Affective valence as a computational signal for learning value

Area of Science:

• Cognitive Neuroscience
• Decision Science
• Affective Science

Background:

• Prevailing theories debate the role of affect in learning, with some viewing it as a byproduct of choice or irrelevant.
• Understanding how emotions impact learning and decision-making is crucial for a comprehensive model of cognition.

Purpose of the Study:

• To investigate the specific role of affect (valence) in value-based learning and choice.
• To differentiate the influence of valence on learning updates versus outcome evaluation during decision-making.

Main Methods:

• Utilized behavioral tasks combined with computational modeling across two experiments and a replication study.
• Analyzed trial-by-trial data to assess the predictive power of valence on choice and beliefs.

Main Results:

• Trial-by-trial valence significantly predicted both choice and beliefs about future outcomes, independent of experienced rewards.
• Computational models incorporating valence for learning updates demonstrated a better fit than models relying solely on reward.
• Valence's influence remained robust under uncertainty, during passive learning, and was particularly sensitive to socio-emotional information.

Conclusions:

• Affect (valence) plays a direct role in learning by shaping expectations and updating beliefs, not merely evaluating choices.
• These findings support a computational account of emotion in decision-making, highlighting its importance in social contexts.