19
Interpretation, 79, 3.4(80)
contingency, 23
Intro, 6.2(118)
contrapositive, 30, 40
isomorphic, 105
converse, 31
Curry-Howard isomorphism, 115 K k-ary, 77
D DeMorgan's laws, 29 L Law of Excluded Middle, 60
Disjunctive Normal Form, 32 lemmas, 45
DNF, 32 Linear logic, 115
DNF clauses, 32
domain, 76
M meta-variables, 21
modus ponens, 31
domain axiom, 25
modus tollens, 41
domain axioms, 37
E edge, 106
N nand, 128
nodes, 106
edges, 78
non-standard interpretation, 3.4(80)
Elim, 6.2(118)
nonstandard interpretation, 3.4(80)
EQUIV, 25
nor, 58
Equivalence, 34
NSF, 5.2(116)
equivalent, 22, 28
exclusive-or, 20
O opportunity cost, 114
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
INDEX 131
oracle, 3.4(80), 82 stronger, 56
students, 5.2(116)
P P vs NP, 3.4(80)
sum-of-products, 32
paradox, 1.1(1)
syllogism, 8
Poincaré disc, 3.4(80)
syntax, 28
Poincare disc, 3.4(80)
polymorphic, 101 T tautology, 23, 24, 34
predicate logic, 91 Temporal logic, 115
premises, 37 term, 91
Prenex CNF, 95 terms, 77, 91
Prenex Conjunctive Normal Form, 95 the law of the excluded middle, 98
Prenex Disjunctive Normal Form, 95 theorem, 55
Prenex DNF, 95 truth assignment, 22
product-of-sums, 32 truth table, 19
proof, 1.1(1) Type checking, 6
proof-by-contradiction, 40 type safe, 49
proposition, 18
propositional logic, 6.2(118)
U unary, 77, 86
unique factorization into primes, 3.4(80)
propositional variable, 18
universal quanti cation, 129
Q quanti er, 4.1.1(89), 4.2.1(95) universe, 76
quanti ers, 89 unsatis able, 23
query, 52
V vacuously true, 97
R reduce, 34 valid, 92
relation modeling waterworld interpretation, vertices, 78
3.3(79)
relation subset domain binary re exive
W weaker, 56
Well-Formed formula (WFF), 20
symmetric antisymmetric transitive, 3.2(76)
Well-Formed Formula (WFF) for rst-order
relativized, 3.4(80)
logic, 91
rhetoric, 114
witness, 99
S SAT, 25
second-order logic, 114
X xnor, 55
xor, 20
semantics, 28
sentences, 94
α α-renaming, 95
shadows, 94
soundness, 35, 35 ∃ ∃Intro, 99
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
132 ATTRIBUTIONS
Attributions
Collection: Intro to Logic
Edited by: Ian Barland, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen, John Greiner
URL: http://cnx.org/content/col10154/1.20/
License: http://creativecommons.org/licenses/by/1.0
Module: "90 = 100: A Proof"
By: Ian Barland, Moshe Vardi, Phokion Kolaitis, Matthias Felleisen, John Greiner
URL: http://cnx.org/content/m12727/1.8/
Pages: 1-4
Copyright: Ian Barland, Moshe Vardi, Phokion Kolaitis, Matthias Felleisen, John Greiner
License: http://creativecommons.org/licenses/by/2.0/
Module: "Introduction: logic motivation"
Used here as: "the need for proofs"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10714/2.36/
Pages: 4-7
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Introduction: logic de nition"
Used here as: "de ning a proof"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12072/1.12/
Pages: 8-14
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: propositions"
Used here as: "propositions"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10715/2.58/
Pages: 17-20
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: formulas"
Used here as: "formulas"
By: Ian Barland, John Greiner, Moshe Vardi, Phokion Kolaitis, Matthias Felleisen
URL: http://cnx.org/content/m12073/1.16/
Pages: 20-25
Copyright: Ian Barland, John Greiner
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: truth tables"
Used here as: "using truth tables"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10716/2.35/
Pages: 25-26
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
ATTRIBUTIONS 133
Module: "Propositional Logic: limitations of truth tables"
Used here as: "the limitations of truth tables"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12074/1.7/
Pages: 26-28
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: equivalences"
Used here as: "propositional equivalences"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10717/2.40/
Pages: 28-32
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: normal forms"
Used here as: "normal forms"
By: Ian Barland, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12075/1.12/
Pages: 32-34
Copyright: Ian Barland, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: soundness and completeness"
Used here as: "soundness and completeness"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12076/1.10/
Pages: 34-35
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: inference rules"
Used here as: "propositional inference rules"
By: Ian Barland, John Greiner
URL: http://cnx.org/content/m10718/2.35/
Pages: 35-40
Copyright: Ian Barland, John Greiner
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: subproofs"
Used here as: "using subproofs"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12077/1.12/
Pages: 40-48
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
134 ATTRIBUTIONS
Module: "Propositional Logic: soundness and completeness revisited"
Used here as: "the soundness and completeness of inference rules"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12078/1.10/
Pages: 48-49
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: type checking"
Used here as: "proofs and programming"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12079/1.7/
Page: 49
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Propositional Logic: conclusions"
Used here as: "conclusions"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12080/1.11/
Pages: 49-50
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Exercises for Propositional Logic I"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10514/2.57/
Pages: 50-58
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Exercises for Propositional Logic II"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12352/1.20/
Pages: 59-63
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Relations and Logic: using relations"
