Deliberative Move Recognition (DMR) detects multi-turn patterns in a Language Game that change deliberative state (e.g., assert → question → justify → revise), rather than labeling single utterances. It helps humans and AI spot the same Deliberative Primitive across media and contexts inside Living Documents.
# What It Detects - Sequences of moves: assert, question, justify, challenge, retract. - Roles in play: speaker, addressee, stakeholder, steward. - Commitment updates: new claims, obligations, metrics, deadlines, exit ramps. - Links between moves: support, attack, reference, approval, concession. - Outcomes: plan revision, harm mitigation, boundary pass/fail, consent achieved.
# How It Differs (Unit of analysis) - Dialogue Act Recognition labels single utterances (e.g., Statement, Question, Backchannel). - Deliberative Move Recognition detects multi-turn patterns that update commitments and plans.
# How It Differs (Ontology) - Dialogue Act sets follow standards like ISO 24617-2 (DiAML) or DAMSL. - Deliberative moves align with Argumentation structures (support/attack, critical questions) and commitment change.
# How It Differs (State tracking) - Dialogue Act pipelines rarely maintain a shared commitment store. - Deliberative Move Recognition maintains a live commitment graph and revision deltas.
# How It Differs (Goal) - Dialogue Act Recognition optimizes per-utterance tagging accuracy. - Deliberative Move Recognition optimizes decision quality and governance outcomes.
# Why It Matters - Makes reasoning patterns legible and teachable as Deliberative Primitive guides. - Connects talk to consequences via a shared commitment graph. - Enables gentle, in-flow prompts that suggest the next helpful move. - Supports auditability and learning across cases in Living Documents.
# Minimal Instrumentation - Tag each turn with a move type and role(s). - Record commitment changes (who owes what, when). - Link moves (supports, attacks, retracts, references). - Log outcomes (revisions, boundary checks, consents). - Keep a small evolution log in the document.
# Quick Checklist - Can the same pattern be recognized across text, audio, video, interactive? - Does detection trigger the right next step for participants? - Are commitment updates visible and reversible? - Do detected patterns correlate with better outcomes over time?
# Example Pattern - Propose → Challenge (evidence?) → Justify (model link) → Revise (parameters) → Commit (care ledger + monitoring) → Boundary Proof (pass).
# Related Deliberative Primitive · Language Game · Argumentation · Living Documents · Commitment Graph · Policy as Code
# References (external)
- ISO 24617-2 (DiAML) – dialogue act annotation 
- Dialogue act recognition – survey
- Switchboard – conversational corpus
- Dung 1995 – abstract argumentation
- Walton & Krabbe – commitment in dialogue
- AI mediator – common ground (Science 2024)
- CICERO – Diplomacy negotiation agent
- Rules as Code – primer