Opening Up to AI or Not: Compassion Mindsets Regulate the Intensity of Venting to AI Versus Human Friends

AI-powered chatbots are increasingly viewed as responsive and tireless conversational partners, raising the possibility that they may serve as alternatives to human friendsfor emotionally taxing self-disclosure, such as venting. Drawing on the disclosure decision-making model (DD-MM) and compassion mindset theory, the present research presents a qualified view on this “AI optimism.” Across two studies (N = 74 and 376), participants with a limited-compassion mindset, measured in Study 1 and experimentally induced in Study 2, vented more intensely to AI friends than to human friends. This pattern was partly explained by perceived compassion fatigue. In comparison, participants with a primed nonlimited-compassion mindset vented more intensely to human friends than to AI friends, partly because they perceived human listeners as more interpersonally accuracy. These findings suggest that disclosure to AI or human friends is goal-oriented and guided by the perceived affordances of each in intimate contexts. Implications, limitations, and directions for future research are discussed.