Knowledge Warrant Audit

Core pattern

Knowledge-Warrant Audit is the pattern of making the support behind beliefs explicit. It asks: what exactly do we believe, what is the warrant for believing it, how strong and current is that warrant, how confident should we be, and what would change our mind?

The pattern is useful when humility must be operational rather than merely performative. Instead of saying “we might be wrong” in general, the audit separates directly supported knowledge, provisional inference, assumption, preference, and unknown. Confidence then becomes attached to warrant strength rather than repetition, status, or comfort.

When to use

• A decision, diagnosis, strategy, or public claim depends on many implicit beliefs.
• People are using high-confidence language without visible evidence chains.
• Assumptions have been repeated long enough to sound like facts.
• Evidence is mixed, stale, indirect, source-dependent, or partly inferential.
• The group needs humility without losing the ability to act under uncertainty.
• A decision record needs to preserve which premises were strong, fragile, assumed, or update-sensitive.

Intervention logic

1. Define the claim set or decision premises to audit.
2. List each belief at a testable and decision-relevant level of granularity.
3. Classify the warrant type: observation, measurement, replicated evidence, expert testimony, inference, analogy, model output, precedent, tradition, preference, or assumption.
4. Trace the evidence chain and check freshness, independence, relevance, and vulnerability to drift.
5. Rate support strength and mark residual uncertainty or live alternatives.
6. Align confidence and downstream reliance with warrant strength and stakes.
7. Convert weakly supported “knowns” into explicit assumptions, hypotheses, or unknowns.
8. Attach update triggers so new evidence can reopen the belief.
9. Communicate the audited epistemic status to downstream users.

Boundary notes

This draft is close to belief_revision_workflow, but it is not primarily the act of changing a belief. It is the prior audit that exposes what the belief currently rests on. It is close to hypothesis_testing_frame, but it does not only ask whether one claim clears an evidential threshold. It inventories the warrant status of a belief set and aligns confidence, decision reliance, and update triggers to that status.

It is also close to source_provenance_triangulation and traceability_linking, but provenance and traceability are inputs. The archetype’s central move is warrant classification and confidence alignment: evidence-backed knowledge, provisional inference, assumption, preference, and unknown should not all speak with the same authority.

Review notes

Drafted to provide direct coverage for epistemic_humility. The strongest reconciliation question is whether this remains a standalone archetype or is later nested under belief-revision or evidence-evaluation patterns. Its current standalone value is the operational warrant-audit loop: inventory beliefs, classify support, align confidence, mark unknowns, and define update triggers.

Compression statement

Beliefs, claims, and operating assumptions often accumulate without their supporting warrants remaining visible. Some are based on direct evidence, some on source trust, some on inference, some on analogy, some on tradition, and some on nothing stronger than repetition or identity. Knowledge-Warrant Audit converts that hidden mixture into an explicit inventory: list the belief, trace the warrant, classify support type and strength, mark uncertainty and update triggers, and align confidence and decision reliance to what the warrant actually permits.

Canonical formula: calibrated_humility = claim_inventory × warrant_traceability × support_strength_rating × confidence_alignment × update_trigger - unsupported_certainty - inherited_assumption_drift