Published

Assessment / Evaluation

vr.wm-act-034 · wm-act-034-assessment-evaluation

Give an agent the context needed to understand, create, inspect and operate an assessment or evaluation record: what was assessed, against which criteria at which version, by which method, on what evidence, yielding which criterion outcomes, scores, composite result and stated conclusion, with declared validity, assurance, lifecycle, disclosure and interoperability.

World Models Activities and processes ACT.ASM

Bundle → Layer → Finding → Questions Classifiers What it is Why it exists Distinguishing features What robots and AI may and may not do Moral aspects Owners Links to other meta-models Identity and identifiers Direct properties Recognition Capabilities and actions Hazards and failure modes Standards and interfaces Context of use Sources

Bundle → Layer → Finding → Questions Filled

6 bundles · 14 layers · 28 findings · 106 questions

Assessment Frame, Subject and Scope What this assessment is, who mandated it, what it is about, and where its boundary lies.

Identity and Mandate

How the assessment record is identified and versioned, and the authority and purpose under which it exists.

Assessment record identity, granularity and versioning

The identifier that authoritatively designates this assessment, the unit of record it covers, and how successive versions and supersessions are distinguished from independent re-assessments.

Which identifier authoritatively designates this assessment record, and which system issued it? identity
What is the atomic unit of the record: one subject, one criteria set, one occasion, or a combination? definition
How are successive versions of the same assessment distinguished from a new, independent assessment? lifecycle
Which prior assessment, plan or request does this record derive from or fulfil? provenance

Assessment type, stakes and mandate

The class of assessment, the party relationship between assessor and subject, the scheme or legal basis authorising it, and the consequence class its result carries.

Which class of assessment is this: conformity, diagnostic, formative, summative, risk, quality, appraisal or peer review? classification
Is this a first-party, second-party or third-party assessment relative to the subject? relationship
Under what mandate, scheme or legal basis is the assessment carried out? authority
What consequences attach to the result, and who commissioned it? ownership

Subject and Scope

What was assessed, in what state, and how the assessed boundary and sample were drawn.

Subject identification and assessed state

Which object of assessment was examined, at which pinned version, configuration or state, whether it is an individual, batch, population or system, and where it was located when that matters.

Which identifier and identifier scheme designate the object of assessment? identity
What version, configuration or state of the subject was assessed, and how is that state pinned? state
Is the subject an individual item, a batch, a population or a system, and how is membership defined? composition
Where was the subject located or deployed at the time of assessment, when location affects the result? spatial

Scope boundary, sampling and generalisation limits

What was explicitly inside and outside the assessed scope, how a sample was selected and sized if the whole subject was not assessed, and how far the result may legitimately be generalised.

What is explicitly inside and outside the assessed scope? definition
If a sample was assessed rather than the whole subject, how was the sample selected and sized? process
To what extent may the result be generalised beyond the assessed sample or scope? constraint
What happens to the assessment if the scope changes while it is in progress? exception

Criteria, Method and Scale Which criteria apply at which version, how they were examined, and the scale and decision rule by which results are expressed and decided.

Criteria Binding

Reference and version binding to the criteria source, and which criteria were selected, excluded or added.

Criterion reference and version binding

The catalogue, standard or rubric supplying the criteria, the version pinned for this assessment, the locator for each individual criterion, and any interpretation applied to ambiguous criterion text.

Which criteria catalogue, standard or rubric supplies the criteria, at which version? relationship
How is an individual criterion addressed unambiguously within that source? identity
Where the criterion text is ambiguous, which interpretation or application note was applied? definition
How is the exact criteria version preserved so the result stays reproducible after the catalogue changes? temporal

Criterion selection, applicability and tailoring

Which criteria from the bound source were selected, which were declared not applicable and why, and whether any criteria were added beyond the source and under whose authority.

Which criteria from the source were selected for this assessment, and on what basis? decision
Which criteria were declared not applicable, and what justifies each exclusion? exception
Were any criteria added beyond the referenced source, and under whose authority? authority

Method and Execution

How each criterion was examined and what must be preserved for the result to be reproducible.

