Looking deeper into the algorithms underlying human planning

Abstract

Humans possess a remarkable ability to form sophisticated multi-step plans even in complex environments. In this review article, we consider efforts that attempt to characterize the mechanisms underlying human planning using a computational framework, primarily focusing on methods that search a tree of possible solutions. These studies range from experimental probes for heuristics that people employ while thinking ahead to normative models for reducing the computational costs of planning. Additionally, we examine the recent successes of artificial intelligence in the domain of planning and how these innovations can be applied to better understand human sequential decision-making. As examples, we highlight this approach in two tasks that require planning many steps into the future, namely 4-in-a-row and chess.