Decision-making and general framework for judgement

Question 1

Decision requires …

Answer 1

1. Two or more courses of action
2. Uncertainty about events
3. Evaluations of the consequences of each possible outcome

Question 2

Rational expectations principle

Answer 2

Utility = Sum (Pr1 * V1)

Question 3

Incomplete thinking is due to…

Answer 3

Due to human limited capacity:

1. Focus on only 1 or 2 nodes of decision tree
2. Focus on gains and losses of node they initially like best (+ primacy effect)
3. Primacy effect

Question 4

Over-inclusive thinking (sunk costs)

Answer 4

1. Sunk costs are non-refundable past expenses
2. People honour them which violates 2nd principle of rationality
3. Why? Because fear of social disapproval (indecisive, wasteful) + Tendency to discard alternative nodes
4. The costs of the sunk cost can be taken in account but the sunk cost per se should not have an influence on future decision-making
5. What should you do? Make the best out of the situation - Good after bad instead of bad after bad

Question 5

How to find out future probabilities of a specific outcome

Answer 5

Multiply the probabilities of each previous outcome which must occur for that specific outcome to occur (Pr1 * Pr2 * Pr3…) - The previous outcomes will never change the probability of the future outcomes !

Question 6

Lens model

Answer 6

Left side - External world (perception) - Criterion to be taken into account

Right side - Psychological processes (cognitive computation) - To reach a judgement

Cue utilisation in the middle describes the psychological impact of each cue on the judgement - Generally done linearly, people weight the cue and add it up

Question 7

Why are human judges bad?

Answer 7

1. Valid for both professionals and amateurs
2. Rely on few cues
3. Most judgement policies are additive and linear
4. Judges lack insight - Cannot assign ex-post weight to the cues
5. Large individual differences, low inter judge agreements
6. More confidence in decision if non-diagnostic but associative evidence is presented

Question 8

3 principles of statistical models

Answer 8

1. Mathematical principles - Linear models approximate monotone relationships and monotone interactions fairly well
2. Principle of nature - Most relationships and interactions are in fact monotone
3. Psychological principle - People struggle to attend more than two non-comparable cues, they use anchor-and-adjust

Question 9

Why should/shouldn’t you use linear models

Answer 9

You should because:
- Better

You shouldn’t because:

• Protect expert’s narcissism
• Wrongful belief that they can predict long-term outcomes better
• 10 000 Frenchmen can’t be wrong - Circular argument

Solution? Let the judges choose which cues to take into account (left side of Lens model) and linear model compute them (right side of Lens model)

Question 10

Dummy and indicator variables

Answer 10

1. Dummy is for 2 variables, indicator is for more than 2 variables
2. Assign an indicator of 0 or 1 (or more depending on the number of variables) to each variable
3. If there is a monotone interaction, the lines will still have the same slope, they will only move a little up (or down)

Question 11

3 assumptions of decisions trees

Answer 11

1. We know numerical values
2. We do the mathematical operations
3. We do consider all options