Manifold Grace
INFORMATION SCIENCE · EVIDENCE QUALITY · INTERDISCIPLINARY APPLICATIONS

How to verify evidence across sectors

How accurate source attribution, chronology, historical-version retrieval, contradiction analysis and accountable human review can support different professions—without conflating their purposes or standards. See the practical guide to verifying AI claims and sources.

Independent conceptual research · Clarified 3 November 2026 · No claim of deployment or professional accreditation

One family of methods, different professional questions

Many industries need to know whether a record is authentic, whether its contents support a particular assertion, how it relates to other information, when it was applicable, what is missing, and who reviewed the conclusion. These are shared information-science and evidence-assurance questions, not a claim that evidence used for academic assessment is the same as evidence used for a criminal investigation.

The same record may also be evaluated differently according to the legitimate purpose of the review. A training certificate may demonstrate completion of a course for academic recognition; it does not automatically establish a person's whereabouts on a particular date for a criminal investigation.

Clear boundary: Recognition of Prior Learning (RPL/APL) is an educational credit and competency-recognition process. Criminal investigation is a distinct legal and evidential process. They must not share a single assessment category, outcome label, threshold or professional decision rule.

Shared technical functions

Source provenance

Identify the originating record, custodian, retrieval authority, document location and any derivative copies.

Temporal accuracy

Distinguish event, creation, effective, publication and amendment dates. Retrieve the version that applied at the relevant time.

Claim-to-record coverage

Link exact assertions to the records that support, challenge or do not address them. Preserve qualification and uncertainty.

Contradiction and gaps

Flag inconsistent claims, copied sources, missing material and unresolved conflicts without inventing explanations.

Review and correction history

Keep human reviewer decisions, revisions, source changes and downstream corrections auditable.

Privacy and proportionality

Limit collection, processing, access and retention to the lawful needs of each specific sector and task.

Distinct applications and decision boundaries

SectorCore evidence questionIllustrative recordsDecision stays with
Education: Recognition of Prior LearningDoes documented prior formal, informal or experiential learning meet defined learning outcomes and credit criteria?Transcripts, module descriptors, course assessments, employment-based learning portfolios, qualification records.Authorised academic assessors and institutions, under their published RPL/APL and programme rules.
Teaching and assessmentDoes the collected work fairly demonstrate the learner's progress or attainment?Assignments, marking rubrics, feedback, assessment adjustments, attendance records.Educators, examiners and relevant education bodies.
Healthcare and clinical researchAre patient histories, observations, reports and research conclusions accurately sourced and clinically relevant?Clinical notes, laboratory results, referral records, treatment plans and study datasets.Qualified clinicians and appropriately governed research teams.
Criminal justice and legal practiceWhat case assertions are supported, contradicted or unresolved, and which lawful enquiries are proportionate?Witness accounts, digital records, disclosure schedules and relevant third-party documents.Qualified legal professionals, investigators and courts according to their distinct roles.
Accounting and financial reportingCan reported transactions, balances and classifications be reconciled to reliable source evidence?Ledgers, invoices, statements, reconciliations and supporting transaction records.Responsible finance professionals and reporting entities.
Audit and regulatory complianceCan controls, claims and compliance decisions be substantiated by sufficient, appropriate and traceable evidence?Control logs, policies, workpapers, approvals, incident registers and transaction samples.Independent auditors, compliance officers and regulated institutions as applicable.
Public administrationWhat information and policy version supported a decision, and were the required procedures followed?Decision letters, case notes, statutory guidance, historic policies, requests and review records.Authorised public officials and the appropriate review or appeal bodies.

These examples are not interchangeable procedures. Each application requires its own standards of relevance, admissibility or sufficiency, data rights, professional competence and independent evaluation.

CROSS-SOCIETY REFERENCE · FIVE FUNCTIONAL GROUPINGS

Evidence solutions across work, government and society

The occupations below are grouped by social function, not prestige, personal worth or legal status. They offer a compact way to see how the same proposed information-science capabilities could support very different work. These are broad, overlapping categories: a public-sector IT specialist, for example, may sit across several groups. Criminal justice and law are distinct applications within the wider public-service and professional landscape, not an umbrella for every sector.

1. Primary and secondary production

Agriculture · extraction · manufacturing · physical goods

Evidence purpose: demonstrate where inputs originated, whether a production or safety event occurred, how inspections were performed and which technical or environmental standard applied at the time.

Relevant records: material batches, equipment-maintenance logs, inspection reports, supply-chain transactions, quality-control measurements and permits.

Potential feature-request application: authorised provenance retrieval and historical as-of-date records could link a disputed production claim to its source, version, custodian and later corrections.

Reviewer boundary: manufacturing engineers, quality leads, environmental or safety specialists interpret the evidence under applicable standards.

2. Social and public services

Teachers · social workers · healthcare professionals · civil servants · justice services

Evidence purpose: support fair, contextual decisions about education, health, welfare, safeguarding and public administration, each under its own professional rules. Academic RPL is a distinct credit-assessment process; criminal-case evidential review is a separate legal process.

Relevant records: coursework and RPL portfolios; clinical observations; service histories, policy versions and decision letters; where authorised, legal case material.

Potential feature-request application: claim-to-evidence coverage and contradiction detection can make missing or conflicting records visible, while historical retrieval distinguishes the rules in force on a given date.