Used here as: "relations"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10724/2.25/
Pages: 75-76
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Relations and Logic: properties of relations"
Used here as: "properties of relations"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10725/2.28/
Pages: 76-78
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
ATTRIBUTIONS 135
Module: "Relations and Logic: interpretations"
Used here as: "interpretations"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10726/2.23/
Pages: 79-80
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Relations and Logic: Non-standard Interpretations"
Used here as: "Nonstandard Interpretations (optional)"
By: Ian Barland, Moshe Vardi, Phokion Kolaitis, Matthias Felleisen, John Greiner
URL: http://cnx.org/content/m12741/1.3/
Pages: 80-83
Copyright: Ian Barland
License: http://creativecommons.org/licenses/by/2.0/
Module: "Relations and Logic: modeling"
Used here as: "modeling with relations"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10727/2.24/
Pages: 83-85
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "First-Order Logic: using quanti ers"
Used here as: "syntax and semantics of quanti ers"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10728/2.35/
Pages: 89-94
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "First-Order Logic: bound variables, free variables"
Used here as: "bound variables, free variables"
By: Ian Barland, John Greiner, Phokion Kolaitis, Matthias Felleisen, Moshe Vardi
URL: http://cnx.org/content/m12081/1.7/
Pages: 94-95
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Matthias Felleisen, Moshe Vardi
License: http://creativecommons.org/licenses/by/1.0
Module: "First-Order Logic: normal forms revisited"
Used here as: "normal forms revisited"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m12082/1.5/
Page: 95
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
136 ATTRIBUTIONS
Module: "First-Order Logic: equivalences"
Used here as: " rst-order equivalences"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10729/2.26/
Pages: 95-98
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "First-Order Logic: inference rules"
Used here as: " rst-order inference rules"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10774/2.27/
Pages: 98-102
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Exercises for First-Order Logic"
By: Ian Barland, John Greiner, Fuching Chi
URL: http://cnx.org/content/m12353/1.28/
Pages: 102-109
Copyright: Ian Barland, John Greiner
License: http://creativecommons.org/licenses/by/1.0
Module: "Conclusion"
Used here as: "Logic: Looking Back"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10775/2.20/
Pages: 113-115
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Acknowledgements"
By: Ian Barland
URL: http://cnx.org/content/m13613/1.1/
Page: 116
Copyright: Ian Barland
License: http://creativecommons.org/licenses/by/2.0/
Module: "Reference: propositional equivalences"
Used here as: "propositional equivalences"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10540/2.25/
Page: 117
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Reference: propositional inference rules"
Used here as: "propositional inference rules"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10529/2.28/
Pages: 118-119
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
ATTRIBUTIONS 137
Module: "Reference: rst-order equivalences"
Used here as: " rst-order equivalences"
By: John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m11045/2.18/
Pages: 119-121
Copyright: John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Reference: rst-order inference rules"
Used here as: " rst-order inference rules"
By: John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m11046/2.18/
Pages: 121-122
Copyright: John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Reference: propositional WaterWorld"
Used here as: "propositional axioms for WaterWorld"
By: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m10528/2.34/
Pages: 122-125
Copyright: Ian Barland, John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Reference: rst-order WaterWorld"
Used here as: " rst-order axioms for WaterWorld"
By: John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
URL: http://cnx.org/content/m11072/2.12/
Pages: 125-126
Copyright: John Greiner, Phokion Kolaitis, Moshe Vardi, Matthias Felleisen
License: http://creativecommons.org/licenses/by/1.0
Module: "Browser support"
Used here as: "browser support"
By: Ian Barland, John Greiner
URL: http://cnx.org/content/m10845/2.3/
Page: 127
Copyright: Ian Barland, John Greiner
License: http://creativecommons.org/licenses/by/1.0
Available for free at Connexions <http://cnx.org/content/col10154/1.20>
Intro to Logic
An introduction to reasoning with propositional and rst-order logic, with applications to computer science.
Part of the TeachLogic Project (www.teachlogic.org).
About Connexions
Since 1999, Connexions has been pioneering a global system where anyone can create course materials and
make them fully accessible and easily reusable free of charge. We are a Web-based authoring, teaching and
learning environment open to anyone interested in education, including students, teachers, professors and
lifelong learners. We connect ideas and facilitate educational communities.
Connexions's modular, interactive courses are in use worldwide by universities, community colleges, K-12
schools, distance learners, and lifelong learners. Connexions materials are in many languages, including
English, Spanish, Chinese, Japanese, Italian, Vietnamese, French, Portuguese, and Thai. Connexions is part
of an exciting new information distribution system that allows for Print on Demand Books. Connexions
has partnered with innovative on-demand publisher QOOP to accelerate the delivery of printed course
materials and textbooks into classrooms worldwide at lower prices than traditional academic publishers.
