How Humans Should Converse with AI: A Problem-First Adaptive Dialogue Conversion Protocol for Durable Human Capability (v1.3 glosa-checked edition)

Status. K0 theory, problem-first adaptive protocol and empirical research agenda; v1.3 is the glosa-checked edition of the 6 September 2026 v1.2 (internal citation table repaired so that each programme paper is cited to its own deposited record and version; "novelty" wording replaced by a contribution boundary; no content change). Part of the series When AI Expands Human Potential (index DOI 10.5281/zenodo.22308201). Companion papers: After Labour and The Human Conversion Imperative. Central claim. A human–AI conversation should be evaluated by the human state it produces after the conversation, and should begin from an agenda that remains human-owned. Outer architecture: Topic Entry → Triage → Minimum Sufficient Dialogue → Human Return → Action/World Feedback; inner eight stages Anchor → Expand → Oppose/Boundary-Test → Discriminate → Verify → Integrate → Remove → Return; DCP-Lite/Standard/Critical triage; ten hypotheses and nine removal conditions. All governing propositions are marked Open. An AI assistant assisted under the author's direction; no AI system is an author or contributor.