HSO 510: Judgement and

Decision Making
UNIT 4: Rational decision theories: Expected utility theory.

Prof. Sathya Narayana Sharma

IIT (ISM) Dhanbad
Rational Choice under Risk and Uncertainty: Expected
Utility Theory
You face a choice under uncertainty when the probabilities of the relevant
outcomes are completely unknown or not even meaningful.

You face a choice under risk when the probabilities of the relevant outcomes are
both meaningful and known.
Choice under uncertainty You don’t know the probabilities of what will happen (like you don’t know if it will rain or not, no idea about chances).

Choice under risk You do know the probabilities (like you know there’s a 70% chance of rain and 30% chance of no rain).
 States
of the
world
Menu

Outcome

} space
Maximin criterion (very cautious person):
Look at the worst possible outcome for each option.
Then pick the option where this worst outcome is least bad.
Basically: “What’s the safest choice if things go badly?”

Maximin criterion: you should choose the alternative that has the greatest
minimum utility payoff.

Maximax criterion: you should choose the alternative that has the greatest
maximum utility payoff.

The maximin reasoner is as cautious as the maximax reasoner is reckless.

The former looks at nothing but the worst possible outcome associated with each
act, and the latter looks at nothing but the best possible outcome associated with
each act. Maximax criterion (very optimistic/reckless person):
Look at the best possible outcome for each option.
Then pick the option with the best best-case.
Basically: “What’s the most awesome thing that could happen?”
Exercise
Having just bought a brand-new watch, you are asked if you also want the optional
life-time warranty.

(a) Would a maximin reasoner purchase the warranty?

(b) What about a maximax reasoner? In short:

Option 1: Buy warranty (pay extra now, but safe if the watch breaks later). Maximin buys warranty (safety-first).
Option 2: Don’t buy warranty (no extra cost now, but risk of big repair cost later). Maximax skips warranty (optimistic).
(a) Maximin reasoner (cautious)
Looks only at the worst-case outcome.
Worst case with warranty You paid extra, but the watch breaks You’re still safe, no big extra cost.
Worst case without warranty The watch breaks, and you must pay a huge repair/replacement cost.
The worst case is much worse without warranty.
So, a maximin reasoner would purchase the warranty.
(b) Maximax reasoner (optimistic)
Looks only at the best-case outcome.
Best case with warranty The watch never breaks, but you wasted money on the warranty.
Best case without warranty The watch never breaks, and you didn’t waste money.
The best case is better without warranty.
So, a maximax reasoner would not purchase the warranty.
Minimax-risk criterion: You should choose the alternative that is associated with
the lowest maximum risk or regret.

for part c in next ques

this is how to make risk matrix

Step 2: For each state of nature (each column), find the best payoff
For S: max(1, 2, 3) = 3
For S: max(10, 9, 6) = 10
Step 3: Compute regret values
Regret = (Best payoff in that column) – (Actual payoff of the option)
For S:
A: 3 – 1 = 2
B: 3 – 2 = 1
C: 3 – 3 = 0
For S:
A: 10 – 10 = 0
B: 10 – 9 = 1
C: 10 – 6 = 4
 (a) Maximin criterion (choose best worst-case payoff) (b) Maximax criterion (choose best best-case payoff)
A min(1, 10) = 1 A max(1, 10) = 10

Exercise B min(2, 9) = 2
C min(3, 6) = 3
B max(2, 9) = 9
C max(3, 6) = 6
Best of these minimums = 3 (from C). Best of these maximums = 10 (from A).
Answer: Choose C. Answer: Choose A.

This exercise refers to the utility matrix below. What course of action would be
favored by (a) the maximin criterion, (b) the maximax criterion, and (c) the
minimax-risk criterion?

Minimax-risk criterion (choose lowest maximum regret)

From regret matrix:
A max(2, 0) = 2
B max(1, 1) = 1
C max(0, 4) = 4 Smallest of these = 1 (from B).
Answer: Choose B.
 Objections to Maximin Criterion
Too pessimistic: It only looks at the worst outcome and ignores all other possible outcomes.
Ignores useful information: Even if an option has good or excellent payoffs in other cases, maximin doesn’t care—it
Objections only cares about the worst case.
Ignores probabilities: If you actually know the chances of each state (like 80% sunny, 20% rainy), maximin doesn’t use
that information. It treats all states as if they were equally scary.