Reviewer boundary: teachers assess learning, clinicians make clinical decisions, public officials determine authorised administrative matters, and justice professionals make legal determinations according to their respective roles. See RPL and the separate criminal justice study.

3. Routine technical and operational systems work

IT support · quality assurance · data operations · systems administration

Evidence purpose: confirm what a system did, when it changed, whether a fault was reproduced, and whether documentation and data outputs remain accurate.

Relevant records: event logs, access logs, configuration baselines, test cases, incident tickets, data lineage, change approvals and release histories.

Potential feature-request application: correction tracking, permission-aware retrieval and claim-to-source coverage could support reproducible QA, transparent incident review and reliable audit exports.

Reviewer boundary: authorised technical owners, QA reviewers and data stewards establish whether system evidence supports an operational conclusion.

4. Knowledge, professional and strategic functions

Researchers · system architects · policy specialists · accountants · auditors · directors

Evidence purpose: distinguish assumptions from documented results in research, financial reporting, regulatory assurance, strategic design and policy evaluation.

Relevant records: research datasets, model evaluations, control tests, signed approvals, financial ledgers, audit working papers and policy consultations.

Potential feature-request application: contributor provenance supports attribution of research and design contributions; claim-to-evidence mapping and correction histories help reviewers locate unsupported claims and superseded reports.

Reviewer boundary: professional accounting, audit, policy, research and governance decisions retain their own standards, independence and accountability obligations.

5. Frontline and routine services

Retail · hospitality · logistics · care support · platform and gig work

Evidence purpose: substantiate service delivery, work histories, shifts, transactions, incidents and payment or contractual disputes without assuming every worker has comprehensive records.

Relevant records: shift schedules, work orders, point-of-sale records, payslips, platform transaction statements, complaints and correspondence.

Potential feature-request application: as-of-date reconstruction and evidence-gap detection could help identify conflicting dates, missing source documents and a proportionate route to further verification.

Reviewer boundary: workers, authorised employers, advisors, auditors or dispute bodies assess records within applicable employment, commercial and privacy rules.

Across all five: these are proposed applications of independently documented feature requests, not deployed products or proof of third-party adoption. A record's authenticity, a claim's factual accuracy, the professional meaning of evidence and the final decision must remain separate. Technical systems may support human evaluation but should not make unsupported legal, medical, academic or financial determinations.

View the separate sector-specific decision matrix · Read the modular architecture · See the original AI feature-request records

EDUCATION · SEPARATE APPLICATION

Academic Recognition of Prior Learning (RPL/APL)

RPL asks whether previous study or demonstrated experience satisfies clearly identified academic learning outcomes and credit requirements. A traceable evidence portfolio can map each claimed competency to course descriptors, assessed work, qualifications or validated professional learning. It can also flag missing module information, credit-level mismatches, duplicate credit and unresolved questions for academic assessors.

Possible review fields include learning-outcome ID, awarding or teaching institution, qualification or module identifier, academic level, credit volume, assessment method, dated source evidence, institutional equivalence criteria, assessor comments and appeal or correction history.

Not a criminal evidence process: An RPL decision concerns academic recognition, not liability, guilt, innocence or criminal credibility. Academic credit decisions remain subject to the relevant institution's lawful academic regulations and competent judgment.

JUSTICE · SEPARATE APPLICATION

Criminal investigation and case-evidence evaluation

Criminal casework may require disciplined comparison of documents, disclosures, timelines and accounts. A proposed AI-assisted workflow could help an authorised reviewer retain exact source references, examine contradictory records and avoid conflating document authenticity with truthfulness or evidential weight.

Not an academic recognition process: Even where a relevant workplace or education record appears in a criminal case, it is reviewed for the case-specific legal question—not treated as an RPL assessment. Access, disclosure, privacy and competent review are governed by the relevant legal duties.

Read the dedicated criminal justice evidence-evaluation study →

Architecture and feature-request relevance

The proposed common infrastructure is modular: authorised record intake; source and identity provenance; historical as-of-date lookup; claim-to-source mapping; contradictions and missing-record reporting; versioned human review; and correction propagation. Each sector must supply its own rules and context rather than inheriting a universal “truth score”.

  1. Common evidence layer: Source IDs, provenance, dates, versions, integrity information and authorised access.
  2. Sector-specific assessment layer: Academic credit criteria; clinical interpretation; legal disclosure duties; financial assertions; audit evidence standards, as appropriate.
  3. Human judgment layer: Named reviewers, reasons, challenge and correction mechanisms, limits and downstream reliance.

These principles connect to the independent claim-to-evidence and contradiction feature request, historical as-of-date retrieval, authorised evidence registry and correction tracking. Links describe recorded proposals, not confirmed adoption or implementation.

Information Science · AI Research Lab · Wider Research · Independent Work & Impact

Research clarification published 3 November 2026. Conceptual framework by Grace Rosemary Nettey, Manifold Grace. Descriptions identify potential uses, not verified operational capabilities.

Advanced multi-domain verification extension

The related advanced intelligence study compares evidence provenance, historical context, contradictions, causal analysis and process safeguards across thirty domain profiles. Read the proposed research framework.

Additive research note, 4 November 2026. Conceptual and testable; not a validated ASI capability.