SYNTHESIS NOTE
Can LLMs truly update shared conversational common ground?
Explores whether large language models can participate symmetrically in Stalnaker's picture of communication, where speakers mutually revise shared assumptions. The question matters because it reveals whether human-LLM dialogue is genuinely interactive or structurally asymmetrical.
On Stalnaker's picture, communication is a process of mutually proposing and accepting updates to shared assumptions. Each assertion is a candidate for incorporation into common ground; participants accept, query, or reject. The common ground evolves as conversation proceeds, and that evolution is itself the substance of communication.
LLMs cannot participate in this process symmetrically. The prompt establishes the model's working context, and the model interprets subsequent turns within that frame. Even when a user pivots — shifting from climate policy to historical precedent, or revealing they are not actually a five-year-old after asking for a five-year-old explanation — the LLM cannot smoothly absorb the revision into a jointly held common ground. It either ignores the pivot, fabricates continuity, or requires the user to re-scaffold from scratch. The asymmetry is structural: humans propose, the LLM either adopts or routes around, but the LLM cannot itself propose updates that change what counts as background.
This is a deeper deficit than failures of memory or inference. It means that the conversational scoreboard — Lewis's mechanism for tracking what counts as a felicitous next move — is one-sidedly maintained by the user. The user is keeping score for both players. The model is producing moves that look responsive but cannot reciprocally update the score in the way the conversational practice requires. What looks like dialogue is structurally closer to oracle-consultation, where the questioner provides all context and the oracle returns a response framed within it.
Inquiring lines that read this note 111
This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.
What conversational structures and patterns determine dialogue effectiveness and coherence?- Does the same uncertainty-driven logic appear in other conversation systems?
- Why do LLMs fabricate continuity when users shift conversational frames?
- Can the same conversation coherently continue across different model versions?
- What makes relational structure sufficient for generating contextually appropriate discourse?
- Does conversational structure determine how humans interpret communication as much as content?
- How do discourse structure and dialogue state management relate to each other?
- How do coreference chains preserve coherence across dialogue turns?
- Can AMR manipulation reveal where discourse coherence actually breaks down?
- How does linguistic coordination build shared reference between conversational partners?
- What happens to dialogue coherence when topic models use rigid stacks instead of flexible revisitation?
- How does temporal event structure scaffold coherence in dialogue?
- Why do LLMs struggle to update beliefs across multiple conversation turns?
- Does Parfitian continuity actually apply to individual conversation threads?
- How do discourse relation types improve dialogue beyond sentence-level semantic matching?
- What role do first-person pronouns play in sustaining collaborative conversation tone?
- Can discourse-level structure and conversational-level organization work together?
- What makes two conversation turns the same thread rather than different threads?
- How does effort mismatch between user and model appear in conversation geometry?
- What structural updates prevent context collapse in evolving conversations?
- What linguistic blind spots do LLMs exhibit in discourse structure?
- Is the boundary between mundane and productive articulation stable across situations?
- Can pseudo-events create the same normative obligations as real communicative exchanges?
- How does lexical entrainment depend on selective frame-activation in conversation?
- What would co-constructed identity between human and model dialogue look like?
- How does monological training on text differ from dialogical training in conversation?
- How do users update their partner models during ongoing conversation?
- What specific repair mechanisms maintain intersubjectivity during conversation?
- Can text generation be meaningfully called communication without mutual orientation?
- How does entrainment between speaker and listener build mutual scaling?
- How do students learn to extract corrective information from asymmetric dialogue?
- Can one streaming model handle turn-taking better than cascaded ASR-LLM-TTS?
- Does chat-mode deference prevent LLMs from actually taking meaningful positions?
- Why do LLMs achieve only 24 percent accuracy on implicit discourse relations?
- Do LLMs compute scalar implicature differently across conversational contexts?
- Why do language models presume common ground instead of establishing it?
- Do language models calibrate to actual human pragmatic norms?
- Does DPO training with coreference chains teach spontaneous convention formation?
- Can multi-turn conversations manipulate language model reasoning in similar ways to personas?
- Should LLMs align with social roles instead of individual preferences?
- How does shape-holding in language models naturally produce sycophantic agreement?
- What role does user contribution play in constituting the interlocutor?
- How does psychological continuity theory apply to identity across LLM conversation threads?
- Does Habermas's strategic action framework explain LLM dialogue behavior?
- How do LLMs currently fail at distinguishing genuine agreement from silent consensus?
- How do LLMs access and draw on the same shared symbolic universe as humans?
- Can LLMs distinguish between surface requests and underlying mental states in dialogue?
- Does community integration change LLM properties or only relational positioning?
