Reasoning Council

What ships when the mind that generates the options also scores and judges them?

Problem: a single reasoning pass locks in — blind spots ship, and the captain must curate what should have arrived argued.

Question: which separated perspectives must argue a call — with the incumbent defended and the judge apart from the debaters — before it reaches the chair?

Decision: whether a judgment call ships as delivered, ships revised, or dies — decided by a chaired verdict over scored options instead of a self-judged conclusion.

This method produces the Debate Meeting's Before artifact: 3–5 scored options with full-sentence rationales and recorded dissent, ready for the human Council to chair.

Inputs

• One consequential pending call, stated as a question with a named beneficiary.
• Base evidence pinned to a commit or dated source set, so a moving canon cannot drift the pool mid-debate.
• The incumbent position with its receipts — what stands today and why it earned its place.
• A bench drawn from the Research Council: Practitioner, Academic, Economist, Historian, with the Skeptic as mandated adversary. Three seats is the floor and often the ceiling; the full bench is for architectural calls. Add a Dreamer seat as change-advocate only when the call is generative.
• Each seat's gift and shadow written down before the first position — what the lens fights for, and what it will systematically miss.

Steps

1. Seat the bench. Assign roles; the Historian always defends the incumbent from precedent and receipts. Output: the seat list with gift and shadow per seat.
2. Form independent positions. Each seat argues from the base evidence only — no peer positions, no contested history yet. Output: one position with rationale per seat.
3. Open the pool. Inject the contested history (prior variants, prior verdicts, the revealed preferences of past chairs) and all peer positions. Each seat may update; a seat whose lens no longer discriminates is re-assigned or dropped — a role is an updateable lens, never a scripted brief. Output: updated positions with named reasons for every change or hold.
4. Run the Skeptic gate. The Skeptic attacks the leading position — strongest case against, blind spots, what would falsify it. Output: the attack, and each affected seat's response.
5. Synthesize apart. A synthesizer who held no seat scores 3–5 options against named criteria — always including "leave it unchanged" — and records the dissent verbatim. Reopening an already-decided question must name the new evidence that justifies reopening, or the record closes as "no grounds to reopen." Output: the scored slate.
6. Chair the verdict. The human Council rules ADVANCE / REVISE / KILL per the Debate Meeting protocol. The synthesizer scores; it never rules. Output: the chaired verdict, routed to the Decision Journal.

Checks

• Before step 3: every seat's initial position exists in writing — no seat saw another seat first.
• Before step 5: at least one position changed or one seat explicitly held against new evidence with a reason; a debate where nothing moved is a script, not a council.
• Before step 6: the incumbent has a scored option and a named defender; the Skeptic attack is answered or absorbed, not ignored.
• The synthesizer held no seat in the same run. Same mind in both chairs is the self-preference failure this method exists to remove.

Boundary

This method is the Before stage of the Debate Meeting — runnable by agents, or by one person switching seats deliberately. The meeting's During and After (vote, binding verdict, commitment) stay with the human Council unchanged.

It is adversarial by design. de Bono's Six Thinking Hats is parallel thinking, built to replace debate; what this method borrows is only the sequencing idea — every seat completes evidence mode before any seat enters verdict mode — and the difference is named here so the sources stay honest.

Failure Modes

• Judge in the arena — the synthesizer argued a seat; self-preference decides the slate.
• Scripted briefs — seats defend fixed conclusions instead of updating lenses; positions never move.
• Shadowless roles — a seat's blind spot is unnamed, so its miss ships as consensus.
• Consensus theatre — every rationale is "seems fine"; the filter is present in form, absent in function.
• Re-litigation — a decided question reopens with no new evidence; the same call appears in three records in a month.
• Ceremony over stakes — a full bench convened for a call whose cost of error is below the cost of the debate.

Proof Of Done

A run record exists showing: the incumbent defended by a named seat, at least one position update with its reason, a Skeptic finding the synthesis absorbed or explicitly overruled, a scored slate with dissent recorded, and a chaired verdict routed to the Decision Journal. The method's own gauge: curation delta — how far the chair's landed result differs from the recommended slate — trending toward zero across runs.

Changes my mind: a run record where the chaired result matches a single-pass answer over the same evidence while the council found nothing the single pass missed — repeated across comparable calls — proves the separation is ceremony, and this page should be retired or reduced to the Skeptic gate alone.

Retrieval

Pull this method when a consequential framing, naming, or doctrine call is about to ship on one mind's single pass, or when the chair keeps having to curate shipped conclusions into landable form.

Version delta: first authoring — protocol drafted from multi-agent debate research, seated on the existing Research Council and Debate Meeting canon; unproven until its first live run, which is this page's falsifier.

Context

• Purpose — name the beneficiary and the worthwhile outcome before seating the bench.
• Debate Meeting — the During/After machinery this method feeds; verdict vocabulary ADVANCE / REVISE / KILL.
• Archetypes — the Research Council bench and the gift/shadow language every seat uses.
• Truth-Seeking Protocol — the evidence discipline each seat argues under.
• Tight Five Loops — the five-seat bench is a Tight Five: same binding discipline, judgment domain.
• Inversion Thinking — the Skeptic gate's core move, as a standalone method.
• Outsider Thinking — where to find a seat when the bench is too close to the call.
• Decision Journal — where the chaired verdict lands as retained precedent.
• Perspective — the observing skill a seat runs inside its lens.
• Persuasion — the rhetoric standards a rationale must meet to earn its score.

Questions

Next question: which pending call is consequential enough to be this method's first live run — and what curation delta would count as proof?

• When one person runs all the seats deliberately, what stops the strongest voice in their head from writing every position?
• What does a healthy dissent record look like a year in — and what does its absence reveal?
• If the Historian wins every run, is the canon strong or is the bench stale?
• At what stakes does the three-seat floor itself become ceremony?