1. Decision making
Decision making is an important area within cognitive psychology, because of the applied interest: everyone would like to make better decisions, including armies, corporations, and individuals.
So, how is the science doing? Do we have a good understanding of human decision making? Not really…
3:15 PM

2. 2 approaches to study of decision making
Behavioural decision making
Basic idea: people make deliberate choices
requires assessment of self-interest, computed from attributes of alternatives
Loewenstein’s embryonic view
People do not compute value, but construe situations and select appropriate rules
3:15 PM

3. Behavioural decision making
Decision making is defined as choice among alternatives.
Traditional view: people analyze alternatives for their attributes. They then evaluate these attributes comparatively, and choose the one that maximizes value.
Dominant view has been Expected Value theory.
3:15 PM

4. Expected value models Originally advanced by economists.
Basic idea: value of an outcome is product of $ value (if it occurs) and probability.
E.g., a one-tenth chance of winning $200 is worth (1/10 * $200) = $20
Problem: humans do not make their choices on the basis of expected value.
3:15 PM

5. Problems with Expected Value models
Three reasons humans do not use expected value approach to decision making:
Specification of relevant attributes for any particular alternative is difficult.
B. Evaluation of alternatives is difficult.
C. For many problems, processing all attributes of all possible alternatives would require more capacity than humans have. 
3:15 PM

6. Problems with Expected Value models
Specification of attributes is hard.
Evaluation of alternatives is hard
Processing capacity is limited
3:15 PM

7. Specification of alternatives
Klein (1989)
150 interviews with 5 populations of decision makers (e.g., managers)
Conclusion: few decisions involve generating a variety of options and comparing their strengths & weaknesses.
Most alternative we choose between can’t be broken down into clear attributes
3:15 PM

8. Problems with Expected Value models
Specification of attributes is hard.
Evaluation of alternatives is hard
Processing capacity is limited
3:15 PM

9. Evaluation of alternatives (1)
Which would you prefer - a 1/10th chance of winning $100, or a 9/10th chance of winning $8?
1/10th X $100 = $10
B. 9/10th X $8 = $7.20
But suppose you are hungry - really hungry. Then, higher probability more attractive, so you choose B. How do we quantify the extra value associated with certainty?
Perhaps we change B to 9/10 X $7 = $6.30. If people still prefer B, make it 9/10 X $6 = $5.40. Just keep reducing the amount. At some point, people would always choose A.
Would that tell us the value of certainty?
Unfortunately, no, because small changes in procedure would lead us to different answers to the question, what is certainty worth? For example, if we get people to choose between A and B, as above, we would get one answer. If we ask people, how much would you pay for a ticket with the conditions of B, we would get another answer. If we ask them, how much would you take to sell me a ticket with the conditions of B, we would get a third answer.
3:15 PM

10. Evaluation of alternatives (2)
There are important differences between objective and subjective value of money: 
Subjective value is not a linear function of objective value.
$1 million = 1000 X $1000
$1 billion = 1000 X $1 million.
Is $1 billion worth 1000 times as much as $1 million to you?
3:15 PM

11. Evaluation of alternatives (2)
2. Subjective value is not symmetric for gains and losses (losses are more important). Kahneman & Tversky (1984):
A. You flip a coin. Heads – you win $20. Tails – you pay $20.
B. You do not flip a coin. You win nothing.
Most people chose B.
3:15 PM

12. Problems with Expected Value models
Specification of attributes is hard.
Evaluation of alternatives is hard
Processing capacity is limited
3:15 PM

13. Processing capacity is limited
Humans often do not consider all possible outcomes because doing so would take too long. Example:
An HR person must hire a new employee.
She has 500 applications. Each takes 10 minutes to review -> 80 hours 20 minutes
She has 6 hours for the task. What should she do?
3:15 PM

14. Processing capacity is limited
In 6 hours, she can review 36 applications.
She reads 36 at random, hires the best of those 36.
This is called satisficing (doing well enough).
Instead of analyzing problems the way a computer would, humans use heuristics such as satisficing.
3:15 PM

15. Heuristics
A heuristic is a 'rule of thumb,' a procedure which is easy to use, though it may not work.
In contrast, an algorithm is a step-by-step procedure guaranteed to produce the correct result.
We do not always have an algorithm, and when we do have one, it may take too long (as in the case of the HR person). Then, we use heuristics
3:15 PM

16. Processing capacity is limited
Because processing capacity is limited, humans often use heuristics. We’ll talk about three important heuristics that have been studied by psychologists:
Availability
Representativeness
Simulation
3:15 PM

17. Three important heuristics
Availability
judgment that the more easily an event comes to mind, the more likely it is to occur.
Often works - e.g., when was the last time you met a professor who liked to be yelled at?
Sometimes doesn’t work – as in cases of illusory correlation.
3:15 PM

