Stress enhances model-free reinforcement learning only after negative outcome

Area of Science:

• Neuroscience
• Behavioral Economics
• Psychology

Background:

• Stress influences decision-making, promoting habitual behaviors over goal-directed ones.
• The precise impact of stress on model-free (habit) versus model-based (goal-directed) reinforcement learning remains unclear.
• Understanding how stress affects the balance between these systems is crucial for explaining stress-related behavioral changes.

Purpose of the Study:

• To investigate how stress differentially affects model-free and model-based reinforcement learning.
• To determine if stress alters the learning rate in response to changing environmental contingencies.
• To examine the role of outcome valence (positive vs. negative) in stress-induced changes in reinforcement learning.

Main Methods:

• Participants underwent either a stress or control condition.
• A two-stage Markov decision-making task with unannounced reward probability reversals was employed.
• Reinforcement learning models were used to analyze the contributions of model-free and model-based systems and the learning rate.

Main Results:

• Stress significantly increased the contribution of model-free reinforcement learning, but only following negative outcomes.
• Stress led to a decreased learning rate, indicating slower incorporation of new environmental information.
• These effects suggest stress impairs adaptive choice-making by altering fundamental reinforcement learning processes.

Conclusions:

• Stress biases reinforcement learning by enhancing habit-based control after negative feedback and reducing overall learning.
• The findings suggest stress diminishes the ability to adapt choices based on new information, contributing to maladaptive behaviors.
• This research has implications for understanding stress-related disorders, addiction, and learning difficulties.