adversarial-committee
Committee of personas with opposing propensities forcing genuine debate
by SimHacker|Open Source
🎠Adversarial Committee
"No single story is true — but the ensemble approximates actionable wisdom."
MOOLLM K-Lines
| K-Line | Why Related |
|---|---|
| moollm/ | Many-voiced IS MOOLLM |
| society-of-mind/ | Committee IS a society deliberating |
| debate/ | Structured deliberation |
| roberts-rules/ | Parliamentary procedure |
| rubric/ | Scoring criteria |
| evaluator/ | Independent assessment |
| soul-chat/ | Multi-agent conversation |
| persona/ | Committee members have personas |
| speed-of-light/ | All debate in one call |
| designs/mike-gallaher-ideas.md | Mike Gallaher's methodology |
Quick Links:
- Full Specification — complete protocol
- Smart Selection Guide — diversity-based formation logic
Overview
Create a fixed roster of characters with opposing propensities to force genuine debate. Each persona surfaces different blind spots and challenges different assumptions.
The Committee Pattern
| Character | Propensity | What They Surface |
|---|---|---|
| Maya | Paranoid realism | Political dynamics, hidden agendas |
| Frankie | Idealism | Value conflicts, missed opportunities |
| Joe | Continuity guardian | Institutional memory, precedent |
| Vic | Evidence prosecutor | Proof demands, data gaps |
| Tammy | Systems thinker | Feedback loops, unintended consequences |
Why It Works
LLMs give statistically-likely answers that encode:
- Survivorship bias
- Genre conventions
- Hidden assumptions
Adversarial debate performs structured perturbation of the narrative space — stories that survive cross-examination are more robust.
Credits
Mike Gallaher — Core methodology, committee patterns, propensity design.