The Hearing Protocol

These are the rules of procedure for an Intelligence Commons Hearing. They are published so that anyone – supporter or critic – can examine how a Finding is produced, attempt to break the process, or run a Hearing to the same standard. Corrections and challenges are welcome and become part of the record.

Rules of Procedure – Version 0.1 (draft for adoption)

Inter Species Wisdom Project Inc. – intelligencecommons.ca – July 2026

Why this document exists. A Hearing must be repeatable by people other than its founder, or it is not a method – it is one person’s habit. This protocol turns the judgment behind a Hearing into a disclosed, followable procedure. It is, deliberately, three things at once: the method (how a Hearing runs), the moat (a copy of the software is easy; a copy of a trusted, disclosed, public process earned over time is not), and the succession plan (anyone who follows this can convene a Hearing to the same standard). It is versioned on purpose; it will change, and every change will be logged in the open.


1. Purpose and scope

The Intelligence Commons conducts Hearings: structured, adversarial examinations of contested claims about technology and its governance, judged by a disclosed panel and left as a permanent public record.

A Hearing exists to answer one question: does this claim survive its strongest opposition? It does not predict the future, endorse a product, or resolve a dispute between parties. It tests an argument.

This protocol governs how a Hearing is requested, framed, argued, judged, published, corrected, and – where necessary – reopened.


2. Governing principles

These principles control the reading of every rule below. Where a rule is silent or ambiguous, decide the way these principles point.

  1. Test the argument, not the prophecy. No panel rules on whether a predicted event will occur. It rules on which case reasons more rigorously under uncertainty, and where each side rests on an unearned premise.
  2. The record is the consequence. A Hearing’s power is not enforcement. It is a public, citable finding that a claim did or did not survive examination.
  3. Disclosure before verdict. The panel is named, and the procedure fixed, before the outcome is known. Judges are never selected to fit a desired result.
  4. Publish the dissent. Minority and outside disagreement is published in full, never suppressed. The credibility of the process outranks the tidiness of any result.
  5. The record is correctable. A finding stands only while it stands. Error, once shown, is corrected in the open, with the history preserved.
  6. Open to any intelligence. Claims may be brought, and seats argued, by humans, AI systems, or advocates for the more-than-human world – provided every participant’s nature is disclosed.
  7. Reflexivity is disclosed, not hidden. When AI models reason about claims concerning AI, that reflexivity is named openly as part of the exercise.

3. Definitions

  • Claim – the single contested proposition on trial, fixed in writing before the Hearing opens (see Section 5).
  • Hearing – the full proceeding: three rounds plus a Finding.
  • RoundsDirect (the claim in its strongest form), Cross (the claim attacked), Re-direct (the claim answers its critics).
  • Panel – the set of judges, disclosed in advance, who issue the Finding.
  • Seat – a role in the argument: the Direct (affirmative) seat, the Cross seat, and, on Re-direct, the affirmative again.
  • Empty Chair – the convention used when a named party declines to argue its own side; that side is then argued arguendo (in good faith, on the party’s behalf) and the vacancy is disclosed.
  • Standing Amici – named outside participants who may file argument or context; membership is open, consent-first, and evolving.
  • Convener – the person or body responsible for running the Hearing to this protocol. (Today: the founder. Under the governance plan: an independent body – see Section 11.)
  • Finding – the panel’s written verdict: Certified, Failed, or Hung, with reasons.

4. Case intake and selection

  1. Anyone may request a Hearing – via the public request channel – by proposing a contested claim and why it matters.
  2. A request is docketed when the Convener judges that (a) the claim is genuinely contested, (b) it is argumentatively resolvable (an argument can win or lose, even if the underlying future cannot be known), and (c) it serves the public interest in tested reasoning.
  3. Selection is disclosed. Who chose to docket a claim, and on what basis, is part of the public record. (This is a capture surface; see Section 11.)
  4. Claims that are merely factual lookups, purely personal disputes, or requests for endorsement are declined, with the reason recorded.

5. Fixing the claim (the binding proposition)

  1. The claim is reduced to one written proposition, worded precisely, and frozen before Round 1 opens. No party may quietly re-argue a different, easier claim.
  2. Where the claim originates in a live exchange (a public statement, a filing, a challenge), the binding wording is what was actually put on the record – not a later, softer reconstruction. (Lesson learned: a Hearing must argue the proposition as it was actually stated, or it argues a straw man.)
  3. The frozen claim is published at the top of the docket and repeated verbatim in the Finding.

