Agent-Based Behavioral Modeling of Human Associative Learning in a Complex Approach-Avoidance Conflict Task.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Behavioral Economics

Background:

• Avoidance learning is central to anxiety disorders, yet its precise mechanisms are not fully understood.
• Current models often lack the complexity to capture nuanced human decision-making in conflict scenarios.

Purpose of the Study:

• To investigate the computational mechanisms underlying human approach-avoidance learning.
• To develop and validate a novel approach-avoidance conflict task for studying associative learning.
• To identify computational models that best predict human behavior in this task.

Main Methods:

• Developed a novel approach-avoidance conflict task with complex visual and outcome stimuli.
• Implemented and tested various agent-based models: control, heuristic, Rescorla-Wagner, and hybrid.
• Validated models against behavioral data from 50 human participants.

Main Results:

• A Rescorla-Wagner learning model incorporating a prior expectation bias parameter demonstrated superior explanatory power.
• The best-fitting model captured participants' associative learning within the novel task.
• Findings highlight the influence of task design on learning dynamics.

Conclusions:

• The Rescorla-Wagner model with expectation bias provides a strong computational framework for understanding approach-avoidance learning.
• This research underscores the importance of task-specific parameters in associative learning models.
• Contributes to a more detailed understanding of the etiology of anxiety disorders.