journal article
Countering division with friendliness: How feeling understood by a friendly AI triggers both openness and resistance
Abstract
Fostering understanding across divides could counter societal polarization but also risks reinforcing existing views. Using AI chatbots to experimentally induce interpersonal experiences, it was tested whether feeling understood enhances openness to opposing information. Across four studies (N = 1839), participants engaged in 6-min chatbot conversations before reading articles that challenged their vaccination or climate change views. These conversations were experimentally varied to be neutral, deflective, friendly, corrective, or understanding-focused. Both understanding-focused and friendly conversations increased feelings of understanding, enhancing perceived credibility of opposing information and predicting counter-attitudinal behavioural intentions—mediated paths that persisted at a 60-day follow-up. While experimental mediation effects were small but consistent (β = 0.02–0.06), correlational relationships were robust (β = 0.12–0.41). The effects were strongest among anti-vaccination and climate-sceptic participants. However, despite producing the strongest initial effects on credibility and intentions, random topic friendly conversations also generated suppressed negative direct effects that became apparent at follow-up, suggesting dual systems: one that automatically responds to social cues, and another that simultaneously detects inauthenticity. Factual corrections, initially without impact, showed positive effects at follow-up. These findings illuminate feeling understood as fundamental to bridging divides—powerful enough that even artificial displays activate openness, yet suggesting potential resilience through simultaneous authenticity detection, with implications for defending against cognitive warfare. © 2025
Author keywords
ai chatbot; cognitive warefare; conspiracy theories; conversational receptiveness; counterattitudinal behavior; division; elaboration likelihood model; feeling heard; feeling understood; misinformation; openness; persuasion; polarization