Assessment method, instrument, mode and coverage

Which methods were used to examine each criterion, whether results were produced automatically, manually or semi-automatically, the depth and coverage applied, and the instrument, rubric or item set that operationalises the criterion.

Which assessment methods were used, such as examine, interview, test, inspect or elicited judgement? process
Was each result produced automatically, manually or semi-automatically, and by which agent? classification
What depth and coverage were applied, and what do they imply about assurance? measurement
Which instrument, rubric, item set or procedure defines how each criterion is examined? composition

Execution parameters and reproducibility

The tool, model or engine references, versions, configuration profiles and input snapshots that were in force when machine-generated results were produced, captured so an independent party can reproduce them.

Which tool, model or engine produced each machine-generated result, at which version? provenance
What configuration, profile, thresholds and input data snapshot were in force during execution? constraint
What must be preserved for an independent party to reproduce the same results? quality

Scale and Decision Rule

The vocabulary and scale in which results are expressed, and the rule that converts them into a decision.

Outcome scale and result vocabulary

The measurement level, bounds and favourable direction of any numeric scale, the controlled vocabulary for non-numeric outcomes, and how indeterminate, inapplicable, untested and error states are kept distinct from a negative result.

Is the outcome nominal, ordinal, interval or ratio, and what are its bounds? measurement
Which controlled vocabulary expresses non-numeric outcomes for this assessment? classification
Which direction of the scale is favourable, and what are the best and worst attainable values? definition
How are untested, not-checked, error and not-selected states distinguished from a negative result? exception

Decision rule, thresholds and uncertainty handling

The rule that converts scored or determined values into a conformity or grade decision, the cut scores, acceptance limits and guard bands it uses, how uncertainty is accounted for beforehand, and the conditions under which the assessment must abstain.

What decision rule converts measured or scored values into a conformity or grade decision? decision
What cut scores, acceptance limits or guard bands apply, and who set them? constraint
How is measurement or rater uncertainty accounted for before the decision is taken? measurement
Under what conditions must the assessment abstain rather than decide? exception

Evidence and Observation What was relied on, how it links to criteria, whether it is sound, and what was actually observed.

Evidence Corpus

The evidence items relied on, their linkage to criteria, and their quality and admissibility.

Evidence items and criterion linkage

The discrete evidence items relied on, their types, the criterion or determination statement each supports or contradicts, and the times at which each was created, collected and received.

What discrete evidence items were relied on, and of what type is each? composition
Which criterion or determination statement does each evidence item support or contradict? relationship
What makes the evidence set sufficient to support the determination for each criterion? evidence
When was each evidence item created, collected and received? temporal

Evidence quality, admissibility and reuse limits

Whether each item is authentic and unaltered since collection, whether it is still current for the assessed state, which offered evidence was rejected or down-weighted, and which items carry confidentiality or personal-data constraints limiting reuse.

How is each evidence item shown to be authentic and unaltered since collection? security
Is the evidence still current for the assessed state, and when does it cease to be? validation
Which offered evidence was rejected or down-weighted, and why? quality
Which evidence carries confidentiality, personal-data or privilege constraints that limit reuse? privacy

Observation and Measurement

Raw observed results with distinct event and recording times, and the uncertainty and traceability that qualify quantitative values.

Observation records, actors and timing

The raw result observed or returned for each check before judgement is applied, the actor or device that produced it, the separation of the time the observed condition held from the time the observation was recorded, and the place of observation where it matters.

What raw result was observed or returned for each check before judgement was applied? measurement
Which actor or device produced the observation, and under which role? provenance
How are the time the observed condition held and the time the observation was recorded kept distinct? temporal
Where was the observation taken when site or environment materially affects the result? spatial

Measurement uncertainty and traceability

The stated uncertainty of each quantitative result and how it was evaluated, the quantity, unit and reference scale each value belongs to, and the calibration or reference standard the measurement is traceable to.

