Evidence Validation and Confidence: Development Notes

Document status. This is an intentionally deferred development document. It identifies a methodology UXR expects to need, preserves useful design notes, and defines the questions that future work must answer. It does not yet establish an adopted scoring system, threshold, formula, or operating procedure.

Why this remains open. Premature precision can create false authority and lock UXR into assumptions that have not survived evidence, domain comparison, affected-person review, adversarial testing, or operational experience. UXR will not define this layer further merely to make the documentation look complete.

How this document should be reviewed. Reviewers are invited to identify missing considerations, risks, dependencies, examples, and candidate approaches. The absence of a final mechanism is intentional. A proposal should not be treated as adopted until UXR has published the evidence, tests, and reasoning supporting it.

Purpose. UXR intends to transform lived experience into durable public knowledge. That requires published methods for identifying where each proposition came from, what supports it, what conflicts with it, and what conclusions the evidence can sustain without treating contributor testimony as presumptively inferior.

Testimony is evidence. A contributor's account is part of the evidence record. UXR should record it plainly and accurately. Documentary support can strengthen, corroborate, refine, or contradict a proposition, but the absence of an artifact does not by itself make the contributor's account suspect.

Provenance without credibility discount. UXR should identify how a proposition entered the record without surrounding testimony with language that rhetorically weakens it. “The contributor reports…” is a provenance statement. If a screenshot, forwarded email, transcript, recording, system record, or other artifact supports the proposition, UXR should say so affirmatively. Terms such as “unverified,” “awaiting verification,” or repeated caveats should be reserved for situations where a real evidentiary issue makes that distinction material, such as conflicting evidence, disputed wording, or a conclusion whose precision depends on the artifact. Evidence provenance should add information, not subtract dignity.

Separate judgments. Evidence provenance, identity confidence, claim credibility, factual accuracy, causal inference, severity, and confidence in a final finding are different judgments. Identifying a statement as contributor testimony says where it came from; it does not itself lower its credibility. An authenticated record can prove where a statement came from without proving every implication drawn from it, and testimony can accurately describe an experience even when no artifact exists.

Direct artifact support and pattern corroboration. A directly inspectable artifact may support or verify a message, transaction, interface state, or interaction, subject to authentication and context. Pattern corroboration asks whether sufficiently independent accounts describe materially similar system behavior. These are different evidentiary routes. Artifact support should be described as added support, not as the point at which a contributor's account first becomes worthy of consideration.

Three record layers. UXR should distinguish the contributor's original account, the internal evidentiary working record, and the public case. The original account should be preserved faithfully. The internal working record should preserve provenance, supplements, corrections, superseded details, material conflicts, and the reasoning used to update the case. The public case should present the strongest current evidentiary account rather than reproduce the entire history of how UXR learned it.

Current-best-evidence public record. When later evidence establishes a more accurate wording, date, time, amount, sequence, or other fact, UXR should update the public case to the better-supported fact and remove the superseded detail when the earlier version no longer helps readers understand the case. The public case should read as though the case manager had all current evidence available when writing it from scratch. New evidence may strengthen, weaken, narrow, or otherwise change the analysis; the public case should change with it.

Superseded details and contributor dignity. Ordinary differences between recollection and later artifacts should not become public credibility events. UXR may preserve the superseded detail privately for provenance and auditability without publishing a narrative that calls out the contributor for having remembered wording, timing, or another detail imperfectly. Accuracy should improve without imposing a reputational tax for ordinary human recollection.

Material discrepancies. A discrepancy should become part of the public case when the discrepancy itself materially affects the case, such as when competing accounts remain unresolved, evidence conflicts in a way that changes the finding, the discrepancy bears directly on causation or responsibility, or understanding the conflict is necessary for a fair public conclusion. The purpose is to disclose relevant evidentiary conflict, not to display every refinement in the contributor's account.

Candidate evidence sources. Future methods may evaluate contributor testimony, authenticated interaction records, original emails, screenshots, photographs, recordings, voicemails, notices, receipts, logs, system records, policies, witness accounts, institutional responses, expert analysis, independent replication, and recurring reports from separate people.

Proof should strengthen a report, not become an absolute gate. UXR should request verifiable material whenever it is reasonably available and useful. A person may nevertheless lack records because an organization controls them, a system failed to preserve them, evidence was never created, the person faced danger in collecting it, or the experience itself prevented documentation. The record should state provenance plainly rather than adding suspicion merely because an artifact is absent.

Pattern evidence. Multiple sufficiently independent accounts may provide evidence of a recurring pattern even when no single submission proves every element. Volume alone does not establish truth. Future methods must evaluate independence, similarity, timing, incentives, duplication, shared sources, and alternative explanations.

Identity and anonymity. Confidence that submissions come from distinct human beings can reduce manipulation, but identity verification must not become forced surveillance or exclude whistleblowers, vulnerable people, undocumented people, or those exposed to retaliation. Identity assurance and claim evaluation must remain separate.

Conflicting and missing evidence. Future methods must address contradictory accounts, incomplete records, sampling bias, institutional control of authoritative data, destroyed or inaccessible evidence, power differences in documentation capacity, and the risk of treating polished institutional records as inherently superior to lived experience. When evidence genuinely conflicts, UXR should describe the conflict directly rather than disguising it as generalized suspicion toward the contributor.

Human-readable inference trail. Any future confidence score or public finding should show, in understandable language, which evidence supports which proposition, what genuinely remains uncertain, which competing explanations were considered, and where human judgment entered the process. Explainability requires more than publishing a number or naming a model, but it also requires restraint: uncertainty language should identify actual uncertainty, not function as a ritual disclaimer attached to testimony.

Casework as a methodology stress test. Real casework should test whether UXR can document harm accurately, fairly, and usefully. When a bounded case exposes a weakness in evidence language, correction handling, publication practice, responsibility routing, privacy, or another UXR method, that discovery should be treated as operational learning and, when durable, recorded as a methodological development. Case-driven refinement is evidence of a maturing method; it is not by itself scientific validation of that method.

Candidate technical and procedural tools. Collected notes include chain-of-custody records, source hashes, authenticated share links, integrity tokens, timestamps, redaction, consent controls, confidence categories, transparent inference trails, human review triggers, correction paths, and appeals. None is adopted merely by appearing here.

Dependencies before adoption. This methodology requires intake pilots, privacy and security design, adversarial testing, representative case studies, legal and ethical review, domain expertise, and published evaluations of false-positive and false-negative risks.