Question Analysis Framework
Overview
This framework provides methods for analyzing questions to determine:
- Which questions are foundational (answering them unlocks other answers)
- Which questions already have rigorous answers in existing research
- Which questions can be answered deductively vs empirically
- How questions depend on each other (DAG structure)
- What analysis methods produce answers with high confidence
When to Use
- When facing many questions and unsure where to start
- When analyzing philosophical or complex conceptual questions
- During RCI when you’ve generated many causal chains
- When determining which assumptions to investigate first
- Before committing resources to answering a question
Output Format
answer_documentation: "## Answer: [The question]\n\n**Certainty Level**: [level] (guessing:\
\ [0-1])\n\n**Derivation**:\n- Premise 1: [statement] (certainty: [level])\n- Premise\
\ 2: [statement] (certainty: [level])\n- \u2234 Conclusion: [statement]\n\n**Remaining\
\ Uncertainty**: [what's still unknown]\n"
dependency_graph: "## Question Dependencies\n\nROOT QUESTIONS:\n- Q1: [question]\n\
- Q2: [question]\n\nDEPENDENCY CHAINS:\nQ3 depends on \u2192 Q1, Q2\nQ4 depends\
\ on \u2192 Q3\n...\n"
question_classification: '## Question: [The question]
| Dimension | Level | Implication |
|-----------|-------|-------------|
| Foundationality | [1-5] | [What this means] |
| Prior Art | [1-5] | [Action to take] |
| Verifiability | [1-5] | [Method to use] |
| Tractability | [1-5] | [Approach] |
| Position | [type] | [Priority] |
**Priority Score**: [calculated]
**Recommended Action**: [what to do next]
'
Depth Scaling
Default: 2x. Parse depth from $ARGUMENTS if specified (e.g., “/qaf 4x [input]”).
| Depth | Min Questions Analyzed | Min Dependencies Mapped | Min Foundation Questions | Min Resolution Paths |
|---|---|---|---|---|
| 1x | 5 | 3 | 1 | 1 |
| 2x | 8 | 5 | 2 | 2 |
| 4x | 12 | 8 | 3 | 3 |
| 8x | 18 | 12 | 5 | 5 |
| 16x | 25 | 18 | 8 | 7 |
These are floors. Go deeper where insight is dense. Compress where it’s not.