Of all criteria for choice under uncertainty, the maximin criterion is the most
prominent.
Maximin reasoning fails to consider relevant utility information, since for each act,
it ignores all payoffs except the worst.

Another objection is that maximin reasoning fails to take into account the chances
that the various states of the world will obtain.
Suppose you live in New Delhi and are offered two jobs at the same time. One is a
tedious and badly paid job in New Delhi itself, while the other is a very interesting
and well-paid job in Mumbai. But the catch is that if you wanted the Mumbai job,
you would have to take a plane from New Delhi to Mumbai (for example, because
this job would have to be taken up the very next day). Therefore, there would be a
very small but positive probability that you might be killed in a plane accident.
Expected Value
The expected value of a gamble is what you can
expect to win on the average, in the long run, when
you play the gamble.

On the average, in the long run, you would get $5

when playing this gamble; the expected value of the
gamble = $5.

That is the same as the figure you get if you

multiply the probability of winning (1/2) by the dollar
amount you stand to win ($10).
Choice between gambles
Multi-stage gamble
The general decision problem
Exercise
You are considering whether to park legally or illegally and decide to be rational
about it. Use negative numbers to represent costs in your expected-value
calculations.

(a) Suppose that a parking ticket costs ₹30 and that the probability of getting a
ticket if you park illegally is 1/5. What is the expected value of parking illegally?

(b) Assuming that you use expected-value calculations as a guide in life, would it
be worth paying ₹5 in order to park legally?
Exercise
You are offered the following gamble: if a (fair) coin comes up heads, you receive
₹10; if the coin comes up tails, you pay ₹10. What is the expected value of this
gamble?
Exercise
You are on a game show and you are facing three boxes. One of them contains ₹900,000, one
contains ₹300,000, and one contains ₹60, but you do not know which is which. Here are the
rules: if you choose to open the boxes, you can open them in any order you like, but you can
keep the amount contained in the last box only.
(a) What is the expected value of opening the three boxes?
(b) The host gives you the choice between a sure ₹400,000 and the right to open the three
boxes. Assuming you want to maximize expected value, which should you choose?
(c) You decline the ₹400,000 and open a box. Unfortunately, it contains the ₹900,000. What is
the expected value of opening the remaining two boxes?
(d) The host gives you the choice between a sure ₹155,000 and the right to open the remaining
two boxes. Assuming you want to maximize expected value, which should you choose?
Cost-benefit analysis
Expected-value calculations form the core of cost–benefit analysis, which is used
to determine whether all sorts of projects are worth undertaking.

The basic idea is to compare expected benefits with expected costs: if the benefits
exceed or equal the cost, the assumption is that it is worth proceeding, otherwise
not.
Computing probabilities from expected values
If a parking ticket costs ₹30, and it costs ₹5 to park legally, what does the
probability of getting a ticket need to be for the expected value of parking legally to
equal the expected value of parking illegally?
Exercise
A tablet computer costs $325; the optional one year warranty, which will replace
the tablet computer at no cost if it breaks, costs $79. What does the probability p
of the tablet computer breaking need to be for the expected value of purchasing
the optional warranty to equal the expected value of not purchasing it?
St. Petersburg Paradox
A gamble is resolved by tossing an unbiased coin as many times as necessary to
obtain heads. If it takes only one toss, the payoff is $2; if it takes two tosses, it is
$4; if it takes three, it is $8; and so forth. What is the expected value of the
gamble?

