Beyond Manipulation: How Users Perceive Harmful AI Chatbot Interactions

Key Findings

• — Exploratory factor analysis identified two dimensions of perceived harm: Experiential (cognitive and affective responses such as feeling deceived, controlled or taken advantage of) and Behavioral (user behaviours perceived as influenced by the chatbot, such as acting beyond original intentions)
• — Manipulative interactions scored higher on the Experiential Dimension than inappropriate interactions; the two conditions did not differ on the Behavioral Dimension
• — Qualitatively, manipulative interactions were associated with evaluative contexts and steering behaviours, whereas inappropriate interactions were linked to relational contexts and social discomfort
• — The authors draw implications for trustworthy-AI design and for regulatory frameworks including the EU AI Act

Methodology Notes

Proceedings of Mensch und Computer 2026 (MuC '26, Duisburg), DOI 10.1145/3820253.3831388, published 2026-08-29 per Crossref (OpenAlex lists 2026-08-20); CC BY 4.0. Both authors at the School of Industrial and Intelligent Systems Engineering, Tel Aviv University (Crossref affiliation strings). N = 100, recall-based self-report of a remembered interaction rather than observed logs; exploratory rather than confirmatory factor structure; adapted rather than newly validated instrument. dl.acm.org blocks automated fetchers from this workspace; verified via Crossref DOI metadata and the OpenAlex-deposited abstract.

Sources