What are the four aims of decision making?
- Goals: aims, targets, needs, desires
- Options: 2+ possible behaviour acts
- Choice: among options to maximise goal attainment
- Consequence: what occurred
What is descriptive?
- Psychology: related to how people make decisions and what goals they have
- Through preferences and irrational behaviour
What is prescriptive?
How ought people make decisions and what goals should they have
What is normative?
- Rational decision analysis is normative: how should people make decisions to optimise goal occurrence e.g I want the best ice cream so must plan a strategy
- Decision analysis does not tell you what goals you should have
- Descriptive DM evaluated against a normative standard
- Descriptive DM indicates practical constrains on normative DM
What are the types of goals?
- Low-level biological needs: oxygen, water/food
- High-level psychological needs: acceptance, love, power
- Needs lead to goals: relationships, career, travel
- Achieving goals has ultity
- DA encourages evaluation: why am I doing what I am doing? BUT does not tell you what you ought to want
What biases might you have?
- Base-rate neglect
- Hindsight bias
- Conjunction fallacy
- Availability heuristic
How to decide to wait?
Cost of DA: time, effort, money relative to utility e.g importance
Options:
- Large number of possible behaviours
- Pragmatic constraints
- Must be mutually exclusive and exhaustive: one state specify can and must happen
How to pick a choice among options to maximise expected utility?
- What is the expected utility for each possible option?
- What ways can the world turn out in states if you pick that option: mutually exclusive but states need to be mutually exclusive and exhaustive
- By taking probability multiplied by utility to work out EU
- Must work out EU of each options of choice
-What outcomes: combo of options and states
How to determine consequences?
What happened in reality, the utility and the goals?
How to evaluate?
- Was it a good decision
- Were all options discovered, and was any critical information missing or inaccurate
- Were there some biases in my decision
How to calculate the Expected Utility of an Act?
EU(Ai)=ΣEU(Oij)= ΣU(Oij) *p(Sj)= U(O11) *p(S1) +… +U(Oij) *p(Sj)
What is the fair three card monty game?
- Three cards: queen and 2 other cards
- Can play the game or can choose not to
- Options: win/lose, and if you choose not to: win/lose
- If you play, costs you £1
If you do not play, how to maximise utility?
- States: win or lose, outcomes are play = win/lose, not play = win/lose
- Multiplication of rows U and p to create EU: not play has the better outcome, assessment means it’s better not to play the game
What is the evaluation of the fair three card monty game?
- Trivial decision: can afford to lose £1 and gut decision was right
- Cost of DA is greater than the possible loss of £1
- Not playing id boring: utility is more than money and fun is worth something THEREFORE inaccurate utilities as utility is subjective
- Were the probabilities accurate?
- Could have watched - not sure
- Inaccurate probabilities: could be a scam
What are some subtleties about decision analysis?
- Other options are possible: weird options that have no effect on the states and utilities should be ignored, wild options that have extremely negative utilities should also be ignored
- Options must be mutually exclusive, if not they need to be respecified
- Other weird states are possible: ignore states with vanishingly small probabilities UNLESS the utilities are extraordinarily large
Why is Maximising Expected utility the right/normative thing to do?
- Intuitive: with accurate information, you’ll get the most from a gambling perspective in the long run using MEU in context of decision analysis
- Formal: MEU follows logically from a set of axioms that seem true: cancellation, transitivity, dominance, invariance
- Inaccurate information leads to a poor decision analysis and a bad decision made
