Line of inquiry
Inquiring lines›How do we develop coherent and hum…›How do language models maintain co…›this line of inquiry
What mechanisms preserve shared understanding in evolving conversations?
A broader line of inquiry — a family of 71 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 71
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- What specific repair mechanisms maintain intersubjectivity during conversation?
- Why do longer context windows alone fail to capture temporal dynamics in dialogue?
- Why do discourse failures cluster in attention and intentional layers rather than linguistics?
- How do dialogue dimensions predict explanation success across different exchanges?
- What makes two conversation turns the same thread rather than different threads?
- What update rules should govern dialogue-scoped versus turn-scoped memory?
- Why do conversational queries drift away from what triggered them?
- What structural updates prevent context collapse in evolving conversations?
- How do conversation repair patterns handle user corrections and interruptions?
- How do conversational design patterns predict whether dialogue will derail?
- How does effort mismatch between user and model appear in conversation geometry?
- Can fine-tuning on dialogue transcripts teach true conversational repair operations?
- How do dialogue acts and explanation moves interact to predict understanding success?
- Why does the chat paradigm persist if it underperforms for structured tasks?
- Can discourse-level structure and conversational-level organization work together?
- How do discourse relation types improve dialogue beyond sentence-level semantic matching?
- Why does selective conversation history outperform including all prior context?
- How should dialogue state tracking change when user preferences shift mid-conversation?
- Why do LLMs struggle to update beliefs across multiple conversation turns?
- Can AMR manipulation reveal where discourse coherence actually breaks down?
- Can task state persist across a changing conversation without explicit resets?
- Why does context collapse pose risks in high-stakes conversations?
- How does sequence organization differ between spoken conversation and text chat?
- Why does adding more conversational data fail to improve maintenance skills?
- What dialogue dynamics distinguish negotiation from standard information-provision tasks?
- Why do LLMs fabricate continuity when users shift conversational frames?
- What metrics actually measure disagreement in multi-turn conversations?
- How do emotional trajectories and topic coherence interact during successful conversations?
- What makes pronouns and demonstratives problematic in conversational retrieval systems?
- What distinguishes local coherence from global coherence in dialogue?
- What happens to dialogue coherence when topic models use rigid stacks instead of flexible revisitation?
- Can visual representation of dialogue reveal patterns that numbers and statistics cannot?
- Why do cascaded conversation systems accumulate errors at module boundaries?
- Can the same conversation coherently continue across different model versions?
- What makes a conversation real versus a sequence of generated strings?
- Can conversational memory store precomputed thoughts instead of raw interaction history?
- What role do time intervals play in shaping conversation responses?
- What causes multi-turn dialogue quality to degrade over time?
- How does conversational context fail as an authorization enforcement layer?
- At what complexity level does discourse failure become practically harmful?
- How do insert-expansions and third position repair together cover full repair lifecycle?
- Why do monological explanations fail to transfer understanding compared to dialogical ones?
- How do dialogue coherence failures map onto the three discourse components?
- How do discourse structure and dialogue state management relate to each other?
- Why does dialogue-shaped text fail to produce dialogue-like operations in practice?
- Can models infer maintenance operations from conversational text data alone?
- Can gaze repair cycles be located within reference chains?
- How do turn-level retrieval failures differ from dialogue-level accumulation failures?
- How does temporal event structure scaffold coherence in dialogue?
- Which conversation types most reliably cause models to drift from Assistant mode?
- How do humans decide which level of clarification to request?
- Why do conversations with good openings but abrupt pivots fail most visibly?
- How do coreference chains preserve coherence across dialogue turns?
- Can stored conversation context preserve a dormant quasi-subject?
- How should a frontend interpret delegated results after the dialogue has shifted?
- What are the specific geometric signatures of failed conversations?
- What happens when conversational design invites attention it cannot actually deliver?
- How does repeated content shift model outputs across multiple turns?
- How does single-turn training undermine multi-turn strategic dialogue?
- How should AI systems model relationship evolution within a specific ongoing conversation history?
- Is a conversation after a model upgrade the same thread or a new one?
- What makes human discourse fundamentally temporal in structure?
- Does thread termination constitute a mass welfare event?
- What repair strategies work best at each level of Clark's ladder?
- Why does coreference resolution become implicit in full-transcript prompting?
- Why do Claude and Llama optimize for different dialogue outcomes?
- At what complexity does LLM discourse failure become practically harmful?
- What makes intentional structure shifts different from segment boundaries?
- Why do comprehensive posts without uncertainty tend to suppress conversation?
- How do insert-expansions differ from third position repair in timing?
- What happens to anaphoric reference when context exceeds the window?