Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session C: Friday, August 15, 2:00 – 5:00 pm, de Brug & E‑Hall

Social context affects policy and counterfactual learning in a two-armed bandit task

Amric Trudel¹, Bence C Farkas², Pierre O. Jacquet, Valentin Wyart³; ¹Ecole Normale Supérieure – PSL, ²Forward College, ³Ecole Normale Supérieure de Paris

Presenter: Amric Trudel

Social contexts shape decision making under uncertainty. While reinforcement learning models explain how individuals learn from rewards, it remains unclear how this process is affected when rewards come from other agents. Here we introduce a novel social decision-making task based on the two-armed bandit paradigm, which allows the isolation of social learning mechanisms while keeping reward structure exactly matched between social and non-social contexts. Model-free analyses revealed increased behavioral switching and reduced blind repetition in the social condition. A reinforcement learning model incorporating a counterfactual learning parameter revealed that the social context primarily altered policy parameters and counterfactual updating, suggesting participants imagined a competitive dynamic between their partners. Our findings indicate that while learning mechanisms about chosen options remain relatively stable across conditions, social framing seems to shift how people explore and infer from unchosen outcomes. This advances our understanding of the cognitive architecture underpinning human social learning.

Topic Area: Reward, Value & Social Decision Making

Extended Abstract: Full Text PDF