Arbitrating between planning and habit in naturalistic environments

TitleArbitrating between planning and habit in naturalistic environments
Publication TypeConference Paper
Year of Publication2019
AuthorsMugan, U., and M. A. MacIver
Conference NameConference on Cognitive Computational Neuroscience
Date Published09/2019
Keywordshabit, Markov decision processes, planning, predator-prey interactions
Abstract

Research into the neural basis of decision making suggests that animals engage in habit- or plan-based action selection. The existence of these two modes of action selection raises an additional problem: how should animals arbitrate between them to efficiently allocate their time and computational resources? Here, we use a naturalistic task, spatial planning of a prey evading a predator in environments with varying complexity, to investigate how the arbitration between these decision making systems should be done. We identify a key signature of complex environments where planning becomes imperative— transitions between poorly and highly connected regions. We suggest an efficient approach, based on environmental connectivity, that switches between plan- and habit-based control during a task. This approach provides a unifying account of experimental data that shows vicarious trial and error at high cost choice points, as well as increased theta coherence between the hippocampus and the prefrontal cortex at transitions from closed to open regions—both situations where there is a transition in spatial connectivity.

Publication PDF: