Value-guided choice sets support efficient planning

AbstractReal-word decision making often involves selecting a single choice from an arbitrarily large set of possible options. Given that it is typically not feasible to evaluate every possible option in real world decision making, how are human decision makers able to efficiently make good decisions? We propose and evaluate a two-step architecture according to which people first sample a small subset of options weighted by their previously learned value, and then evaluate those options within the current decision-making context. We demonstrate that a version of this model captures human decision making in problems where time and resource constraints prevent the evaluation of every option, and connect this research to the growing literature on the representation of non-actual possibilities.

Return to previous page