1. Application of EUT
One simple application of EUT is to list all features of your options,
bring their subjective value onto the same scale (e.g., a utility scale or
money), and assign probabilities to the likelihoods that the options will have
these features in the future. You can then multiply the expected value of each
feature and add up each option's expected feature values. (I gave an example
in class.) The accuracy of this procedure is less important than its power to
"test your intuition." We often have preferences but cannot identify or
express them clearly. An algorithm such as EUT can challenge those preferences
by recommending a particular choice. If you are happy with this
recommendation, go for it; if you are unhappy with it, you may want to choose
the other option.
EUT is not a theory that describes how people make decisions. It is a
mathematical apparatus with certain properties. One of its properties is that
decisions made according to its axioms (see textbook) will maximize expected
value. If, for a given decision, you have all the information you need and
maximizing expected value seems desirable, then you may want to apply EUT.
Many decision problems do not provide you with all the information, so EUT
would be hard to apply. And many domains (e.g., relationships) seem too
emotional to apply a mathematical model. But keep in mind that you will make a
decision based on reasons anyway--your beliefs and desires are the basis on
which you make most decisions; by using EUT you can assist yourself in becoming
aware of those beliefs and desires as well as their interrelations.
2. Calculating your own utility function
You start with a simple gamble, such as a 50:50 chance of either winning
$100 or winning nothing, and determine your "cash equivalent"--i.e., the amount
of money for sure that you would find as attractive as an opportunity to play
the gamble. Say your cash equivalent (ce) is $40. Because you made the
utility of the gamble equal to the utility of the ce, you can set up the
equation that u($40) = .50[u($100)] + .50[u($0)]. Then we scale your utility
such that u($100) = 1 and u($0) = 0. We can solve the equation for u($40) and
find that it is 0.5. You continue this process with different gambles, and
soon you will be able to plot your whole positive utility graph. You repeat
the same with gambles about losses and thus plot the negative utility graph.
The result will be some version of the S-shaped curve proposed by Kahneman
& Tversky's (1979) prospect theory, nicely depicted in your textbook on p.
96.
3. Features of prospect theory
Gains vs. losses: The first important feature of prospect theory is
that the loss curve is shaped differently than the gain curve. The concave
gain curve implies a risk-averse attitude when you face possible gains (you
perfer $8 for sure over a 50:50 gamble to win either $20 or nothing; in
general, you prefer lower sure gains over a risky chance of possibly winning
higher gains). In contrast, the convex loss curve implies a risk-seeking
attitude when you face possible losses (you prefer a 50:50 gamble to lose
either $20 or nothing over a sure loss of $8; in general, you prefer a risky
chance of possibly avoiding a loss altogether over losing a small amount for
sure). Both of these risk attitudes can be quite harmful, especially for the
gambler who has lost already $800 and risks a lot to avoid losing the money for
sure. For any single case, it may be justifiable to be risk averse in gains
and risk-seeking in losses (that justification would take into account factors
such as your anticpiated regret). But EUT reminds you that your decision
strategies have "expected" outcomes over many decisions. Consider again
the game of choosing between a $20/$0 gamble and $8 for sure. You may be fine
to be risk averse (i.e., prefer the $8) when playing the game only once; but
suppose you play the game 10 times--your risk aversion would earn you $80,
whereas the risk-neutral stance of choosing the gamble would earn you an
expected $100.
The asymmetry of gains and losses also implies that bundled losses are easier
to bear than spread-out losses, whereas spread-out gains are more enjoyable
than bundled gains (e.g., you may prefer to pay all your bills at once but to
get your birthday presents one a day).
The gain-loss curve has parallels in social perception (negative traits and
behaviors reiceve disproportional weight in impression formation), self
perception (negative views of the self are particularly strongly held and hard
to get rid of), child rearing (one punishment is much more impactful than
several rewards), and several other domains.
Loss aversion. In general, people will avoid sure losses as much as
they can. If, however, people reframe a sure loss as an "insurance premium"
(i.e., a purchase of security) they are much more willing to accept that loss
(Slovic, Fischhoff, & Lichtenstein, 1982).
People's loss aversion leads to curious effects: once they own something (have
made it part of themselves), they will ask an unduly high selling prices for it
("endowment effect") and they are unwilling to give it up for a new option
("status-quo bias"). We have seen similar effects when discussing the
difficulty of breaking up ("sunk-cost bias") and dissonance reduction
("committment").
Decision weights: Another impotant feature of prospect theory is that it
shows how people interpret (or transform) probabilities. In contrast to EUT,
prospect theory shows that the difference between p = .40 and p = .50 is
trivial for most people, whereas the difference between p = .90 and p = 1.0 (or
between p = .10 and p = 0.0) is powerful ("certainty effect"). As a result,
sure gains are much more attractive, and sure losses much more unattractive,
than even highly probable gambles. This is why insurance is attractive only if
it appears to cut the risk of a big loss to p = 0.0.
People also treat very low and very high probabilities differently. Very low probabilities are overweighted (people play lottery despite extremely small chances of winning; and they feel threatened by tiny likelihoods of diseases or disasters). Very high probabilities are underweighted (neurosurgery with 80% chance of success is a medical achievement but may still frighten many people).
In sum, prospect theory gives a more realistic image of people's actual decision making than does EUT. Just as EUT, however, it relies heavily on normed scales, especially money and probabilities. All experimental vignettes to test prospect theory make use of these normed scales, while most everyday decisions neither have an exact value nor an exact expectation. For dealings with money and known probabilities, however, prospect theory shows us some of our dangerous habits (e.g., gambling on a losing streak...) and reminds us to consider EUT as a possible norm after which to model our decisions.