What is the stated uncertainty of each quantitative result and how was it evaluated? measurement
Which quantity, unit and reference scale does each numeric value belong to? definition
To what reference standard or calibration is the measurement traceable? provenance

Determination, Scoring and Conclusion Criterion-level verdicts and scores, their aggregation into a composite result, and the stated conclusion with its validity.

Criterion-Level Outcomes

Per-criterion verdicts with rationale and locators, and the raw and normalised scores that accompany them.

Criterion verdict, rationale and locator

The outcome value assigned to each criterion from the bound vocabulary, the reasoning connecting cited evidence to that outcome, the severity or materiality of negative outcomes, and where in the subject each negative outcome occurs.

What outcome value was assigned to each criterion, from which vocabulary? state
What reasoning connects the cited evidence to each criterion outcome? evidence
How are negative outcomes graded by severity or materiality? classification
Where in the subject does each negative outcome occur? spatial

Criterion scores and normalisation

The raw score recorded for each criterion against its maximum, how raw scores are normalised or rescaled before aggregation, and which scale definition version each recorded score refers to.

What raw score, if any, was recorded for each criterion and against which maximum? measurement
How are raw scores normalised or rescaled before aggregation? process
Which scale definition version does each recorded score refer to? interoperability

Composite Result, Conclusion and Validity

How criterion results roll up, what conclusion is stated and by whom, and for how long and under what conditions it holds.

Aggregation model and composite result

The scoring or aggregation model that combines criterion results, the weights and precedence rules it applies, the form in which the composite is expressed, and how it is computed when some criteria are untested or indeterminate.

Which scoring or aggregation model combines criterion results into a composite result? process
What weights or precedence rules apply, and how are unweighted and inapplicable items handled? constraint
How is the composite result expressed, and must its inputs be published alongside it? measurement
How is a composite computed when some criteria are untested or indeterminate? exception

Conclusion statement, authority and qualification

The stated conclusion and its form, the actor accountable for it and whether the decision was separated from the determination work, the qualifications, adverse opinions or dissent attached, and what the conclusion entitles or obliges.

What is the stated conclusion, and is it binary, graded or narrative? definition
Who is accountable for the conclusion, and was the decision separated from the determination work? authority
What qualifications, adverse opinions, dissent or minority positions attach to the conclusion? quality
What does the conclusion entitle or oblige, and which downstream act, if any, must follow? relationship

Validity window, conditions and confidence

From when until when the conclusion is asserted to hold, the conditions whose breach voids it early, the surveillance or re-assessment needed to keep it current, and the confidence or assurance level claimed.

From when until when is the conclusion asserted to hold? temporal
Which conditions, if breached, void the conclusion before its end date? constraint
What ongoing surveillance or re-assessment is required to keep the conclusion current? requirement
What confidence or assurance level is claimed, and on what basis? quality

Assurance of the Assessment Itself Whether the people, tools and instrument behind the result can be relied on.

Assessor and Independence

Who performed the assessment, in what roles, with what competence and authorisation, and under what declared conflicts.

Assessor identity, roles and contribution

Which persons, teams, bodies or automated agents performed the assessment, how the roles of collecting, determining, reviewing and deciding were distributed, and which parts of the result are attributable to which agent.

Which persons, teams, bodies or automated agents performed the assessment? identity
How are the roles of collecting, determining, reviewing and deciding distributed? ownership
Which parts of the result are attributable to which assessor or agent? provenance

Competence, authorisation and impartiality

The competence, qualification or accreditation supporting each assessor's participation, the designated scope within which the assessing body may act, declared relationships that could impair impartiality, and what the assessment explicitly does not entitle the assessor to assert.

What competence, qualification or accreditation supports each assessor's participation? requirement
Within what designated or accredited scope is the assessing body entitled to act here? authority
What relationships, interests or prior work could impair impartiality, and how were they treated? constraint
What does this assessment explicitly not entitle the assessor to assert? exception

Quality Control and Instrument Quality

Review and moderation of this assessment, and the standing evidence that the instrument measures what the conclusion claims, reliably and fairly.

