【公开学术报告-第165期】Distrust Despite Human Oversight: AI Assistance Undermines Trust in Feedback

Distrust Despite Human Oversight: AI Assistance Undermines Trust in Feedback

【即便有人工把关，依旧难以信任：AI 辅助会降低大家对评价反馈的信任感】

Speaker: Mingyu Li, HKUST

Date & Time:  Wednesday, 18 November 2026, from 14:00 to 15:30 AM (Beijing Time)

Location: Tongji Building A2101

ABSTRACT

Human-in-the-loop (HITL) systems are widely promoted for combining the strengths of AI and human decision makers, yet little is known about how individuals evaluated by them respond. Across four preregistered experiments with U.S. adults (N = 2,413), we find that HITL feedback is trusted less than feedback from humans working alone. The gap holds when all feedback is in fact produced by AI, and only the stated source varies, as well as when HITL feedback is of higher quality as judged by blinded third-party evaluators, an LLM, and NLP text analysis. Recipient distrust tracks perceived warmth: AI involvement lowers how warm the provider seems, and expressed warmth converts into trust only when the feedback is believed to come from an unassisted human. Warmth conveyed by an AI-assisted source is discounted, and human oversight alone does not make AI evaluation more trusted.

Keywords: Trust, Disclosure, Human-in-the-loop (HITL), Feedback, Warmth

GUEST BIO

Mingyu Li is a Ph.D. candidate in Management at the Hong Kong University of Science and Technology. She holds an M.Sc. in Human Resource Management from LSE and a B.Sc. in Business Administration from Shandong University. Her research centers on trust, AI‑enabled management, the future of work, and negotiation. She has published in journals including the Journal of Applied Psychology, received the Hong Kong PhD Fellowship, and presented widely at leading international academic conferences.