Quantifying Uncertainty

ACTING UNDER UNCERTAINTY

 Agents may need to handle uncertainty, whether due to partial
observability, nondeterminism, or a combination of the two.
 An agent may never know for certain what state it’s in or where
it will end up after a sequence of actions.
 Summarizing uncertainty:
 Let’s consider an example of uncertain reasoning: diagnosing a
dental patient’s toothache.
 Toothache ⇒ Cavity
 Toothache ⇒ Cavity ∨ GumProblem ∨ Abscess . . .
 Cavity ⇒ Toothache
 Trying to use logic to cope with a domain like medical
diagnosis thus fails for three main reasons:
 Laziness:
 Theoretical ignorance:
 Practical ignorance:
 Trying to use logic to cope with a domain like medical
diagnosis thus fails for three main reasons:
 Laziness:
 Theoretical ignorance:
 Practical ignorance:
 The agent’s knowledge can at best provide only a degree of
belief in the relevant sentences.
 Our main tool for dealing with degrees of belief is probability
theory.
 Probability provides a way of summarizing the uncertainty that
comes from our laziness and ignorance, thereby solving the
qualification problem.
 Uncertainty and rational decisions:
 Example: Automated Taxi
 Plan 1: A90 – leave 90 minutes earlier
 Plan 2: A180 – leave 180 minutes earlier
 Plan 3: A1440 – leave 24 hours earlier
 Which is rational choice?
 To make such choices, an agent must first have preferences between the
different possible outcomes of the various plans.
 Probability provides a way of summarizing the uncertainty that comes
from our laziness and ignorance, thereby solving the qualification
problem.
 An outcome is a completely specified state, including such factors as
whether the agent arrives on time and the length of the wait at the
airport.
 Preferences, as expressed by utilities, are combined with probabilities in
the general theory of rational decisions called decision theory:
 Decision theory = probability theory + utility theory .
 BASIC PROBABILITY NOTATION
 In probability theory, the set of all possible worlds is called the sample
space. Example: two dices are rolled – 36 possible worlds.
 The Greek letter Ω (uppercase omega) is used to refer to the sample
space, and ω (lowercase omega) refers to elements of the space, that is,
particular possible worlds.
 A fully specified probability model associates a numerical probability
P(ω) with each possible world.
 0 ≤ P(ω) ≤ 1 for every ω and  ω∈Ω P(ω) = 1 .
 unconditional or prior probabilities (and sometimes just “priors” for
short) refer to degrees of belief in propositions in the absence of any
other information. For example, rolling the two dices and they add up to
11, P(Total = 11) = P(<5, 6>, <6,5>)
 conditional or posterior probability for example, rolling doubles given
that the first die is a 5. This probability is written P(doubles | Die1 =5)
 Ex. P(cavity | toothache), P(cavity)
 Note that the precedence of “ | ” is such that any expression of the form
P(. . . | . . .) always means P((. . .)|(. . .)).
 The assertion that P(cavity | toothache)=0.6 does not mean “Whenever toothache is true, conclude that cavity is true with
probability 0.6” rather it means “Whenever toothache is true and we have no further information, conclude that cavity is true
with probability 0.6., P(cavity|toothache ∧ ¬cavity)=0.
 Mathematically speaking, conditional probabilities are defined in terms
of unconditional probabilities as follows: for any propositions a and b,
we have

 The definition of conditional probability can be written in a different

form called the product rule: P(a ∧ b) = P(a | b)P(b)
 Random Variables: Variables in Probability Theory and their names
begin with an uppercase letter. Thus, in the dice example, Total and
Die1 are random variables.
 Every random variable has a domain—the set of possible values it can
take on. The domain of Total for two dice is the set {2, . . . , 12} and the
domain of Die1 is {1, . . . , 6}.
 “The probability that the patient has a cavity, given that she is a
teenager with no toothache, is 0.1” as follows: P(cavity | ¬toothache ∧
teen) = 0.1 .
 We say that the P statement defines a probability distribution for the
random variable Weather.
 The P notation is also used for conditional distributions: P(X | Y ) gives
the values of P(X = xi | Y =yj) for each possible i, j pair.
 For continuous variables, it is not possible to write out the entire
distribution as a vector, because there are infinitely many values.
 Instead, we can define the probability that a random variable takes on
some value x as a parameterized function of x.
 For example, the sentence P(NoonTemp =x) = Uniform[18C,26C](x)
expresses the belief that the temperature at noon is distributed
uniformly between 18 and 26 degrees Celsius.
 We call this a probability density function (pdf).
 Probability axioms and their reasonableness

 We can also derive the well-known formula for the probability of a

disjunction, sometimes called the inclusion–exclusion principle:
 P(a ∨ b) = P(a) + P(b) − P(a ∧ b) .
 INFERENCE USING FULL JOINT DISTRIBUTIONS
 Example: A domain consisting of just the three Boolean variables
Toothache, Cavity, and Catch (the dentist’s nasty steel probe catches in
my tooth).
 The full joint distribution is a 2×2×2 table as shown in

 How to determine the probability of any proposition?

 Notice also that Equation on RHS gives us a direct way to calculate the
probability of any proposition, simple or complex: simply identify those
possible worlds in which the proposition is true and add up their
probabilities.
 For example, there are six possible worlds in which cavity ∨ toothache
holds:
 P(cavity ∨ toothache) = 0.108 + 0.012 + 0.072 + 0.008 + 0.016 + 0.064 =
0.28
 adding the entries in the first row gives the unconditional or marginal
probability of cavity: P(cavity) = 0.108 + 0.012 + 0.072 + 0.008 = 0.2
 This process is called marginalization, or summing out.
 We can write the following general marginalization rule for any sets of
variables Y and Z:
 To Computer Conditional Probability
 INDEPENDENCE
 Independence between propositions a and b can be written as
 P(a | b) = P(a) or P(b | a) = P(b) or P(a ∧ b) = P(a)P(b) .
 For Example, add Weather variable in our previous example:
 The full joint distribution then becomes P(Toothache, Catch, Cavity,
Weather ), which has 2 × 2 × 2 × 4 = 32 entries.
 What relationship do these editions have to each other and to the
original three-variable table?
 For example, how are P(toothache, catch, cavity, cloudy) and
P(toothache, catch, cavity) related? We can use the product rule:
P(toothache, catch, cavity, cloudy)
= P(cloudy | toothache, catch, cavity) P(toothache, catch, cavity) .
 Now, unless one is in the deity business, one should not imagine that
one’s dental problems influence the weather.
 And for indoor dentistry, at least, it seems safe to say that the weather
does not influence the dental variables.
 BAYES’ RULE AND ITS USE
 we defined the product rule. It can actually be written in two forms:
Applying Bayes’ rule:
Ex: Consider two medical tests, A and B, for a virus. Test A is 95% effective at
recognizing the virus when it is present, but has a 10% false positive rate
(indicating that the virus is present, when it is not). Test B is 90% effective at
recognizing the virus, but has a 5% false positive rate. The two tests use
independent methods of identifying the virus. The virus is carried by 1% of all
people. Say that a person is tested for the virus using only one of the tests, and
that test comes back positive for carrying the virus. Which test returning
positive is more indicative of someone really carrying the virus? Justify your
answer mathematically.
• The Wumpus World Revisited

• Representing and Using Domain Knowledge

• Expert System Shells

• Explanation

• Knowledge Acquisition