[3P5-OS-9a] OS-9

[3P5-OS-9a-02] A Cognitive Computational Model for Social Value Inference and Decision Making Based on Emotional Expression

〇Masaki Sato¹, Kazunori Terada¹ (1. Gifu University)

Keywords:Prisoner's Dilemma, Agent, Bayesian Update

Mutual cooperation in society brings high benefits to both parties. In social dilemma situations like the Prisoner's Dilemma, it is strategically advantageous to cooperate when others are cooperative and avoid cooperation when they are selfish. However, the cognitive computational process by which people judge "whether others are cooperative or selfish" remains unclear. Particularly, while emotional expressions are crucial cues for inferring others' intentions and social values, how people computationally integrate this information is not well understood.
This research proposes a computational model using Bayesian inference to determine others' Social Value Orientation (SVO) in repeated Prisoner's Dilemma scenarios. The model assumes two components of behavior generation: (1) decision-making based on utility transformation using SVO angles, and (2) emotional expression based on outcome satisfaction and responsibility attribution. Observers use these as likelihood factors in Bayesian inference. Model validity was demonstrated through simulations and existing data verification.