This study employed a “risky helping task” combined with cognitive computational modeling techniques to investigate the intrinsic cognitive computational mechanisms underlying individual moral decision-making in risky contexts. Through two experiments, the moderating role of emotions (positive/negative) was further explored. Experiment 1 revealed that college students’ behavior in risky helping decisions was simultaneously influenced by two cognitive biases: loss aversion (α parameter) and guilt aversion (β parameter). As the risk of helping failure increased, individuals’ loss aversion intensified while guilt aversion weakened, leading to a decrease in the subjective value (ΔSV) assessment of risky helping options and a reduction in helping behavior. The turning point for risk aversion occurred at a 30% risk level. Experiment 2 further found that, compared to the neutral emotion group (risk aversion turning point at 40%), the positive emotion group exhibited a delayed risk aversion turning point at 60%, with significantly reduced loss aversion and guilt aversion. Moreover, they assigned higher subjective value to risky helping options and ultimately made more helping choices. The negative emotion group showed stronger loss aversion, assigned lower subjective value, and made less helpful choices compared with the neutral group. The results indicate that moral decision-making in risky contexts follows an integrated “loss-guilt aversion” processing model, where the risk of helping failure suppresses helping behavior by enhancing loss aversion, while positive emotions can effectively buffer the negative impact of risk. This study provides computational evidence for understanding the cognitive processes of moral decision-making and the role of emotional interventions.