18. Illusory Correlation
IC is found in cases of rare events that happen to co-occur.
E.g., sports announcer says "X hasn't dropped the ball for 100 plays," and X drops the ball on the next play.
You only notice this when the rare, newsworthy event happens - not when X doesn't then drop the ball.
IC also may have a role in racism.
3:15 PM

19. Three important heuristics
2. Representativeness
An event considered typical of a large class of events will be considered more probable than an atypical event.
So, if you are judging the likelihood of something, you are influenced by how typical it is.
Judging likelihood part of choice.
3:15 PM

20. Representativeness Tversky & Kahneman (1983):
Linda is 31 years old, single, outspoken and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in antinuclear demonstrations
3:15 PM

21. Kahneman & Tversky (1983) Which of these is more likely?
A. Linda is a bank teller.
B. Linda is a bank teller and is active in the feminist movement.
Most subjects picked B - but that is mathematically impossible. The set of people (bank tellers who are feminists) is smaller than the set of people (bank tellers).
3:15 PM

22. Three important heuristics
3. Simulation
Rushing to the airport, you miss your plane by 5 minutes – or by 2 hours. Which is more annoying?
It’s a lot easier to simulate (imagine) catching the plane if you only missed it by 5 minutes – so that is more annoying.
3:15 PM

23. Simulation heuristic
The simulation heuristic lets us "see" the consequences of actions before we do them.
What if Frodo hadn’t volunteered to take the ring to Mordor? What if you took a year off from university and went travelling
Simulation shows both alternative futures and alternative pasts that you could learn a lesson from.
3:15 PM

24. Expected Value models - Summary
We considered a number of problems with EV:
Specification of attributes is hard.
Evaluation of alternatives is hard
Evaluation is tricky
Subjective & objective value of money differ
Processing capacity is limited
So we often use heuristics instead of computing which alternative is best.
3:15 PM

25. Expected utility models
Because Expected Value models don’t capture human decision making, psychologists developed concept of Expected Utility:
essentially, “how much is it worth to you?”
Problem: there’s no way to specify this in advance. It’s subjective.
Expected value models do not predict or explain human behaviour. Lowenstein (2001): “Research on judgment has documented numerous systematic biases in human information processing, such as underweighting base rates, overconfidence, hindsight bias, and misperceptions of random sequences. Research on choice has explored ways in which decision makers violate such stylized assumptions as linear probability weighting, exponential time discounting, and various forms of preferential independence. All of these topics have been studied with a variety of methods, including laboratory experiments, field studies, and economics-style market experiments.”
3:15 PM

26. How can we have a science of value?
Expected value theory is meaningful but doesn't work. Expected utility theory works but is essentially empty: Subjective value is impossible to predict.
The first problem with a science of human decision making is that we can’t objectively measure value, so we can’t predict behaviour. (Perhaps this is why there are markets.)
Add some consideration of markets as assessors of value.
3:15 PM

27. How can we have a science of value?
What should we do?
Some psychologists and economists say, keep trying. We’ll eventually develop a science of decision making.
Others say, don’t bother. We don’t need a science of decision making, because decision making doesn’t exist…
3:15 PM

28. Some more recent alternative views
Some psychologists have argued that we don’t, in fact, make many decisions. Rather, they say, human behaviour is automatic (e.g., Bargh & Chartrand, 1999), or rule-guided (e.g., Anderson’s production rule system in his ACT model).
Loewenstein (2001) summarized arguments against the idea that humans make decisions.
3:15 PM

29. Loewenstein’s (2001) arguments
Context effects
Intra-individual variation
Effects of careful consideration
Decision making anomalies
Do we know what we want?
3:15 PM

30. Loewenstein’s arguments
Context matters.
E.g., choice between 2 gambles affected by whether decision was said to involve a gamble or insurance (Hershey et al. 1982).
Context effects influence construal of the situation, thus which set of rules of behaviour is applied (e.g., tough negotiator with a rival, soft negotiator with a child – but what if child becomes a rival?)
Note: insurance is a gamble, though we don’t usually think of it that way.
Ross & Ward (1996) showed that responses in the prisoner’s dilemma profoundly affected by whether the game was called the Wall Street Game or the Community Game. Loewenstein: these effects seem to involve change in the subject’s construal of the kind of situation they are in, and thus of the rules of behaviour that should apply. Note that this makes individual valuation or interpretation central to explaining decision making behaviour. And this is amplified by the intra-individual variability reported by e.g., MacCrimmon & Wehrug, (1990).
3:15 PM

