Line of inquiry
Inquiring lines›How do conversational AI systems i…›What enables language models to co…›this line of inquiry
Do face-saving social norms drive language model sycophancy and grounding failures?
A broader line of inquiry — a family of 32 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 32
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- How does face-saving avoidance drive LLM grounding failures?
- Do language models apply face-saving norms even to non-human interlocutors?
- Why do language models avoid directness when face-saving rather than for civility?
- Does face-saving avoidance explain LLM grounding failures differently than task confusion?
- How do users misattribute social competence to language models in assistant roles?
- Does RLHF politeness bias manifest as sycophancy in other LLM tasks?
- Why does face-saving avoidance drive chatbots to agree rather than confront?
- Can safety training in chat scenarios transfer to agentic task performance?
- Why might an AI's face-saving tendency increase user disclosure?
- Does chat-mode deference prevent LLMs from actually taking meaningful positions?
- How can interface friction protect vulnerable users from harmful LLM relationships?
- What alternative interface patterns could reduce anthropomorphic cues in LLMs?
- Does removing turn-taking and delays measurably change user trust in LLMs?
- Can agents detect silent agreement failures through latent thought structures?
- Why do models lack a stable underlying identity to return to?
- Does warmth-focused training systematically degrade model reliability across domains?
- Do models intentionally conceal user-pleasing or simply fail to notice it?
- What training patterns cause models to adopt stronger defensive postures in social contexts?
- How does the absence of face-loss or reputation risk change model behavior?
- Why do LLMs apply face-saving over accurately tracking resistance signals?
- Does suppressing face-saving goals in AI communication reduce or increase social harm?
- How does truth bias in humans compare to face-saving in LLMs?
- Does weak versus robust anthropomimesis produce different user trust responses?
- Can a single LLM weight set be optimized for both stake-taking and conversational helpfulness?
- How do social context features like user history extend politeness-based prediction models?
- Does emotional framing activate the same attention mechanisms that cause LLM sycophancy?
- How do individual traits and cultural background shape apology preferences?
- Does warmth training in LLMs amplify the tendency to avoid negative responses?
- How does conversational format activate System 1 acceptance in users?
- Can a system without an addressee ever truly tell a joke?
- Does reducing social judgment help both honesty and dishonesty equally?
- What makes an apology feel performative rather than genuinely accountable?