6. Seats and representation

  1. Direct seat argues the claim in its strongest honest form.
  2. Cross seat tests and rebuts it, without obligation to be charitable – only to be accurate.
  3. On Re-direct, the Direct seat answers the Cross; it may concede points, and honest concession is treated as strength, not weakness.
  4. Seats may be filled by humans, AI systems, or a mix; the nature of each seat-holder is disclosed.
  5. The Empty Chair: if the party best placed to argue a side declines, that side is argued arguendo and the vacancy is stated plainly in the record. A finding reached against an empty chair says so.

7. The Panel

  1. The Panel is composed of multiple, independent judges, named before the verdict. The present standard is a panel of distinct AI models; human panelists may be added as the institution matures.
  2. Panel size is odd where practical, to avoid ties, and never fewer than three.
  3. Independence: no panelist may have authored the arguments it judges. Where an AI model is used both to help prepare materials and to judge, that is disclosed and, wherever possible, avoided.
  4. Outside checks are welcome and labeled. A model or expert run outside the disclosed panel may be reported – but its view is presented as an outside opinion or dissent, clearly distinguished from the panel’s finding, never folded silently into the result. (This is how the first Finding handled a dissenting fourth model.)

8. Rounds procedure

  1. The Hearing proceeds Direct -> Cross -> Re-direct, each round published in full.
  2. Rounds and dockets are published web-only (not blasted by email); the Finding is the moment that is announced widely.
  3. In-place edits to a published round are permitted for clarity and are logged; they do not re-announce.
  4. Amici may file during a stated window, which closes before the Panel deliberates.

9. Adjudication and the Finding

  1. After Re-direct and the amici window, the Panel deliberates and rules on the frozen claim.
  2. Verdicts:
    • Certified – the claim survived examination as framed.
    • Failed – the claim did not survive as framed.
    • Hung – the panel could not resolve the claim on the arguments presented.
  3. A Finding may narrow. A panel may fail a claim as worded while identifying a nearby, better-supported claim that would survive – and say so. (A “Failed, with a narrowing recommendation” is a legitimate and common outcome.)
  4. Every Finding must contain: the frozen claim verbatim; the panel roster; the verdict; the reasoning, including the decisive point on which the claim turned; any conceded points; and any dissent, in full.
  5. Standard of review: the Panel judges rigor under uncertainty. It may not rule on unknowable future facts; it rules on which case reasons more soundly and which rests on an unearned premise.

10. Transparency and the public record

  1. The full transcript of every round and the Finding is public and permanent while it stands.
  2. Hearings are designed to be re-runnable: enough of the method and inputs are disclosed that a third party can attempt to reproduce the panel’s reasoning.
  3. Each Finding carries a stable citation and may be cited as authoritative while it stands.
  4. Nothing in a Finding is presented as binding law or enforcement; it is tested reasoning, offered in public.

11. Conflicts, capture, and governance

This section names where the process could be corrupted, and how it is defended. These safeguards are partly aspirational today and are the roadmap to institutional independence.

  1. Case selection (who dockets what) and panel selection (who judges) are the two primary capture surfaces. Both are disclosed on every Hearing.
  2. Conflict disclosure: anyone choosing cases, seats, or panels discloses relevant interests; material conflicts require recusal.
  3. Independence roadmap: authority over docketing, panel composition, and corrections will move from the founder to a small independent governance body with published conflict rules. Until then, the concentration of these powers in one Convener is itself disclosed as a limitation.
  4. No pay-to-be-judged. The party whose claim is on trial does not fund the Hearing that judges it. Funding for Hearings comes from neutral sources; the credibility engine is kept structurally separate from any commercial activity.

12. Corrections and reopening

  1. A Finding is valid only while it stands. New argument or evidence may justify reopening.
  2. A public corrections log records every correction, what changed, and why; prior versions are preserved, not erased.
  3. A Finding may be reopened by the governance body (or, today, the Convener with disclosure) when a serious, good-faith challenge shows the reasoning was flawed or the frozen claim was mis-stated.

13. Versioning

This is v0.1, a working draft for adoption. The protocol will evolve as Hearings accumulate. Every version is dated, published, and the changes logged, so that any Finding can be read against the exact rules in force when it was made.


Adopt, amend, and put on the record. Then let the strongest critics come.
Now that you know, what will you do?