Investigating the relationship between affective valence and reinforcement learning

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Psychiatry

Background:

• The connection between affective valence and reinforcement learning (RL) is recognized but not fully understood.
• Existing research highlights potential links, but a clear mechanistic explanation is lacking.

Purpose of the Study:

• To investigate how reinforcement learning (RL) computations influence affective valence.
• To examine how affective valence, in turn, shapes RL processes.
• To develop and validate a computational framework for studying affect and RL.

Main Methods:

• Development of an original computational method to analyze affective experience during RL tasks.
• Modeling affective experience based on reward, prediction errors, and counterfactual comparisons.
• Behavioral analysis of choice preferences in relation to external rewards and affective responses.

Main Results:

• Affective experience in RL tasks is best explained by a combination of reward, prediction errors, and counterfactual comparisons.
• Affective responses serve as a reinforcement signal, influencing choice preferences alongside external rewards.
• Participants actively preferred options associated with more positive affect.

Conclusions:

• Reinforcement learning computations play a significant role in shaping affective experience.
• Affective valence acts as a crucial modulator of reinforcement learning and decision-making.
• The study provides mechanistic insights into the interplay of affect, learning, and choice, validating a novel computational framework.