Week 6: Decision & Games - Decision Theory Flashcards
Acts, states and outcomes
Acts: actions you might take; States: mutually exclusive, exhaustive possibilities of how the world might be; Outcomes: results of doing an act.
Preferences
A ranking or evaluation of outcomes.
Probabilities
Evaluation of the likelihood of states (optional).
A priori truth
A proposition that is true by definition, independently of experience or observation.
Strong/weak dominance
Strong dominance: Act A is better than Act B in every state; Weak dominance: Act A is at least as good as Act B in every state, and better in some state.
Dominance principle
Never take dominated acts.
Maximax
Optimist’s rule; Maximize maximum: choose the action with the best possible outcome.
Maximin
Pessimist’s rule; Maximize minimum: choose the action whose worst outcome is the best.
Regret
Difference between the value of the action at that state and the value of the best action at that state.
Minimax regret rule
Minimize maximum regret: choose the action with the lowest possible highest regret.
Expected utility of an action
Sum of utilities of possible outcomes weighted by the probability of each outcome.
Decision under risk
Consensual rule: Maximize expected utility.
Difference between Decision Theory and Game Theory
Decision Theory: study of decisions made by an individual; Game Theory: study of decisions made by multiple individuals in a strategic context.
Decision rules
Guidelines that select or rank acts in decision problems based on the structure of the problem.
Newcomb Problem: Causal theorist and evidentialist
Causal theorist: choose one box, focusing on causal influence; Evidentialist: choose two boxes, focusing on correlation.