You might also like
- Grade 10 Mathematics 2nd Term Test Paper 2019 English Medium North Central Province100% (1)Grade 10 Mathematics 2nd Term Test Paper 2019 English Medium North Central Province13 pages
- Introduction To Logic 3rd Edition Michael Genesereth Instant Download100% (4)Introduction To Logic 3rd Edition Michael Genesereth Instant Download56 pages
- Calvin Jongsma - Introduction To Discrete Mathematics Via Logic and Proof (Instructor's Solution Manual) (Solutions) - Springer (2019)No ratings yetCalvin Jongsma - Introduction To Discrete Mathematics Via Logic and Proof (Instructor's Solution Manual) (Solutions) - Springer (2019)227 pages
- Building Blocks For Theoretical Computer ScienceNo ratings yetBuilding Blocks For Theoretical Computer Science273 pages
- Discrete Mathematics and Functional Programming PDF100% (1)Discrete Mathematics and Functional Programming PDF247 pages
- Michael Genesereth, Introduction To Logic-Morgan (2017)100% (1)Michael Genesereth, Introduction To Logic-Morgan (2017)161 pages
- Coq - Modeling and Proving in Computational Type TheoryNo ratings yetCoq - Modeling and Proving in Computational Type Theory433 pages
- MATH 135 Reading, Discovering, and Writing Proofs E-BookNo ratings yetMATH 135 Reading, Discovering, and Writing Proofs E-Book342 pages
- An Introduction To Proof Theory. Normalization, - Paolo Mancosu, Sergio Galvan, Richard Zach - 2021 - Oxford University PressNo ratings yetAn Introduction To Proof Theory. Normalization, - Paolo Mancosu, Sergio Galvan, Richard Zach - 2021 - Oxford University Press430 pages
- Introduction To Formal Reasoning (COMP2065) : Release 0.1No ratings yetIntroduction To Formal Reasoning (COMP2065) : Release 0.187 pages
- (Micha L. Walicki) Introduction To Logic (Bookos - Org) 181pp-2006No ratings yet(Micha L. Walicki) Introduction To Logic (Bookos - Org) 181pp-2006181 pages
- Building Blocks For Theoretical Computer ScienceNo ratings yetBuilding Blocks For Theoretical Computer Science268 pages
- Two Categories of Refutation Decision Procedures For Classical and Intuitionistic LogicsNo ratings yetTwo Categories of Refutation Decision Procedures For Classical and Intuitionistic Logics156 pages
- Introduction To Formal Reasoning (COMP2065) : Release 0.1No ratings yetIntroduction To Formal Reasoning (COMP2065) : Release 0.179 pages
- Relevant Logic A Philosophical Examination of Inference Read PDF DownloadNo ratings yetRelevant Logic A Philosophical Examination of Inference Read PDF Download52 pages
- C25 MATH Practice Sheet - Complex NumberNo ratings yetC25 MATH Practice Sheet - Complex Number26 pages
- Buy Ebook Geometric Patterns With Creative Coding: Coding For The Arts 1st Edition Selçuk Artut Cheap Price100% (1)Buy Ebook Geometric Patterns With Creative Coding: Coding For The Arts 1st Edition Selçuk Artut Cheap Price57 pages
- Arxiv:Math/0601709V1 (Math - GM) 29 Jan 2006No ratings yetArxiv:Math/0601709V1 (Math - GM) 29 Jan 2006124 pages
- DFMFullCoverage-SolvingLinearEquations 2No ratings yetDFMFullCoverage-SolvingLinearEquations 26 pages
- The Absolutely True Diary of A Part Time IndianNo ratings yetThe Absolutely True Diary of A Part Time Indian4 pages
- UKMT - IMC - Intermediate Mathematical Challenge 2007 - SolutionsNo ratings yetUKMT - IMC - Intermediate Mathematical Challenge 2007 - Solutions10 pages
- Summative and Reflective Portfolio (Math)No ratings yetSummative and Reflective Portfolio (Math)14 pages
- Mathematics Tamil English WWW - Governmentexams.co - in PDFNo ratings yetMathematics Tamil English WWW - Governmentexams.co - in PDF16 pages
- 2020 Euclid Contest: The Centre For Education in Mathematics and Computing Cemc - Uwaterloo.caNo ratings yet2020 Euclid Contest: The Centre For Education in Mathematics and Computing Cemc - Uwaterloo.ca15 pages
- Victoria Vaughn - Codominance, Incomplete Dominance and Multiple Alleles - SinghNo ratings yetVictoria Vaughn - Codominance, Incomplete Dominance and Multiple Alleles - Singh7 pages
- Triangles and Its Properties: Master Maths From HomeNo ratings yetTriangles and Its Properties: Master Maths From Home1 page
- Research Project: Writing Your Lab ReportNo ratings yetResearch Project: Writing Your Lab Report7 pages
- Options Trading for Income: Learn the strategies and techniques for maximizing returns and minimizing risk in the options market (2023 Guide for Beginners)From EverandOptions Trading for Income: Learn the strategies and techniques for maximizing returns and minimizing risk in the options market (2023 Guide for Beginners)No ratings yet
