Extending and Automating Basic Probability

Theory with Propositional Computability

Logic
Keehang Kwon
arXiv:1909.07375v1 [cs.AI] 16 Sep 2019

[email protected]

Abstract: Classical probability theory is formulated using sets. In this pa-

per, we extend classical probability theory with propositional computability
logic[1] (CoL). Unlike other formalisms, computability logic is built on the
notion of events/games, which is central to probability theory. The prob-
ability theory based on CoL is therefore useful for automating uncertainty
reasoning. We describe some basic properties of this new probability theory.

1 Introduction
Classical probability theory[2] is formulated using sets. Unfortunately, the
language of sets lacks expressiveness and is, in a sense, a low-level ‘assembly
language’ of the probability theory. In this paper, we develop a ‘high-level
approach’ to classical probability theory with propositional computability
logic[1] (CoL). Unlike other formalisms such as sets, logic and linear logic,
computability logic is built on the notion of events/games, which is central
to probability theory. Therefore, CoL is a perfect place to begin the study
of automating probability theory.
To be specific, CoL is well-suited to describing complex (sequential/parallel)
experiments and events, and more expressive than set operations. In con-
trast, classical probability theory – based on ∩, ∪, etc – is designed to repre-
sent mainly the simple/additive events – the events that occur under a single
experiment.
Naturally, we need to talk about composite/multiplicative events – events
that occur under two different experiments. Developing probability along
this line requires a new, powerful language. For example, consider the fol-
lowing events E1 , E2 :

E1 : toss a coin two times (events 1 and 2) and get H,T in that order.
E2 : toss two dices (which we call 1, 2) and get at least one 5.

1
Suppose a formalism has the notion of △, ▽ (sequential-and/or) and ∧ , ∨
(parallel-and/or). Then E1 would be written as H d △T d . Similarly, E2 would
be written concisely as (51 ∨ 52 ). The formalism of classical probability theory
fails to represent the above events in a concise way.
Computability logic[1] provides a formal and consistent way to represent
a wide class of experiments and events. In particular, multiplicative experi-
ments (sequential and parallel experiments) as well as additive experiments
(choice-AND, choice-OR) can be represented in this formalism.

2 The Event Formulas

The language is a variant of the propositional computability logic. For sim-
plicity, we will not include sequential operators. The class of event formulas
is described by E-formulas given by the syntax rules below:

E ::= a(t1 , . . . , tn ) | ¬E | E ⊔ E | E ⊓ E | (E|E)

| E kE | E ∧E | E ∨E

In the above, a(t1 , . . . , tn ) represents an atomic event called proposition where

each ti is a term. For example, d(6) represent the event where we get 6 from
tossing a dice d. For readability, we often write 6d (or simply 6) instead of
d(6).
Often events can be composed through the use of logical connectives. We
will describe the definitions of these connectives.
First, we assume here experiments to be generated in stages. Thus our
sample space is a tree diagram of the form

{(a1 , . . . , an ), (b1 , . . . , bn ), . . .}

where each a, b is an atomic event. As we shall see, this can also be repre-
sented as an event formula in set normal form of the form

(a1 ∧ . . . ∧ an ) ⊔ (b1 ∧ . . . ∧ bn ) ⊔ . . .

In the sequel, we use the above two forms interchangeably.

An event space is a subset of the sample space. In the sequel, we in-
troduce a mapping ∗ which converts an event formula E to an event space

2
{E1 , . . . , En } of mutually exclusive points. This mapping makes it much
easier to compute the probability of E. That is, p(E) = p(E1 ) + . . . + p(En ).
To mentioned earlier, a(t) represents an atomic event. In addition,

a(t)∗ = {a(t)}

The event ¬E represents the complement of the event E relative to the

universe U.

(¬A)∗ = U − A∗

The choice-OR event A ⊔ B represents the event in which only one of the
event A and event B happen under a single experiment. For example, 4d ⊔ 5d
represents the event that we get either 4 or 5 when a dice d is tossed. This
operation corresponds to the set union operation.

(A ⊔ B)∗ = A∗ ∪ B ∗

The choice-AND event A ⊓ B represents the event in which both event A

and event B happen under a single experiment. For example, (2 ⊔ 4 ⊔ 6) ⊓ (2 ⊔ 3 ⊔ 5)
represents the event that we get both an even number and a prime number
in a single coin toss. This operation corresponds to the set intersection op-
eration.

