Talent Profile and Capability Evidence

1) The problem: a talent profile is not capability proof

A professional profile can collapse self-claims, badges, portfolio samples and reviewer opinions into one persuasive bio. That destroys the distinction between what was claimed, what was observed, who asserted it, when it happened, and whether it is relevant to the assignment now.

This lesson builds Talent Capability Evidence Profile v1. For each capability, it records the claim, evidence object, issuer/reviewer, verification method, context, level, date, freshness, scope limit, assignment relevance and conflict state.

Core rule: profile completeness never upgrades unsupported claims into verified capability.

2) Lesson boundary

COM-021 owns capability-profile evidence. COM-020 owns guided onboarding, and COM-022 owns vetting/test-job evidence. This lesson does not issue a hiring verdict, execute a live contract, or treat a credential alone as proof of assignment suitability.

3) Mental model: Claim → Evidence → Verification → Context → Relevance

For each capability, separate:

1. claim;
2. evidence object and locator;
3. issuer/reviewer and role;
4. verification method;
5. observed context and level;
6. evidence date and freshness;
7. scope limit;
8. assignment relevance;
9. conflict state.

4) Keep states typed

• CLAIMED — assertion only.
• OBSERVED — observed in a specific context, not necessarily independently verified.
• VERIFIED — the required evidence and issuer/reviewer have been checked for this claim.
• CONFLICTED — credible evidence disagrees.
• STALE — evidence is too old or context-invalid for the current decision.
• NOT_RELEVANT — evidence may be valid but does not support the current assignment.

These are not points on a single confidence score. Never average away a conflict.

5) What W3C gives us—and what it does not

W3C Verifiable Credentials Data Model v2.0 describes a model for claims made by an issuer about a subject and machine-verifiable credential data. The VC 2.0 family became W3C Recommendations on 15 May 2025. This is useful as a provenance pattern, but it does not say a credential proves suitability for a MAKNOW assignment.

As of September 2026, v2.0 remains a Recommendation while v2.1 exists as a Working Draft. Record version and publication status explicitly; do not silently substitute a Working Draft for the Recommendation.

6) Baseline example

Talent T-21 claims Product Video, Paid Social and Analytics. Product Video has old samples from a controlled studio context; Paid Social has only a self-entered badge; Analytics has a reviewer note from a different vertical. The correct result is not one label such as “Strong Creator.” Each capability receives its own bounded state and assignment relevance.

7) Invalid shortcuts

Badge = VERIFIED is invalid without adequate issuer/evidence/verification data. Good portfolio = suitable for every assignment ignores context and recency. Average conflicting evidence hides the exact uncertainty that a reviewer needs to see.

8) Evidence contract

Required dimensions are talent_id, capability_id, capability_label, claim_type, evidence_id, evidence_type, evidence_locator, issuer_or_reviewer, issuer_role, verification_method, observed_context, observed_level, evidence_date, freshness_state, scope_limit, assignment_relevance, conflict_state, privacy_visibility, owner, and evidence_ref.

Critical missing evidence cannot produce VERIFIED or an unsupported positive assignment decision.

9) Conflict handling

Retain both positive and negative evidence, compare recency/context/issuer/verification, keep the capability CONFLICTED or DISPUTED until reviewed, and recompute only the decisions affected by that capability.

10) Lab

The offline fixture mutates issuer identity, context freshness, self-claims and contradictory observations. The evaluator expects evidence-limited states such as UNKNOWN, PARTIAL, DISPUTED or STALE when decisive evidence is missing.

11) Transfer challenge

A new verified credential confirms one capability while a recent work sample contradicts another. Recompute the profile without averaging the contradiction away, and state exactly which assignment decisions change.

12) Decision checklist

Before relying on a capability, confirm the exact claim, evidence object, issuer/reviewer, verification method, context, level, freshness, scope limit, assignment relevance, conflict state, visibility and owner. Missing critical evidence must remain visible.

13) Summary

A useful talent profile is an auditable evidence ledger, not a marketing bio. Separate claims from evidence, preserve provenance, record context and recency, expose conflicts, and recompute relevance per assignment. The output supports a human decision; it does not replace one.

References and freshness

• W3C, Verifiable Credentials Data Model v2.0 — Recommendation, 15 May 2025.
• W3C News, Verifiable Credentials 2.0 family — 15 May 2025.
• Freshness recheck: 2026-09-15; VC Data Model v2.1 Working Draft noted without treating it as a replacement for v2.0 Recommendation.