A Participatory Methodology for Human-AI Co-Inquiry: Full Methodology and Lite Companion

This deposit presents a proposed practice methodology for structured human–AI inquiry, developed from Kon Lionis’s A Participatory Framework for Human-AI Co-Inquiry. The methodology organises an inquiry episode around five stages: orient and frame; explore and distinguish; apply productive friction; ground and diagnose; and audit, revise, and close. It combines exploratory dialogue with explicit assumptions, substantive challenge, appropriate checking, and traceable revision. A central distinction is between a claim’s category and its verification status: identifying something as a factual claim does not establish that it is true. The procedure also separates completing the required steps from answering the question. Agreement, fluent explanation, and complete documentation do not by themselves establish evidential support. Two complementary documents are included: • Full methodology — working draft v0.2, dated 13 October 2026. Sets out the stages, participant responsibilities, claim ledger, verification priorities, closing decisions, documentation requirements, and proposed evaluation approach. It includes guidance for assessing understanding and developing bounded research leads. • Lite companion — working draft v0.3, dated 14 October 2026. Provides a self-contained prompt block for interactive use in an AI conversation, with Quick and Standard depths and referral to the full methodology for consequential inquiries. It retains the central safeguards while reducing record-keeping. Both versions require assumptions and hypothetical premises to retain their status throughout the conversation. When evidence, tools, time, or expertise are insufficient, conclusions must be narrowed, made conditional, or suspended. Available closing outcomes are Close, Iterate, Escalate, and Abstain. These documents are working drafts offered for practical use, criticism, adaptation, and empirical evaluation. They do not report a completed validation study or establish improvements in accuracy, understanding, or efficiency. They make no claim that AI systems possess subjective awareness or that a model’s limitations define the limits of human knowledge. Question everything, including the question.