(A ⊓ B)∗ = A∗ ∩ B ∗

The additive conditional event A|B represents the dependency between

A and B: the event in which the event A happens given B has occurred
under a single experiment.
We can generalize the definition of events to joint events to deal with
multiple experiments. Here are some definitions.
The parallel-AND event A ∧ B represents the event in which both A and
B occur under two different experiments. For example, (H 1 ∧ T 2 ) ⊔ (H 2 ∧ T 1 )
represents the event that we get one head and one tail when two coins are
tossed. It is defined by the following:

(A ∧ B)∗ = A∗ × B ∗

Here, the (flattened) Cartesian conjunction of two sets A and B, A × B

is the following:

3
 A × B = {((a1 , . . . , am , b1 , . . . , bn ))|(a1 , . . . , am ) ∈ A and (b1 , . . . , bn ) ∈ B}

For example, let A = {(0, 1), (1, 2)} and B = {0, 1}. Then A × B =
{(0, 1, 0), (0, 1, 1), (1, 2, 0), (1, 2, 1)}.
The parallel-OR event A ∨ B represents the event in which at least one of
event A and event B happen under two different experiments. For example,
((41 ⊔ 51 ) ∨ (42 ⊔ 52 ) represents the event that we get at least one 4 or one 5
when two dices are tossed. Formally,

(A ∨ B) =def (A ∧ B) ⊔ (¬A ∧ B) ⊔ (A ∧ ¬B)

The parallel conditional event A k B (usually written as B → A in logic)

represents the event in which the event A happens given B has occurred
before/after A under two experiments.
The following theorem substantially extends the traditional probability
theory with new properties. These new properties are obtained by consider-
ing relative frequencies.

Theorem 2.1 Let A be an event. Then the following properties hold.

(1) p(¬A) = 1 − p(A) % complement of A

(2) p(A ⊔ B) = p(A) + p(B) − p(A ⊓ B) % choice-or

For example, H ⊔ T represents the event that H or T occur in a single
coin toss. Now, it is easy to see that p(H ⊔ T ) = 1.

(3) p(A ⊓ B) = p(A)p(B|A) = p(B)p(A|B) % choice-and

For example, H ⊓ T represents the event that H and T occur in a sin-
gle coin toss. Now, p(H ⊓ T ) = 0. Note also that p(H ⊓ H) = p(H).
p(A|B) is the conditional probability of B given A in a single experi-
ment. For example, H|T represents the event that H occurs given T
occurs in a single coin toss. Now, p(H|T ) = 0.

(4) p(A ∨ B) = p(A ∧ B) + p(¬A ∧ B) + p(A ∧ ¬B) =

1 − p(¬A ∧ ¬B) % parallel-or
For example, p(H ∨ 6) = 1 − p(T ∧ (1 ⊔ 2 ⊔ 3 ⊔ 4 ⊔ 5)) = 1 − 10/24 =
14/24.

4
(5) p(A ∧ B) = p(A)p(B k A) = p(B)p(A k B) % parallel-and
For example, suppose two coins are tossed. Now, p(H 1 ∧ H 2 ) = 1/4.
In addition, suppose a coin and a dice are tossed simultaneouly. Then
p(H ∧ 6) = p(H)p(6) = 1/12.
% Below, p computes the probability of an event space rather than an
event formula.

(6) p(A) = p(A∗ ) % A* is the event space of A.

For example, ((3 ⊔ 4) ⊓ 4)∗ is {4}.

(7) p({E1 , . . . , En }) = p({E1 }) + . . . + p({En }) % event space with n mu-

tually exclusive elements.

(8) p({(a1 , . . . , an )}) = p(a1 ∧ . . . ∧ an ) % event space with single element.

Here, each ai is an atomic event.

In the above, we list some properties of our probability theory. In addi-

tion, we sometimes need to deal with sharing experiments among two ex-
periments. Sharing experiments represents experiments which have been
completed. Mathematical facts such as prime(2), odd(3) are such examples.
For this reason, we need to include the following:

p(A ∧ A) = p(A ∨ A) = p(A).

