City Research Online

Abstract

Bidirectional planning refers to a form of goal-directed decision-making process that combines forward and backward planning. Forward planning expands decision trees from the current state towards simulated futures, while backward planning starts the tree expansion from specific goal points in the opposite direction. Previous research has highlighted the impact of Pavlovian approach bias on behavior, showing that animals move towards appetitive outcomes regardless of the appropriateness of such behavior for achieving those outcomes. However, it remains unexplored whether the Pavlovian approach influences behavior by biasing backward planning. This research introduces a spatial navigation task to investigate the involvement of backward planning in humans' action-selection process and to determine whether the Pavlovian approach biases behavior through backward planning. The results reveal the behavioral signature of backward planning in humans and show that Pavlovian approach bias can influence both forward and backward planning, leading to decisions that are not necessarily instrumentally more efficient. Additionally, we developed a bidirectional planning algorithm based on reinforcement learning to simulate the participants' decisions. The simulation results suggest that the observed behavioral patterns can be parsimoniously explained by assuming that the Pavlovian approach bias acts as a pruning mechanism when expanding decision trees in both forward and backward directions.

Publication Type:	Conference or Workshop Item (Paper)
Publisher Keywords:	Reinforcement Learning; Computational Modeling; Bidirectional Planning; Decision-Making; Pavlovian Bias
Subjects:	G Geography. Anthropology. Recreation > GA Mathematical geography. Cartography Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments:	School of Health & Psychological Sciences School of Health & Psychological Sciences > Psychology
SWORD Depositor:	Symplectic Administrator

Export

Downloads