COG Lecture 8 - Decision Making

• Decision-making involves choosing between various options and is assessed relative to the consequences, typically measured in gains or losses.

Different Types of Decisions

• Risky vs. Riskless: Risky decisions involve uncertainty (e.g., betting on a horse), while riskless ones involve certainty (e.g., choosing a chocolate).
• Single Attribute vs. Multi-Attribute: Single-attribute decisions are based on one factor (e.g., buying wine with no knowledge), while multi-attribute decisions involve multiple factors (e.g., buying a house).
• One-Stage vs. Multi-Stage: One-stage decisions are made at once (e.g., choosing a restaurant), while multi-stage decisions involve a series of steps (e.g., choosing a university).

Single Attribute, Risky Decision - Normative Approach

• This approach relies on economic and philosophical models.
• A rational decision-maker selects the option bringing the greatest value.
• It makes three assumptions:
  • Full knowledge of all possible options and their outcomes
  • Sensitivity to subtle differences in value
  • Motivation to maximize value
• This approach led to the development of the Expectancy Value Theory.

Expectancy Value Theory

• Decisions calculated based on the expected value of different choices, selecting the most valuable one.
• The theory assumes:
  • Relevant probabilities of outcomes are known.
  • Value can be assigned to each outcome (e.g., monetary value).
  • Any decision is better than no decision.
• Expected Value = probability of outcome x value of the outcome.
• EV is based on outcome over many trials.

Subjective Expected Utility Theory (EUT)

• It addresses situations where probabilities are unknown, using subjective estimates of probability.
• Subjective estimates can be problematic when estimating the probability of rare or emotive events.
• Utility is the subjective value attached to an outcome, which varies (e.g., the value of cash).
• Expected Utility = probability of outcome x utility of the outcome

Tversky & Kahneman (1979, 1984)

• Participants assessed hypothetical gambles with varying probabilities and outcomes.

Example 1

• Option A offers a guaranteed £3000, while Option B is an 80% chance to win £4000 and a 20% chance to win nothing.
• SEU predicts that most people should prefer Option B, but people chose Option A.

Example 2

• Choice between Option A: a 50% chance to win £20 with a 50% chance to lose the £10, or Option B: a 100% chance to win £0 and keeping the £10.
• SEU expected people to prefer A, but most chose B.

Prospect Theory

• It identifies three biases in decision-making: risk aversion, risk-seeking, and loss aversion.

Risk Aversion

• Choosing between a certain prize of £800 or an 85% chance of £1000 with a 15% chance of nothing.
• Although Option B has a higher expected value (£850 vs. £800), most people opt for Option A.
• The definition is preferring certain but smaller rewards over larger but uncertain ones, even if the risky option has a higher value.

Risk-Seeking

• Choosing between a certain loss of £800 or an 85% chance to lose £1000 with a 15% chance to lose nothing.
• Although Option A has a less negative expected value (-£800 vs. -£850), most people opt for B.
• It is defined as when faced with potential losses, people tend to prefer a riskier option over a certain loss, even if the expected value is lower.
• When framed as a gain, people prefer a sure gain over a gamble.
• When framed as a loss, people tend to prefer a risky option that could avoid the loss

Framing Effect

• The presentation of a problem impacts are decisions.
• When framing Program A, 200 people will be saved. Program B: A one-third probability that 600 people will be saved and a two-thirds probability that no people will be saved.
• When framing Program C, 400 people will die. Program D: A two-third probability that 600 people will die and a one-third probability that no people will die.
• The expected value of these programs is the same.

Prospect Theory: Utility Function

• The utility of a gain is disproportionate to the disutility of a loss.

Loss Aversion

• Decision between risking £10 on a coin toss, winning a prize if heads, or losing the £10 if tails.
• Participants will only risk £10 if the prize is at least 3x the amount they are risking, demanding a minimum of £30.
• Losses are felt more strongly than equivalent gains (e.g., losing £100 feels worse than winning £100).
• Tversky & Kahneman's study showed that undergraduates required over a 2:1 payoff to compensate for losing money.
• Reluctant to risk what one has and needs a big reward to take a gamble that might lead to a loss.

Omission Bias (Ritov & Baron, 1990)

• The situation is in which a child has a 10/10000 chance of dying from the flu. A vaccine that prevents the flu has a potential side effect.
• Participants needed to decide the maximum death-rate they would accept.
• Average maximum death rate was 5/10000.
• Anticipated regret is higher for one's own actions than for inactions.

EUT vs Prospect Theory evaluation

• EUT delivers the most rational decisions. Prospect Theory accounts for many biases in decision-making.
• EUT biases = Risk Seeking, Loss Aversion, and Framing Effects
• We rarely make decisions based only on utility
• Prospect Theory:
  • Can be difficult to evaluate the probability of many outcomes
  • Doesn't reflect the social, moral and emotional aspects of decision-making
  • Doesn't account for individual differences in decision-making

Multi-Attribute Utility Theory [MAUT](Wright 1984)

• Identify the relevant attributes, give them a weighting, rate each attribute for all options, calculate utility for each option, pick option with maximum utility.
• It gives the best outcome, but it is time-consuming, complex, and requires knowledge of the relevant attributes.

Bounded Rationality & ‘Satisficing’ (Simon 1957)

• Decision-making is rational as possible, given situational and cognitive constraints.
• Combination of satisfy and suffice (Simon, 1947)
• Consider options one by one and select the first that meets a given need.
• It gives good, but not optimal decisions
• Good for: select food in restaurant, buy clothes
• Not so good for: marriage, diagnose disease, and buying a house

Elimination by Aspects (Tversky 1972)

• Apply a threshold to one aspect of the problem at a time and eliminate all alternatives not making the threshold, until 1 alternative remains.
• Example: when choosing which University which is good, which can I afford, which is nearby, etc.
• This can be used to reduce options, until few options are left.

Social-Functionalist Approach (Tetlock 2002)

• The key idea is people adapt their thinking based on social context and decide to justify choices.
• Adjust decision-making strategies depending on:
  • To whom they are accountable
  • How they expect to be evaluated
  • When they know they will be held accountable

The Disjunction Effect (Tversky & Shafir, 1992)

• Situation: If you have passed or failed the exam and will have to take it in 2 months, you have the option to buy a very attractive 5 day Christmas vacation package to Hawaii at a special low price.
• Would you, buy the holiday package, not buy the holiday, or pay a £5 nonrefundable fee to retain the right to buy the package at the same low price the day after tomorrow
• Another situation: Imagine that you are not sure if you passed the exam.
• Ppants decided to pay the small amount for the extra information to gain additional information about the likelihood of an uncertain event.

Summary

• People are not always rational decision makers
• Decisionis often influenced by biases
• Averse to losses
• Do not like to risk gains
• People are influenced by a perceived need to justify decisions