For example, let A be a (uncertain) statement that there are aliens in

Vega. Then it is easy to see that these rules hold, as every occurrence of A
represents the same event here.

3 Examples
Let us consider the following event E where E = roll a dice and get 4 or 5.
The probabilities of E and E ∨ E is the following:

p(E) = p(4 ⊔ 5) = p(4) + p(5) = 1/3

p(E1 ∨ E2 )= 1- (2/3 x 2/3) = 5/9.

5
 As another example, (H 1 ∧ H 2 ) ⊔ (H 1 ∧ T 2 ) ⊔ (T 1 ∧ H 2 ) represents the event
that at least one head comes up when two coins are tossed. Now, it is easy
to see that p((H 1 ∧ H 2 ) ⊔ (H 1 ∧ T 2 ) ⊔ (T 1 ∧ H 2 )) = 3/4.

As the last example, suppose two dice are tossed and we get 6 from one
dice. What is the probability that we get 5 from another dice? This kind
of problem is very cumbersome to represent/ solve in classical probability
theory. Fortunately, it can be represented/solved from the above formula in
a concise way. It is shown below:
% computing the following probability requires converting the event to
its event space.

p(((61 ∧ 52 ) ⊔ (62 ∧ 51 )) ⊓ (61 ∨ 62 )) =

p({(5, 6), (6, 5)} ∩ {(6, 1), . . . , (6, 6), (1, 6), . . . , (5, 6)}) =
p({(5, 6), (6, 5)}) = p(5 ∧ 6) + p(6 ∧ 5) = 2/36

% computing the following does not require converting the event to its
event space.

p(61 ∨ 62 ) = p(61 ∧ 62 ) + p(¬61 ∧ 62 ) + p(61 ∧ ¬62 ) = 11/36

From these two, we obtain the following in a purely algorithmic way:

p(((61 ∧ 52 ) ⊔ (62 ∧ 51 ))|(61 ∨ 62 )) = 2/11

4 Two Versions of the Bayes Rule

In this section, we raise questions related to the interpretation of ∩ in the
Bayes rule. Most textbooks interpret ∩ in an ambiguous, confusing way:
sometimes as ⊓ , and as ∧ in others. This is problematic, especially in
automating probabilistic inference.
In considering automation of probability, it is very problematic/cumbersome
to use the Bayes rule in its current form. Suppose A can be partitioned into

6
k disjoint events A1 , . . . , Ak . Understanding ∩ as ⊓ , the Bayes rule can be
written as:
p(Ai ⊓ B)
p(Ai |B) =
p(A1 ⊓ B) + . . . + p(Ak ⊓ B)
or
p(Ai ⊓ B)
p(Ai |B) =
p(A1 )p(B|A1 ) + . . . + p(Ak )p(B|Ak )

This rule can easily be generalized to ∧ as follows:

p(Ai ∧ B)
p(Ai k B) =
p(A1 ∧ B) + . . . + p(Ak ∧ B)
or
p(Ai ∧ B)
p(Ai k B) =
p(A1 )p(B k A1 ) + . . . + p(Ak )p(B k Ak )

That is, we need two versions of the Bayes rule and it is crucial to apply
the correct version to get the correct answer. As a well-known example of
the Bayes rule, consider the problem of sending 0 or 1 over a noisy channel.
Let r(0) be the event that a 0 is received. Let t(0) be the event that a 0
is transmitted. Let r(1) be the event that a 1 is received. Let t(1) be the
event that a 1 is transmitted. Now the question is: what is the probability
of 0 having been transmitted, given 0 is received? In solving this kind of
problem, it is more natural to use Bayes rule on ∧ , rather than on ⊓ . This
is so because p(t(0) ∩ r(0)) = p(t(0) ⊓ r(0)) = 0.

5 Conclusion
Computability logic [1] provides a formal and consistent way to represent
a wide class of experiments and events. For this reason, we believe that
probability theory based on computability logic is an interesting alternative
to the traditional probability theory and uncertainty reasoning.

7
References
[1] G. Japaridze, “Propositional computability logic II”, ACM Transaction
on Computational Logic, vol.7(2), pp.331–362, 2006.

[2] R.D. Yates and D.J.Goodman, “Probability and Stochastic Processes”,

Wiley, 1999.