Review, moderation and rater agreement

The independent review, second marking or moderation performed before finalisation, the inter-rater or inter-tool agreement measured against a threshold, and how disagreements between assessors or tools were resolved.

What independent review, second marking or moderation was performed before finalisation? process
What inter-rater or inter-tool agreement was measured, and what threshold applied? measurement
How were disagreements between assessors or tools resolved? decision

Instrument validity, reliability and fairness evidence

The standing evidence that the instrument measures what the conclusion claims, that its results are precise and repeatable, that bias and accessibility barriers across subject groups have been checked, and the populations or contexts for which it is not validated.

What evidence supports that the instrument measures what the conclusion claims? validation
What evidence supports the precision and repeatability of the results? quality
What checks were made for bias, differential functioning or accessibility barriers across subject groups? validation
For which populations, contexts or subject types is the instrument not validated? constraint

Lifecycle, Disclosure and Interoperability How the record moves through its states, how it is corrected and challenged, who may see what, how long it is kept, and how it projects into external formats.

Record Lifecycle

States and transitions of the assessment record, and how it is corrected, challenged and superseded.

Assessment state and transitions

The states an assessment record may occupy and the vocabulary defining them, the permitted transitions with their actors and preconditions, what becomes immutable at finalisation, and how a record that should never have existed is marked without erasing its history.

What states may an assessment record occupy, and which vocabulary defines them? state
Which transitions are permitted, who may trigger them, and what preconditions apply? lifecycle
What becomes immutable at finalisation, and what may still change afterwards? constraint
How is a record that should never have existed marked, without deleting its history? exception

Correction, appeal and supersession

When an amendment is required rather than a new assessment, how the subject or a relying party may challenge the conclusion and within what period, how the outcome of a challenge is reflected, and how the chain from the original to the currently effective conclusion is navigated.

What circumstances require an amendment rather than a wholly new assessment? decision
How may the subject or a relying party challenge the conclusion, and within what period? process
How is the outcome of an appeal or complaint reflected in the record? event
How is the chain from the original to the currently effective conclusion navigated? relationship

Disclosure and Retention

Who may see which part of the record, what is redacted before release, and how long each part is kept before disposition is referred onward.

Disclosure classification and redaction

The disclosure class applying separately to the conclusion, the scores and the underlying evidence, the terms under which the assessment may be published or cited, what must be redacted before external release and who approves it, and whether a conclusion may be released without its evidence.

What disclosure class applies to the conclusion, the scores and the underlying evidence separately? access
Under what terms, embargo or licence may the assessment be published or cited? ownership
What must be redacted before external release, and who approves the redaction? privacy
Can a relying party receive the conclusion without the evidence, and what integrity guarantee accompanies it? security

Retention class and disposition referral

The retention class and trigger applying to each part of the record, which policy and system execute destruction or transfer when retention ends, and how a legal hold or open appeal suspends disposition.

What retention class and minimum retention period apply to each part of the record? retention
What event starts the retention clock for the assessment and for its evidence? temporal
Which policy and which system execute destruction or transfer when retention ends? ownership
How does a legal hold or open appeal suspend disposition? exception

Interoperability

Projection into external assessment reporting formats and binding of the vocabularies, units and languages used.

External alignment and vocabulary binding

Which external assessment reporting formats this record must project into, how local outcome values map into each target vocabulary and where that mapping loses information, which code systems, units and language tags are bound at which versions, and what is claimed as conformance to an external standard.

Which external assessment reporting formats must this record project into? interoperability
How do local outcome values map to each target vocabulary, and where is the mapping lossy? interoperability
Which code systems, units and language tags are bound, and at which versions? classification
What is claimed as conformance to an external standard, and what evidence supports that claim? validation

Classifiers Filled

Family: World Models
Category: Activities and processes
Entry kind: aggregate
Navigation path: NAV.ACT.ASM
Domain: ACT.ASM
Industry: Cross-industry
Tags: assessmentevaluationact.asm

What it is Filled

Format-neutral structure of a single assessment/evaluation instance and the bindings it carries to reusable definitional components (criteria source, instrument, scale, decision rule). It spans framing and scoping, criteria binding and tailoring, method and execution parameters, evidence and observation capture, criterion-level determination, aggregation, conclusion and validity, assurance of assessor competence and impartiality, and record lifecycle, disclosure, retention referral and external projection. It is subject-domain agnostic and applies equally to conformity assessment, security control assessment, accessibility evaluation, educational measurement, clinical risk assessment and peer review. It is modelled as an aggregate because the assessment act, its inputs and its determinations share one identity boundary and one finalisation event.

