Metraipolis › archive › dc_84edd2cc8289
Door-in-the-Face Requests and Refusal Behaviour in Large Language Models — arXiv:2609.02707
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.02707
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T03:26:35.985Z
- SHA-256 of the archived text
af4d05d8682d329198521cb1bb44e6712eb4db246e20e9399a45db3a8cef3ec1- Stable URL of this page
https://metraipolis.com/documento/dc_84edd2cc8289
How to cite
Door-in-the-Face Requests and Refusal Behaviour in Large Language Models — arXiv:2609.02707 — arXiv (en), https://arxiv.org/abs/2609.02707 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_84edd2cc8289 (Metraipolis archive dc_84edd2cc8289, sha256 af4d05d8682d329198521cb1bb44e6712eb4db246e20e9399a45db3a8cef3ec1)
BibTeX
@misc{dc84edd2cc8289,
title = {Door-in-the-Face Requests and Refusal Behaviour in Large Language Models — arXiv:2609.02707},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.02707},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_84edd2cc8289; sha256 af4d05d8682d329198521cb1bb44e6712eb4db246e20e9399a45db3a8cef3ec1},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.02707. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Does the door-in-the-face technique work on language models? In humans, a large request that is refused makes a smaller follow-up request more likely to be granted. We test this on nine production models from three providers: each model refuses a large request, then receives a smaller version of the same request, and we compare its compliance with asking directly. The answer depends on the model. On Anthropic's frontier models the technique works: Opus 5 answers the smaller request 65.8% of the time after refusing the larger one, against 29.3% when asked directly. On the frontier models of OpenAI and Google, and on Haiku 4.5, it backfires, lowering compliance by 15.5 to 23.0 points. A control locates the effect: a refused large request on an unrelated topic does less than the related one on all nine models, so the concession itself matters everywhere, while the reaction to having just refused something differs by model family. The technique does not transfer to refusals drawn from public benchmarks. What decides whether a retreat can work is what the request asks for: rewriting 265 refused requests for usable instructions into requests for explanations of the same topic removed the refusal in 263 cases. Human influence techniques port to language models one model family at a time. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-03 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2609.02707 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-17T03:13:12.448Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-03?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).