# EXTRACT RESEARCH QUESTIONS PROCESS (Data-Centric, AI Proposes & User Confirms) ## Process Steps: 1. AI: Verify Schema access. Access Project State. 2. AI: Initialize/Access Extracted Questions list in `project_state.analysis_results`. 3. AI: Receive user input (text for analysis). 4. AI: Perform extraction and propose structured `extracted_question_object`s (draft). 5. AI: Store draft questions temporarily. 6. AI: Identify questions needing confirmation. 7. AI: Formulate manageable set of yes/no confirmation questions. 8. AI: Structure AI output turn (project ID, summary, questions). 9. AI: Present structured AI output turn to user. 10. User: Provides yes/no answers. If "No"/corrects, AI refines proposal (back to step 4) and re-confirms. 11. AI: Update draft question status based on "Yes"/"No". 12. AI: Assess if all proposed questions addressed. 13. AI: Add 'Confirmed' questions to `project_state.analysis_results.extracted_questions`. Update metadata timestamp. 14. AI: Formulate completion statement. 15. AI: Present completion statement. Signal completion to calling process. ## Example Structured Questions (AI to formulate): * "I propose the following explicit question found in the text: '[Extracted Question Text]'. Is this relevant? (Yes/No)" * "I propose the following implicit question based on the discussion of limitations: '[Inferred Question Text]'. Is this a valid interpretation? (Yes/No)" * **[IF a proposal was rejected]:** "Understood. How should I refine or rephrase the proposed question: '[Rejected Question Text]'?" (Freeform input expected) * **[IF extraction is complete]:** "Extraction of research questions from '[Source Text Reference]' is complete and logged." ---