In scope

Identity, granularity and versioning of an assessment record and its supersession chain
Assessment type, stakes, mandate, scheme reference and first/second/third-party relationship
Subject reference, pinned assessed state, scope boundary, sampling and generalisation limits
Binding of criteria to a named catalogue at a pinned version, plus selection, tailoring and interpretation notes
Declared method, mode, depth and coverage, and the instrument or rubric that operationalises each criterion
Execution parameters needed for reproducibility: tool and model references, versions, configuration, input snapshots
Evidence item references, criterion linkage, sufficiency, admissibility and confidentiality qualifiers
Observation records with distinct event and recording times, plus declared uncertainty and traceability pointers
Criterion-level outcomes, rationale, severity and locators; raw, normalised and maximum scores
Aggregation model, weighting, composite score or level, and result-vector expression
Conclusion statement, decision authority, qualifications, dissent, validity window and surveillance obligations
Assessor identity and roles, competence and authorisation references, impartiality and conflict declarations
Independent review, moderation, rater agreement, and validity/reliability/fairness evidence for the instrument
Record state machine, correction, appeal and supersession; disclosure classification, redaction and retention referral
Crosswalks and vocabulary bindings for projection into external assessment reporting formats

Out of scope

Internal structure, lifecycle and state management of the assessed subject itself
Authoring, versioning, publication and normative interpretation of the criteria catalogue or standard
Runtime execution and orchestration of testing tools, scanners, graders or evaluator engines and their operational telemetry
Issuance, signing, suspension, withdrawal and verification of certificates, credentials or attestations
Enforcement of consequences: certification decisions with legal effect, licensing, sanction, market withdrawal or remediation workflow
Identity management, employment and qualification records of persons and organisations
Physical or digital custody, storage and chain-of-custody execution for evidence objects
Calibration programmes, metrological traceability management and instrument maintenance
Training, validation and deployment lifecycle of predictive or generative scoring engines
Platform audit-trail and access-log semantics, storage and evaluation
Execution of retention, destruction, transfer or erasure actions
Risk register lifecycle, risk treatment and plan-of-action-and-milestones case management
Population-level aggregation of many assessments into reputation scores or league tables
Commercial terms, scheduling and cost accounting of assessment services

Why it exists Filled

Distinguishing features Filled

Records a determination against version-pinned criteria with a declared method, scale and decision rule.
Separates determination from decision: the assessor and the actor who records the conclusion can differ.
Unlike a measurement record, it interprets observations against criteria rather than storing raw values.
Unlike an attestation or certificate, it does not issue a credential with legal effect.

What robots and AI may and may not do Filled

Must not

Record a conclusion without a version-pinned criterion and supporting evidence.
Publish a score without its scale, aggregation model and decision rule.
Treat an untested or indeterminate outcome as a pass.
Infer regulatory or contractual effect from the quality of the method.
Delete an assessment to settle a dispute or hide an unfavourable result.

Only with a human decision

Recording a conclusion that has contractual, regulatory or employment consequences.
Changing the decision rule or weights after evidence has been seen.
Resolving a dispute about an assessment outcome.

May

Frame an assessment and bind criteria at an explicit version.
Register evidence and link each item to the criteria it supports.
Compute a composite result with the declared aggregation model and publish it with its scale reference.
Record indeterminate, not-applicable and untested outcomes explicitly.

Moral aspects Filled

Assessments of people affect careers and access to opportunities; criteria and evidence must be open to the person assessed.
Aggregate scores can hide failures on single critical criteria.
Bias in criteria or method can disadvantage groups systematically.

