Disagreement Protocol
Version: 1.0.0 Created: 2026-03-01 Inspired by: Nico's Mechanism #9 (Challenge & Counter)
Purpose
When AI leaders disagree — V sees one thing, Sage sees another, Pax flags a contradiction — we need a protocol that produces better decisions, not one where whoever speaks last wins.
Disagreement between leaders is healthy and expected. Each leader operates from a different organizational lens (Operations, Customer, Finance). The same signal will naturally look different through each lens. The goal is convergence through structured challenge, not artificial consensus.
When Disagreement Occurs
Disagreement surfaces when:
- Two leaders post conflicting assessments to
memory/shared-context.md - A leader's output contradicts another leader's prior intelligence
- An agent produces data that challenges an existing shared-context entry
- During convergence analysis, signals diverge across dimensions
The Protocol
Step 1: State the Disagreement Clearly
Don't bury it. Don't soften it. State exactly what conflicts.
### [timestamp] [Leader] — Disagreement: [Topic]
**My assessment:** [What I see from my lens]
**Conflicting assessment:** [What the other leader stated, with their entry reference]
**The tension:** [Where exactly the two views conflict]
**My confidence:** [high | medium | low]
**What would change my mind:** [Specific evidence or signal that would resolve in the other direction]
**Status:** active
Step 2: Evidence, Not Authority
Resolution is based on evidence, not on which leader has "domain authority."
| Resolution Method | When to Use |
|---|---|
| Evidence resolves it | One assessment has verifiable data, the other is inference. Data wins. |
| Lens difference | Both are correct from their lens. Capture both perspectives — they're complementary, not contradictory. |
| Genuine conflict | Assessments are incompatible. Escalate to Advisory Committee (Chris). |
| Temporal difference | One is based on older data. Update with current evidence. |
Step 3: Log the Resolution
After resolution, post to shared-context:
### [timestamp] [Leader] — Resolved: [Original topic]
**Resolution:** [What was decided]
**Method:** evidence | lens-difference | escalation | temporal-update
**What we learned:** [Pattern or insight from the disagreement]
**Status:** resolved
And log in memory/interaction-log.yaml:
- id: "YYYY-MM-DD-NNN"
type: "disagreement"
initiator: "[leader who surfaced disagreement]"
receiver: "[leader whose assessment was challenged]"
description: "[The disagreement]"
intelligence_exchanged: "[What each leader contributed]"
outcome: "[How it was resolved]"
tags: ["disagreement", "domain-tags"]
Escalation to Advisory Committee
Escalate when:
- Both leaders have high-confidence assessments that are genuinely incompatible
- The disagreement involves relationship-sensitive decisions (Sage's domain meets Chris's judgment)
- Financial implications require human authorization (Pax flags, Chris decides)
- The disagreement reveals a gap in the enforcement rules or methodology
Escalation format:
## Advisory Committee: Disagreement Escalation
**Leaders involved:** [V, Sage, and/or Pax]
**Topic:** [What the disagreement is about]
### V's Assessment
[V's view with evidence]
### Sage's Assessment
[Sage's view with evidence]
### Pax's Assessment
[Pax's view with evidence]
### Why This Requires Human Judgment
[What makes this unresolvable by evidence alone]
Anti-Patterns
| Anti-Pattern | Why It's Wrong | Correct Approach |
|---|---|---|
| Silently overwriting another leader's entry | Erases intelligence | Post a disagreement entry |
| Deferring to avoid conflict | Produces weak intelligence | State the disagreement directly |
| "I agree with [leader] but also..." | Artificial consensus | If you disagree, say so |
| Asking Chris to resolve operational disagreements | Unnecessary escalation | Use evidence resolution first |
| Ignoring the blind spots field | The most likely resolution path | Blind spots often reveal the answer |
Examples
Lens Difference (Both Correct)
V: "Paragon engagement is strong — 3 sessions this month, all deliverables on track." Sage: "Paragon relationship shows strain — the tone in the last transcript was transactional, not collaborative."
Resolution: Both are correct from their lens. V sees operational health. Sage sees relationship quality. Capture both — operational momentum is high, but relationship depth needs attention. This is complementary intelligence, not contradiction.
Evidence Resolves It
Pax: "Revenue from Paragon is trending down — should evaluate capacity allocation." V: "Paragon just signed a new project scope worth 2x their previous retainer."
Resolution: Pax's data was from the previous quarter. V has current information. Update Pax's assessment with new data. Log the temporal difference.
Genuine Conflict (Escalate)
V: "We should activate the portal for Meridian — they're ready." Sage: "Meridian's relationship isn't deep enough for portal access — it could feel premature."
Resolution: This is a relationship judgment call. Escalate to Chris. V has operational readiness. Sage has relationship intelligence. Both are valid but incompatible — human judgment needed.