1/2 ∗ $2 + 1/4 ∗ $4 + 1/8 ∗ $8 +… = $1 + $1 + $1 +… = $∞

Expected Utility
St Petersburg gamble with utilities

1/2 ∗ log (2) + 1/4 ∗ log (4) + 1/8 ∗ log (8) +…= 2 log (2) ≈ 0.602 < ∞
Expected Utility
Exercise
Suppose that your utility function is u (x) = √x. Should you accept or reject the
gamble?
The utility of rejecting the gamble is EU (R) = u (4) = √4 = 2.
The utility of accepting the gamble is
EU (A) = 1/2*u (10) + 1/2 ∗ u (0) = 1/2 ∗ √10 ≈ 1.58.
What if your utility function is u(x) = x2?
Exercise
A patient with hearing loss is considering whether to have surgery. If she does not have
the surgery, her hearing will get no better and no worse. If she does have the surgery,
there is an 85 percent chance that her hearing will improve, and a five percent chance
that it will deteriorate. If she does not have the surgery, her utility will be zero. If she
does have the surgery and her hearing improves, her utility will be ten. If she does
have the surgery but her hearing is no better and no worse, her utility will be
minus two. If she does have the surgery and her hearing deteriorates, her utility
will be minus ten.
(a) What is the expected utility of not having the operation?
(b) What is the expected utility of having the operation?
(c) What should the patient do?
Exercise
Suppose you are contemplating whether to go home for Midsem break. You would
like to see your family, but you are worried that your aunt may be there, and you
genuinely hate your aunt. If you stay in campus you are hoping to stay with your
roommate, but then again, there is some chance that she will leave town. The
probability that your aunt shows up is 1/4, and the probability that your
roommate leaves town is 1/3. The utility of being with your family without the
aunt is 12 and with the aunt is -2. The utility of staying in your hostel without
your roommate is 3 and with the roommate is 9. What should you do?
Indifference
If the probability of rain is p, what does p
need to be for the expected utility of taking
the umbrella to equal the expected utility of
leaving it at home?

EU (Take umbrella) = EU (Leave umbrella).

3 = p * 0 + (1 – p) * 5
Attitudes towards risk
Suppose you own $2 and are offered a gamble giving you a 50 percent chance of
winning a dollar and a 50 percent chance of losing a dollar. Your utility function is u
(x) = √x, so that marginal utility is diminishing. Should you take the gamble?

What if u (x) = x2 ?
accept
E(U)=1/2*sqrt(3)+1/2*sqrt(1) Since
reject sqrt(2)1.414
Keep 2 dollars and (sqrt(3)+1)/2=1.366
E(U)=sqrt(2) rejecting gives higher utility.
Whether you should reject or accept a gamble depends on whether your utility
function gets flatter or steeper.

You are risk averse if you would reject a gamble in favor of a sure dollar amount
equal to its expected value, risk prone if you would accept, and risk neutral if
you are indifferent.

You are risk averse if your utility function bends downwards (as you move from left
to right), risk prone if your utility function bends upwards, and risk neutral if your
utility function is a straight line.
A concave utility function (bends
downward) means marginal utility
decreases with wealth. This describes a
risk-averse individual, who prefers a
certain outcome over a risky one with the
same expected value.

A convex utility function (bends upward)

indicates marginal utility increases with
wealth. This describes a risk-prone or
risk-seeking individual, who prefers the
gamble to the certain outcome.

A linear utility function (straight line)

implies marginal utility is constant. This is
a risk-neutral individual, indifferent
between gambling and certainty if the
expected value is the same.

Classification Based on Behavior

Risk averse: Rejects a fair gamble; utility

function is concave.

Risk prone (risk seeking): Accepts a fair

gamble; utility function is convex.

Risk neutral: Indifferent; utility function is

linear.
(a) People who invest in the stock market rather than in savings accounts
Risk attitude: Generally risk prone (risk seeking)
Reason: The stock market offers higher expected returns, but with much more risk (volatility and uncertainty). People choosing stocks over
guaranteed savings typically have higher risk tolerance or are seeking higher gains, accepting the possible losses.
(b) People who buy lottery tickets rather than holding on to the cash
Risk attitude: Risk prone (risk seeking)
Reason: Lotteries have negative expected value (statistically, people lose money), but the potential for a huge win entices some to take the risk.
Opting for the gamble over certainty is a classic risk-seeking behavior
As far as you can tell, are the following people risk prone, risk averse, or risk
neutral?

(a) People who invest in the stock market rather than in savings accounts.

(b) People who buy lottery tickets rather than holding on to the cash.

(c) People who buy home insurance.

