Modeling Human Information Acquisition Strategies

The focus of this paper is the development of a computational model for intelligent agents that decides on whether to acquire required information by retrieving it from memory or by interacting with the world. First, we present a task for which such decisions have to be made. Next, we discuss an experiment that shows that humans do not apply rational expected utility analysis to make this decision, but instead adopt a simpler heuristic strategy. Then, we introduce a computational model that incorporates the rational as well as various heuristic task strategies. The human data is compared to the behavior of the model under several parameter settings. We were able to match the human actions with model actions for various task strategies, suggesting that humans differ in the task strategies they apply, and that our manner to deduce heuristic task strategies is feasible.