31. Loewenstein’s arguments
If decision-making is algorithmic, why is there so much intra-individual variation?
Fear is not based on objective risk: Many people are afraid of some risks (e.g., flying) but not of others (eating red meat).
Many people have long-term goals but also indulge in short-sighted behaviour.
So whether we are afraid, whether we are goal-directed, is subjectively determined.
Loewenstein (2001) p. 501:
“In the domain of risk, for example, many people are scared by risks that are objectively rather harmless, such as flying or social interactions, but not by risks that are truly hazardous, such as driving or eating fat red meat. People will appear to be inconsistent if you fail to take into account the way that they construe risks – that is, what they find scary – but appear much more consistent if you take into account their perceptions of riskiness.”
In other words, we need a theory not only for every person but for each construal of each situation a person might be in.
3:15 PM

32. Loewenstein’s arguments
If decision making theory is correct, making people consider their choices more carefully should lead to better choices. It doesn’t. In fact, it can make things worse.
Wilson et al. (1993): students selected favourite poster from a set. Those asked to provide reasons for choice less happy with choice, less likely to display poster. Wilson: deliberation interferes with “gut” reaction.
Klein (1989, p. 79)
“With a few exceptions, decision training has not been shown to be very effective, and under time pressure such training has not shown any benefit.”
3:15 PM

33. Loewenstein’s arguments
Decision making anomalies
People prefer sequences of outcomes that improve over time. But expected utility tells us that delayed rewards are discounted.
Diversification bias: if people choose several things from a set, they go for more diversity when they choose simultaneously than when they choose sequentially. Why?
Diversification bias: Read & Loewenstein (1995) – J.E.P: Applied, 1 (March)
3:15 PM

34. Loewenstein’s arguments
Do we know what we want?
Ariely et al. (2001): some subjects asked, “Would you listen to 10 minutes of poetry if we gave you $10?” Others asked, “Would you pay $10 to hear 10 minutes of poetry?”
Later, first group had to be paid to listen to poetry; second group willing to pay.
Both groups named higher amounts for longer durations! They didn’t know whether listening to poetry was positive or negative, but knew it was more positive or more negative if longer.
Ariely et al. (2001) asked some subjects if they would be willing to listen to 10 minutes of Walt Whitman’s poetry (Leaves of Grass) if they were paid $10. Other subjects were asked if they would be willing to pay $10 to listen to the same poems. Then, subjects bid for listening to 1, 5, or 10 minutes of poetry. “Those asked if they would pay were willing to pay to listen to the poetry, and those asked if they would listen if they were paid needed to be paid to do it. Whether they valued the experience positively or negatively, however, they named higher money amounts for longer durations of poetry. Participants had no idea whether this was a positive or negative experience, but they knew that it was more positive or negative if it was longer.” (Loewenstein, 2001, p. 502)
3:15 PM

35. Loewenstein’s alternative perspective
Consciousness mainly makes sense of behaviour afterwards.
People rely on highly-developed cognitive capacities such as perception and categorization, more than computations.
Two stage process: first, construe the situation. Then apply relevant rules.
#3 – paying people a small amount to donate blood seems to decrease donations (Titmuss, 1987); charging parents a small amount for picking up their children late at daycare increases late pickups (Gneezy & Rustichini, I Camerer et al., 2002).
3:15 PM

36. Loewenstein’s alternative perspective
Intra-individual differences to be explained via situation construal.
5. Identify choice heuristics people use and cues that trigger them.
6. Don’t assume people evaluate outcomes consciously, or ‘know’ what they want.
Loewenstein (2001) - Happiness is “only weakly related” to objective circumstances. Some people see through ‘rose coloured glasses’ others through ‘grey glasses.’
This, of course, is where the market shines – because individuals are not good at evaluating options, but millions of people expressing opinions that appear to be guesses (e.g., gut feelings) produce a useful metric of value.
3:15 PM

37. Loewenstein’s alternative perspective
Overall, 2 themes:
Importance of subjective construals of situations, and
Importance of role of heuristics.
Question is, will we be able to have a science of such phenomena? We don’t yet know the answer.
Loewenstein (2001, p. 504)
“Gilboa and Schmeidler’s (1995) “case-based” decision theory posits that people make decisions by judging the similarity of the current situation to past situations and by applying decision rules that produced desirable results in similar situations in the past… There are also a number of important discussions of the role of choice heuristics. Prelec (1991, p. 133), for example, in an insightful article titled “Values and Principles,” draws a distinction between “action governed by cost-benefit calculation – however broadly the relevant costs aThe key step, in my opinion, will be to integrate these elements and to formalize them in a fashion that produces concrete, testable predictions.”
nd benefits are understood – and action governed by a binding rule or principle.”
3:15 PM