Recognising and mitigating LLM Pollution in online behavioural research
Online behavioural research faces a growing methodological and epistemic threat as participants increasingly rely on large language models: LLM Pollution. Amid accumulating empirical evidence of contamination, we introduce a conceptual framework that distinguishes three variants — Partial LLM Mediation, Full LLM Delegation, and LLM Spillover. Their interaction distorts samples, biases inferences, and fuels an escalating methodological arms race. We outline mitigation strategies spanning researcher practices, platform accountability, and community adaptation.
Authors
- Iyad Rahwan (ORCID: https://orcid.org/0000-0002-1796-4303)
- Hiromu Yakura (ORCID: https://orcid.org/0000-0002-2558-735X)
- Anne-Marie Nußberger (ORCID: https://orcid.org/0000-0002-1805-9399)
- Tobias Werner (ORCID: https://orcid.org/0000-0003-2985-2760)
- Raluca Rilla (ORCID: https://orcid.org/0009-0007-7588-2262)
Institutions
- National University of Ireland, Maynooth (IE)
- Max Planck Institute for Human Development (DE)
- Max Planck Institute for Evolutionary Biology (DE)
- University of Southampton (GB)
Publication Details
- Journal
- Nature Communications
- Published
- 2026-06-25
- DOI
- https://doi.org/10.1038/s41467-026-74621-9
- Primary Topic
- Survey Methodology and Nonresponse
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Japan Science and Technology Agency
- Precursory Research for Embryonic Science and Technology