(d) People who consistently maximize expected value.

c) People who buy home insurance
Risk attitude: Risk averse
Reason: Insurance buyers pay a certain premium to avoid possible large losses, preferring certainty to risk. Taking steps to insure against
low-probability, high-impact events is a hallmark of risk aversion.
(d) People who consistently maximize expected value
Risk attitude: Risk neutral
Reason: Risk-neutral individuals focus only on maximizing average (expected) returns, indifferent to the riskiness or variability of outcomes.
They neither avoid nor seek risk for its own sake.
Certainty equivalent
The certainty equivalent represents what the gamble is worth to you. The certainty
equivalent determines your willingness-to-pay (WTP) and your
Certainty Equivalent (CE):
willingness-to-accept (WTA). The certainty equivalent is the guaranteed amount of
money (or utility) that makes you just as satisfied as
playing a gamble.
Risk averse: certainty equivalent < expected value
Formally:
u(CE)=EU(G)
Risk prone: certainty equivalent > expected value
If you are risk-averse CE < Expected Value (EV)
If you are risk-prone CE > EV
If you are risk-neutral CE = EV

The certainty equivalent of a gamble G is the number CE that satisfies this

equation: u(CE) = EU(G). So, the CE tells you how much the gamble is truly “worth”
to you and shows your willingness-to-pay (WTP) or
willingness-to-accept (WTA).
1/2 ∗ log (2) + 1/4 ∗ log (4) + 1/8 ∗ log (8) +…= 2 log (2) ≈ 0.602 < ∞
What is the certainty equivalent of the gamble?
log(CE) = 0.602 Here, u(x)=log(x), so:
log(CE)=0.602
CE = 100.602 ≈ 4.00
Exercise
Suppose that you are offered the choice between ₹4 and the following gamble,
G: 1/4 probability of winning ₹9 and a 3/4 probability of winning ₹1.

(a) Suppose that your utility function is u (x) = √x. What is the utility of ₹4? What is
the utility of G? What is the certainty equivalent? Which would you choose?

(b) Suppose instead that your utility function is u(x) = x2. What is the utility of ₹4?
What is the utility of G? What is the certainty equivalent? Which would you
choose?
a) 1/4*sqrt(9)+3/4*(sqrt(1))=1.5. u(CE)=EU(G) b)1/4*9*9+3/4*1*1=21
sqrt(CE)=1.5 CE^2=21
CE=(1.5) ^2 CE=SQRT(21)=4.58
=2.25 Since
Since CE=4.58>4, the Gamble is worth more than the sure 4.
CE=2.25<4, the Gamble is worth less than the sure 4. Choose the Gamble.
Choose 4.
Suppose that your utility function is u (x) = √x, and that you are offered a gamble
which allows you to win ₹16 if you are lucky and ₹4 if you are not.

(a) Suppose that the probability of winning ₹16 is 1/4 and the probability of
winning ₹4 is 3/4. What is the expected utility of this gamble?

(b) What is the certainty equivalent of the gamble?

(c) Imagine now that the probability of winning ₹16 is p and the probability of
winning ₹4 is (1 – p). If the expected utility of the gamble equals 9/4, what is p?

(d) Are you risk averse or risk prone, given the utility function above?
 Decision context: Sometimes it's unclear which decision is the "right" one ahead of time due to unknown future outcomes.
Rational decision: The theory assumes that rational decision-makers choose the option that maximizes expected future utility.
Probabilities: People assign probabilities to states of the world (outcomes), and these probabilities must adhere to the laws of probability.
Utilities: Individuals assign utilities (values satisfying their preferences) to each possible outcome.
Choice criterion: The decision selected is the one with the greatest expected utility, calculated by weighting utilities by their respective
probabilities.

Whether we treat a decision as a choice under uncertainty or under risk can have
real consequences.
We do not know ahead of time which decision is the right one. That is why we aim
for the rational decision – being the one with the greatest expectation of future
utility.
The theory assumes that people assign probabilities to states of the world, that
these probabilities satisfy the axioms of the probability calculus.
People assign utilities to outcomes, and that they choose that alternative which
has the greatest expected utility given the probabilities and utilities.