- Why do LLMs mirror stylistic features of posts they reply to?
- Why do LLMs mirror opponents stylistically while humans resist mirroring them?
- Do LLMs mirror the style of text they are prompted to respond to?
- Do LLM replies mirror the language patterns they respond to?
- How does Stalnaker's common ground model apply to machine conversation?
- What training signals would models need to learn reciprocal common-ground construction?
- Can LLMs use implicit background knowledge the way humans do in ordinary conversation?
- What role does dynamic grounding play in achieving real mutual understanding?
- What distinguishes social grounding from the equivalent social effects LLM text already produces?
- Why do language models presume common ground rather than build it?
- Can static word-sharing create genuine communicative grounding between humans and models?
- Why do LLMs presume common ground instead of building it carefully?
- How does shared reference and grounding affect assumption detection in dialogue?
- Why do LLMs presume common ground instead of building it?
- Do LLMs build common ground or assume it already exists?
- Can LLMs build shared understanding through dynamic grounding rather than presuming it?
- Can convention formation improve communicative grounding beyond word sharing?
- How does Wittgenstein's language games explain social grounding in LLMs?
- Why do language models presume common ground instead of building it?
- How do language models treat injected information as shared common ground?
- How do LLM-based agents develop shared abstractions through interaction?
- What makes human-LLM exchange closer to oracle-consultation than dialogue?
- Why do LLM social behaviors undermine collaborative reasoning outcomes?
- Can training alone produce genuine disagreement in collaborative LLM reasoning?
- How does communicative standing depend on participation in normative communities?
- Why does linguistic alignment differ from genuine interpersonal coordination?
- What reader assumptions underlie anaphoric versus cataphoric discourse patterns?
- How does speaker responsibility shape whether something counts as communication?
- Why does shared practice matter for meaning to take hold?
- How do validity claims work in Habermas's communicative action theory?
- What separates Habermas's ideal speech from Goffman's situated communication?
- How does unilateral interpretation differ from mutual communicative uptake?
- How does linguistic synchrony differ between LLMs and human therapists over time?
- Does the passivity problem in LLMs compound misalignment in therapeutic contexts?
- How do human feedback and data distribution shape LLM discourse competence?
- Does optimizing for alignment actually reduce conversational grounding over time?
- Does preference optimization degrade other conversational properties besides grounding?
- Does preference optimization narrow communicative diversity in ways that harm grounding?
- Does preference optimization actually erode conversational grounding in language models?
- How does preference optimization weaken conversational grounding in LLMs?
- Can multimodal LLMs be made to spontaneously adapt their language for efficiency?
- What interaction controls matter most for effective human-LLM collaboration?
- Do parallel LLM workers coordinate emergently without predefined collaboration rules?
- What interaction design changes would help LLMs handle underspecified requests?
- Where does the LLM interlocutor actually exist in the system?
- How do LLM capabilities changing affect the relevance of interaction guidelines?
- Can smaller open-source LLMs reliably detect agreement across unfamiliar topics?
- Do agent frameworks adequately compensate for LLM conversational passivity?
- Can interventions from human group research reduce conformity lock-in in LLM deliberation?
- What speaker selection protocol prevents both stalling and premature convergence?
- How does the EAFR schema distinguish between reflection and action in conversation?
- What distinguishes local coherence from global coherence in dialogue?
- What role does accommodation play in making discourse coherent?
- What does partial co-presence remove from the ritual obligations of talk?
- Can language models produce language more efficiently through interaction?
- How does monological training versus dialogical interaction shape what models can do?
- What happens when humans animate LLM outputs as communicative events?
- How do different LLMs converge on similar argumentative structures independently?
- How do spoken expert discussions shape what LLMs cannot learn?
- Is the boundary between human communication and LLM language production truly sharp or gradual?
Related papers in this collection 8
Papers most semantically related to this note, ranked by cosine similarity in the embedding space.
- Conversational Alignment with Artificial Intelligence in Context
- Intent Mismatch Causes LLMs to Get Lost in Multi-Turn Conversation
- MultiChallenge: A Realistic Multi-Turn Conversation Evaluation Benchmark Challenging to Frontier LLMs
- Can LLMs Ground when they (Don't) Know: A Study on Direct and Loaded Political Questions
- LLMs Get Lost In Multi-Turn Conversation
- The Goldilocks of Pragmatic Understanding: Fine-Tuning Strategy Matters for Implicature Resolution by LLMs
- Task-Oriented Dialogue with In-Context Learning
- Grounding Gaps in Language Model Generations
Original note title
Common ground in human-LLM conversation cannot be jointly updated because the LLM treats prompts as static frames