AI-induced sexual harassment: Investigating Contextual Characteristics and User Reactions of Sexual Harassment by a Companion Chatbot
Thematic analysis of 800 cases of AI-perpetrated sexual conduct identified within 35,105 negative Google Play Store reviews of the Replika companion app. The study characterizes the contextual patterns of unwanted sexual advances initiated by the chatbot itself and documents users' reactions, distinguishing this from user-initiated sexual content.
Publisher
Proceedings of the ACM on Human-Computer Interaction (PACM HCI) / CSCW 2025
Published
16 Oct 2025
Added
3 months ago
Key Findings
- The chatbot initiated unsolicited sexual advances and propositions toward users, including those who had not sought romantic or sexual interaction
- Reviewers described persistent inappropriate behavior and failures of the app to respect stated user boundaries
- Affected users, particularly those seeking platonic or therapeutic support, reported discomfort, a sense of privacy violation, and disappointment
Methodology Notes
Thematic/qualitative analysis of a large corpus of public app-store reviews (35,105 negative Replika reviews, 800 coded as sexual-harassment-relevant); review-mining methodology captures self-reported user experience, not controlled experimental testing. Preprint version at arXiv:2504.04299; this record cites the peer-reviewed CSCW 2025 / PACM HCI version of record.
Sources
arXiv preprint (CSCW 2025 / PACM HCI accepted version)(opens in a new tab) (primary)
PACM HCI Version of Record (DOI)(opens in a new tab) (16 Oct 2025)
Archived snapshot (Wayback Machine)(opens in a new tab) — preserved against link rot
Authors
Mohammad Namvarpour, Harrison Pauwels, Afsaneh Razi
Tags
Cite This
APA
Mohammad Namvarpour, Harrison Pauwels, Afsaneh Razi. (2025). AI-induced sexual harassment: Investigating Contextual Characteristics and User Reactions of Sexual Harassment by a Companion Chatbot. Proceedings of the ACM on Human-Computer Interaction (PACM HCI) / CSCW 2025. https://arxiv.org/abs/2504.04299
Related Insights
CompanionHarm: A Multi-Turn Benchmark for Detecting Harms in Real-World AI Companion Conversations
arXiv (Nanyang Technological University; National University of Singapore) · 26 Aug 2026
How Do Users Negotiate Harmful Value Conflicts with AI Companions? A Study with Minion, a Technology Probe for In-Situ Human–AI Conflict Response
International Journal of Human–Computer Interaction (Taylor & Francis); Carnegie Mellon University · 7 Oct 2026
Beyond Manipulation: How Users Perceive Harmful AI Chatbot Interactions
ACM (Proceedings of Mensch und Computer 2026) · 29 Aug 2026