The Gold Standard of Anthropology | Third-Party Evaluators, Audit Access and Community-Controlled Disclosure

1. Begin with the evaluation purpose and decision

Back to contents

State who commissioned the evaluation, what decision it will inform and which claims it must test. A learning review for partners, a financial audit and a public impact evaluation need different evidence. Without that distinction, “full access” becomes a convenient but unjustified demand.

Translate each question into the minimum evidence needed. If the question is whether community review occurred on time, dated review logs may be sufficient; private testimony about why a reviewer disagreed may not be.

2. Map the evaluator as a new governance actor

Back to contents

Record the evaluator’s legal relationship, funder obligations, professional standards, conflicts of interest, subcontractors, storage systems and onward-reporting rights. Independence from the delivery team does not mean independence from ethics, consent or community governance.

Place the evaluator in the existing authority map. Identify who approves access, who can correct a misunderstanding, who hears a complaint and who decides whether restricted findings may leave a secure setting.

3. Separate audit rights from unrestricted possession

Back to contents

A right to inspect evidence can be implemented through controlled viewing, sampling, redacted extracts, secure rooms or trusted intermediaries. It need not mean copying the entire corpus to the evaluator’s device. Match the access mode to the risk and the proposition being verified.

Write down whether access is view-only, time-limited, purpose-limited or extractable. Prohibit reuse for unrelated research or method development unless separately authorised.

4. Classify evidence before the evaluator requests it

Back to contents

Use practical tiers such as public, project-confidential, evaluator-confidential, community-restricted and prohibited from third-party access. Attach the responsible authority, permitted purpose, retention period and output rule to each tier.

Classification should consider group harm as well as individual identification. A de-identified account of a rare event, office or lineage can still reveal the people involved.

5. Create an evidence-substitution ladder

Back to contents

Start with the least revealing evidence that can credibly answer the question: aggregate metrics, process logs, signed confirmations, redacted samples, controlled inspection and only then direct restricted access. Escalation should require a stated gap, approval and a record of what additional disclosure will resolve.

A substitute is not weaker merely because it protects content. Its adequacy depends on whether it lets the evaluator test the claim and trace the process.

6. Worked example: evaluating a digitised oral-history programme

Back to contents

An external evaluator is asked whether 600 recordings were preserved, described and made accessible under agreed protocols. Two hundred recordings contain restricted kinship and ceremonial material. The initial contract gives the evaluator a copy of every file, although the evaluation does not concern narrative content.

The partners revise the design. The evaluator inspects the preservation manifest, checks a stratified sample inside the community archive, reviews access-permission logs and receives a community-signed statement for restricted deposits. Only cleared excerpts may appear in the report. The evaluator can verify completion and governance without taking possession of the restricted corpus.

7. Protect independence without excluding community review

Back to contents

Community review should not allow inconvenient findings to disappear, but it can correct facts, identify disclosure risks and distinguish evaluator interpretation from community evidence. Define the review stages and the limits of each. The evaluator retains responsibility for conclusions; authorised partners retain control over protected material.

Record disagreements transparently. A final report can describe a contested interpretation and include a response rather than forcing false consensus or publishing sensitive detail.

8. Changed-context transfer check: a safeguarding evaluation

Back to contents

Move from an archive to a youth programme where the evaluator encounters a disclosure of immediate harm. The transferable structure is purpose-limited access, classification and a documented escalation route. The changed context adds safeguarding duties that may require disclosure to named authorities.

Consent and confidentiality materials should explain these limits in advance. The evaluator should disclose only what the safeguarding process requires, record the reason and avoid turning an emergency exception into general access to all participant data.

9. Control evaluator outputs, metadata and quotations

Back to contents

Review tables for small cells, maps for revealing locations, quotations for voice or event recognition and appendices for filenames or hidden metadata. Define whether working papers, recordings and code may leave the secure environment.

Require versions and correction logs. A disclosure repaired in the final report may persist in slides, dashboards or draft annexes unless outputs are traced.

10. Audit the evaluation access chain

Back to contents

Retain the evaluation question matrix, classification register, access approvals, viewing logs, extracted files, redactions, reviewer comments, disclosure decisions and destruction certificates. Test whether every reported finding has an authorised evidence path.

Near misses matter. An emailed spreadsheet, an evaluator’s cloud backup or a copied quotation can reveal weak controls even when the final PDF is safe.

11. Frequently asked questions

Back to contents

Does independence require unrestricted data access? No; it requires freedom to assess evidence credibly within agreed controls. Can communities review negative findings? Yes, for accuracy, interpretation and disclosure risk, without guaranteeing removal. May an evaluator keep data for future work? Only with separate authority. What if the funder asks for raw evidence? Use the contract and governance process to negotiate proportionate verification.

12. The gold standard makes independent scrutiny governable

Back to contents

Credible evaluation needs evidence that can be challenged, traced and checked. It also needs legitimate authority for every disclosure. Define purpose, map the evaluator’s role, classify evidence, use proportionate substitutes and preserve a documented path for disagreement and urgent exceptions.

The gold standard does not weaken accountability to protect secrecy. It designs accountability so that independent scrutiny can occur without making confidentiality and community control the price of being believed.