This is a computational study of spatial planning under uncertainty using a novel Maze Search Task (MST), in which people search mazes for probabilistically hidden rewards. The MST is designed to resemble real-life spatial planning, where costs and rewards are physically situated as distances and locations. The researchers found that people’s spatial planning is best explained by planners with limited planning horizon, as opposed to both myopic heuristics or the optimal expected utility, showing that a limited planning horizon can generalize to spatial planning tasks. They also find that our results do not exclude the possibility that in human planning action values may be affected by cognitive mechanisms of numerosity and probability perception.

This funding will help enable the Atlas Network to scale up the pilot versions of George Ayittey Society and the Champion’s Summits.