To choose between moves in chess, products in the supermarket, or whether to try out a new restaurant, an agent has to look ahead. Stochasticity in the outcomes of actions and the size of the decision tree can make such decisions difficult.
|Date||13 January 2016|
|Time||16:00 - 18:00|
I will examine three strategies that people might be using to overcome these difficulties:
1) reducing risk through accumulation of evidence over time;
2) heuristic (suboptimal) evaluations of states;
3) smart pruning of the decision tree. I will present behavioral tests of these strategies in a free-choice task between consumer goods, an exploration/exploitation task with "safe returns", and a board game.