When attitudes and habits don’t correspond: Self-control depletion increases persuasion but not behavior

Conversational artificial intelligence (AI) can be harnessed to provide supportive parasocial interactions that rival or even exceed social support from human interactions. High-quality listening in human conversations fosters social connection that heals interpersonal wounds and lessens loneliness. While AI can furnish advice, listening involves the speakers’ perceptions of positive intention, a quality that AI can only simulate. Can such deep-seated support be provided by AI? This research examined two previously siloed areas of knowledge: the healing capabilities of human interpersonal listening, and the potential for AI to produce parasocial experiences of connection. Three experiments (N = 668) addressed this question through manipulating conversational AI listening to test effects on perceived listening, psychological needs, and state loneliness. We show that when prompted, AI could provide high-quality listening, characterized by careful attention and a positive environment for self-expression. More so, AI’s high-quality listening was perceived as better than participants’ average human interaction (Studies 1–3). Receiving high-quality listening predicted greater relatedness (Study 3) and autonomy (Studies 2 and 3) need satisfaction after participants discussed rejection (Study 2–3), loneliness (Study 3), and isolating attitudes (Study 3). Despite this, we did not observe downstream lessening of loneliness typically observed in human interactions, even for those who were high in trait loneliness (Study 3). These findings clearly contrast with research on human interactions and hint at the potential power, but also the limits, of AI in replicating supportive human interactions.

Can perceived responsiveness, the extent to which an individual feels understood, validated, and cared for by close others, reduce prejudiced attitudes? We hypothesized that perceived responsiveness by meaningful other people would increase recipients’ intellectual humility and attitude ambivalence and that these changes would reduce prejudice. Five studies (total N = 3362), four of which were preregistered, manipulated perceived responsiveness by a specific person (Studies 1–3, 5) or measured the effects of perceived responsiveness by the closest social network of the recipient (Study 4). All studies supported the hypotheses. Specifically, Studies 1 and 2 found that perceived responsiveness increased intellectual humility and attitude ambivalence and reduced prejudice toward a group from a pre-determined list. Study 3 replicated these findings when participants freely chose the social group. In Study 4, perceived responsiveness from individuals’ closest social networks predicted the dependent variables a few days afterward, controlling for positive and negative affect and social desirability. Finally, in Study 5, we added a condition of positive social interaction to rule out the possibility that the prior findings were due to recalling an affectively positive experience. The effect of perceived responsiveness on prejudice reduction (i.e., increased attitude favorability toward the social group) was not moderated by attitude certainty (Study 2), anxious or avoidant attachment style (Study 2), or attitude morality (Study 3). This work suggests that fostering perceived responsiveness can serve as a strategy for mitigating prejudice and promoting more open-minded attitudes.