Door-in-the-Face Requests and Refusal Behaviour in Large Language Models
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.…
Opening of the original on arXiv cs.AI
Summary
Researchers explored how large language models like OpenAI's ChatGPT respond to the "door-in-the-face" persuasion technique. This method involves making a large, unreasonable request first, followed by a smaller, more reasonable one, to increase compliance. The study investigates whether LLMs exhibit similar refusal and concession behaviors to humans when presented with these sequential requests. Understanding this could refine how AI interacts with users and handles complex conversational strategies.
Why it matters
This research probes LLM's understanding of human social psychology and persuasion tactics. If models can be trained to recognize and respond appropriately to techniques like door-in-the-face, it could significantly improve their ability to negotiate, persuade, and manage user expectations in conversational AI. This affects how AI assistants interact with consumers and developers. Future work should examine if this behavior is consistent across different model architectures and training datasets, and how it might be exploited or mitigated.
Related: Google: Google’s latest AI weather model gives you no excuse to forget your umbrella · Anthropic: AGI IS HERE · Microsoft: What's new in AI? · Meta: Trump Administration Sides With OpenAI in New York Times Copyright Lawsuit · xAI: Sam Altman "AGI by December"
Rated low: routine. Worth knowing, not worth rearranging your day for.
- Sam Altman :‘AGI in 2026’, just as Models Start to [Mis]Train ThemselvesAI Explained · Web · Aug 27, 2026
- AGI IS HEREWes Roth · Web · Sep 3, 2026
- GPT-6 Astra Just Went CRITICAL...Wes Roth · Web · Sep 2, 2026
- ReDeck: Step-Level Render-Grounded Refinement for Document-to-Slide GenerationarXiv cs.AI · Web · Aug 31, 2026
- GPT‑6 AstraSimon Willison · Web · Sep 3, 2026
- Four major AI models suffer rare overlapping downtimeArs Technica AI · Web · Sep 3, 2026
- GPT-6 Astra Just Went CRITICAL...Wes Roth
- ReDeck: Step-Level Render-Grounded Refinement for Document-to-Slide GenerationarXiv cs.AI
- Sam Altman :‘AGI in 2026’, just as Models Start to [Mis]Train ThemselvesAI Explained
- SciTrue: Reliable Scientific Claim Validation with Frontier and Open Language Models at the NTCIR SciClaimEval TaskarXiv cs.AI
- EvoFlint: An Evolutionary Atlas of Multi-Turn LLM VulnerabilitiesarXiv cs.AI
- AgentProv: Auditing Agentic LLM API Providers via Tool-use Policy ProbesarXiv cs.AI
Questions people ask
- What is the 'door-in-the-face' technique?
- It is a persuasion strategy where a large, unreasonable request is made first, followed by a smaller, more reasonable request to increase the likelihood of compliance with the second request.
- What does this research investigate about LLMs?
- The study investigates whether large language models exhibit refusal and concession behaviors similar to humans when subjected to the 'door-in-the-face' request strategy.
More from arXiv cs.AI 12 more
Page generated Sep 3, 2026. Summaries are Subvolts' own; the story belongs to arXiv cs.AI.