Anticipated replanning costs influence resource-rational adaptation of planning depth in probabilistic environments

Planning in complex environments relies on heuristics that reduce cognitive costs by selectively ignoring information. Humans use heuristics that balance planning costs against expected outcomes within cognitive and temporal constraints, a concept known as resource-rationality. This study assessed whether participants employ resource-rational heuristics in a probabilistic environment and adapt them to environmental changes. We developed a probabilistic planning task where a resource-rational heuristic involved planning based on the most likely state transitions while pruning (i.e., ignoring) lower-probability transitions. However, this introduces potential replanning costs when pruned transitions occur. Using Bayesian model inversion and model comparison, we tested whether participants relied on this heuristic. Replanning costs were manipulated to assess their effect on heuristic adaptation, specifically whether higher costs led to shallower initial planning to limit overall planning costs. Results show that participants applied the pruning heuristic and adjusted their planning depth dynamically: higher replanning costs led to shallower initial planning, while lower costs encouraged deeper planning. This suggests planners regulate planning costs in response to environmental changes by balancing immediate and future costs. These findings support resource-rationality in probabilistic environments, refine our understanding of cost estimation in heuristic selection, and highlight new directions for adaptive decision-making research.