Temporal dynamics of model-based control reveal arbitration between multiple task representations

Published: June 15, 2024

Predominant frameworks categorize decisions dichotomously (e.g. “goal-directed” vs. “habitual”; “model-based” “model-free”). However, extensive work has shown that many human behaviors exhibit features of both systems, such as those require foresight (a goal-directed feature) but are not sensitive to environmental perturbations during action execution rigidity characteristic habits). Here, we introduce and explain a new subdivision behaviors, linking the format in which decision-maker represented contingencies memory. We this distinction by employing novel variant standard, two-stage decision task, allows us behaviorally capture within- across-trial dynamics planning. jointly fit choices response times with computational model revealed how people select among multiple task representations planning environments differing state-space complexity. In particular, examined reliance on changed function experience, within-subject, complexity, across-subjects (total n = 426). show complexity environment experience given contingency structure inform kinds use make decisions: at early stages start “conjunctive” (combining co-occurring first-stage states) simpler environments, “separated” representation (splitting states according their second-step outcomes) is preferred more complex environments. With pattern reversed. Finally, shift governed change approaches optimizing reward rate: initially, focus minimizing uncertainty, once reached asymptote, they transition prioritizing efficiency. Taken together, only arbitrate between different modes control, also types for efficient

Language: Английский

Memory precision and age differentially predict the use of decision-making strategies across the lifespan.

Published: May 31, 2023

Memory function declines in normal aging, a relatively continuous fashion following middle-age. The effect of aging on decision-making is less well-understood, with seemingly conflicting results both the nature and direction these age effects. One route for clarifying mixed findings to understand how age-related differences memory affect decisions. Recent work has proposed sampling as specific computational role decision-making, alongside well-studied mechanisms reinforcement learning (RL). Here, we tested hypothesis that episodic alter sampling. Participants (total N=361; ages 18-77) performed one two variants standard reward-guided decision experiment additional trial-unique mnemonic content separately-administered task assessing precision. When fit participants’ choices hybrid model implementing memory-based RL-driven valuation side-by-side, found precision tracked contribution choice. At same time, corresponded decreasing influence RL increasing perseveration. A second confirmed further revealed sampled memories. Together, suggest across lifespan may be related function, interventions which aim improve former benefit from targeting latter.

Language: Английский