A Bayesian brain model of adaptive behavior: an application to the Wisconsin Card Sorting Task

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Neuroscience

Background:

• Adaptive behavior relies on cognitive agents interacting with dynamic environments.
• Assessing cognitive processes often uses tasks like the Wisconsin Card Sorting Test (WCST), but current scoring methods are limited.
• Understanding information processing in adaptive tasks is crucial for cognitive and clinical insights.

Purpose of the Study:

• To propose and validate a novel computational Bayesian model for analyzing individual performance on the WCST.
• To formalize the information processing mechanisms underlying rule inference and adaptation in the WCST.
• To integrate the model within the Bayesian Brain Theory (BBT) framework.

Main Methods:

• Developed a computational Bayesian model to capture information processing in the WCST.
• Embedded the model within the Bayesian Brain Theory (BBT) framework for dynamic belief updating.
• Validated the model through extensive simulations and real behavioral data analysis.

Main Results:

• The model successfully accounts for individual performance variations in the WCST.
• It decomposes adaptive behavior into separable, neurobiologically plausible, information-theoretic constructs.
• The model recovers trial-by-trial cognitive dynamics at an individual level.

Conclusions:

• The proposed Bayesian model offers a powerful tool for understanding adaptive behavior and cognitive dynamics in the WCST.
• It provides a framework for mapping cognitive processes to information-theoretic metrics and potential neuroanatomical correlates.
• The model has significant empirical benefits for clinical assessment and model-based neuroscience research.