Figures and data
Behavioral results.
(A) Participants took shorter to make decisions as the probability level increased in risk task but not in cooperation task. (B) Increased principal magnitude decreased the decision time more pronouncedly in cooperation task than in risk task. (C) Fixed effects of probability and principal on the decision time as a function of task type. (D–E) Participants were less willing to take risks in the cooperation task than in the risk task, with the effect being more pronounced at high principal magnitude. (F) Fixed effects of probability on the acceptance rate as a function of task type and principal level (low vs. high). Shaded areas depict the 95% confidence intervals.
Computational results.
(A) AIC values (bars; lower values indicate better fit) from the comparison of the computational models. The insert shows the distribution of the model selection across participants. Model 3 (incorporating CPT, betrayal aversion, and prosocial utility) was selected as the winning model. (B) Parameter recovery of Model 3: gain sensitivity (α_gain), loss sensitivity (α_loss), probability sensitivity (γ), loss aversion (λ), noise (β), betrayal aversion (κ), and prosocial preference (ω). The correlation matrix shows the Pearson correlation between the actual recovered and simulated data from the winning model. The high correlations (all r > 0.80) indicate excellent parameter recovery. (C) The distribution of the best loss (top) and defect (bottom) parameters for the winning model, which maximizes log-likelihood during fitting. (D–E) Decision time was becoming shorter as loss aversion increased, and betrayal aversion decreased, and the willingness to cooperate became lower as loss aversion and betrayal aversion increased. Betrayal aversion significantly predicts both decision time (D) and acceptance rate (E). Shaded areas depict the ± 1 standard errors. (F–G) Focusing more on others’ benefits was associated with increased betrayal aversion (F), but did not affect loss aversion (G). Loss represents the subject value for the best-fitting financial loss aversion (−λ ∗ (0.5 ∗ V)αloss ∗ PWeight) and defect is the subject value of betrayal aversion κ ∗ PWeight). ***p < 0.001, **p < 0.01.
Grand-average ERP waveforms and topographic maps of the P3 and LPP as a function of task type (risk task vs. cooperation task) separately for probability (A) and principal (B) trials.
Gray shaded bars represent time windows used for quantification. The insets show the fixed effects of betrayal and loss aversion on the P3 (A) and LPP (B). A betrayal-inhibition effect emerged during the early P3 phase and persisted through the late LPP phase, while the loss-inhibition effect appeared solely during the late LPP phase. Shaded areas depict the 95% confidence intervals.
Experimental tasks and rating results.
(A) The risk task (top). Participants chose a reject option, keeping their current principal (¥0.2, ¥0.4, ¥0.6, ¥0.8, or ¥1.0), and an “accept” option, where they gambled to double their principal, with a varying probability (10%, 30%, 50%, 70%, 90%) of success, or lose half the principal if they fail. The cooperation task (bottom). Participants as agents chose whether to cooperate with their partner, with the same probabilities and payoffs as the risk task. If successful, both agent and partner doubled their principal; if it failed, the partner took half of the agent’s principal. ISI = interstimulus interval; ITI = intertrial interval. (B) The decision tree depicts an example of a decision to cooperate, and the partner could choose cooperation or betrayal. (C) The distribution of the accepted numbers during the risk task and the cooperation task. (D) Rating data. Participants felt more liking for positive outcomes and more disliking of negative outcomes during the cooperation task than the risk task. Error bars represent the within-subject standard error of the mean.
Results of linear mixed-effects models predicting decision times (left) and choices (right) as a function of probability, principal, and task type.
Model comparison
Parameter recovery analysis
Results of linear mixed-effects models predicting decision times (left) and choices (right) as a function of loss aversion and betrayal aversion.