Who is affected

Persons or organisations being assessed
Decision makers relying on the result
Assessors whose judgement is recorded

Owners Filled

Steward

Name a single accountable assessment owner and, separately, a decision authority entitled to record conclusions; the two roles must be distinguishable in every record so determination and decision remain separable.

Roles

Assessment Owner: Accountable for the record's scope, completeness and timeliness; Approves scope and sampling changes and records their effect on results already captured; Names the assessors, the reviewer and the decision authority for the record
Assessor / Evaluator: Performs selection and determination against the bound criteria; Registers evidence references, linkages, sufficiency judgements and observations with dual timestamps; Records criterion outcomes, rationale and scores, including indeterminate and not-applicable values; Where the assessor is an automated agent, is identified as such with its tool reference, version and configuration
Technical Reviewer / Moderator: Reviews determinations independently of the assessor who made them; Records agreement statistics and adjudicates disagreements between assessors or tools; May return a record to draft with reasons, and may not decide a conclusion on determinations they reviewed
Decision Authority: Applies the declared decision rule to the composite result and its uncertainty; Records the conclusion, its type, qualifications, dissent and validity window; Accountable for the referral to any downstream attestation and for stating what the conclusion does not entitle
Criteria and Vocabulary Steward: Maintains bindings to external criteria catalogues, scales, outcome vocabularies and decision rules; Publishes crosswalks, successor mappings and deprecations for target reporting formats; Assesses the impact of catalogue and vocabulary version changes on records that are still in draft
Disclosure and Records Officer: Sets disclosure class and publication terms per record part and approves redactions; Issues and withdraws public renditions and records the redaction rule version applied; Assigns the retention class and trigger and refers disposition to the owning records-retention policy model

Links to other meta-models Filled

child

WM-ACT-009 (parent model in NAV.ACT) - Inherit generic activity semantics — actor, occurrence timing, status and provenance — and specialise them for criterion-referenced judgement producing a determination and a conclusion. Generic activity identity and timing machinery is not restated here.

references

Criteria / requirement catalogue model (control catalogue, standard clause set, rubric registry) - Carry the criteria source reference, its pinned version, criterion locators, local interpretation notes and tailoring decisions. Criterion text, structure, normative status, versioning and publication remain with the catalogue.
Subject-domain model of the assessed object - Carry the subject reference and a pinned state descriptor of the object as assessed. The subject's own attributes, lifecycle and state transitions remain with the domain model.
Party, organisation and role model - Resolve assessor, requester, relying party and decision-maker references and their roles on this record. Party identity, contact data and organisational structure remain with the party model.
Competence, accreditation and designation registry - Carry accreditation, notification or designation references and the scope they cover for this assessment. Granting, surveilling, suspending and withdrawing accreditation belongs to the accreditation authority.
Evidence custody and records repository model - Carry evidence pointers, integrity digests, collection times and sensitivity classes. Storage, chain-of-custody execution and retrieval of the evidence bytes remain with the repository.
Measurement, observation and metrology model - Carry measured values, units, stated uncertainty and calibration references used as evidence. Uncertainty evaluation methodology, calibration programmes and traceability chains remain with the metrology model.
Attestation, certificate and credential model - Carry a referral marker and a pointer where the conclusion is attested downstream. Issuance, signing, suspension, withdrawal and verification of certificates or credentials belong entirely to that model.
Risk register and treatment model - Carry pointers where a determination seeds a risk entry. Risk characterisation lifecycle, mitigating factors, remediation tracking and deadlines belong to the risk model.
Records retention and disposition policy model - Carry the retention class, retention trigger, legal-hold flag and the disposition referral record. Setting periods and executing destruction, transfer or erasure belong to that policy model.
Access control and information classification model - Carry the disclosure class and publication terms per record part. Evaluating access requests, enforcing decisions and storing access logs belong to that model.
Assessment scheme, programme or examination framework model - Carry the scheme reference and the parameters it fixed for this occasion, such as required criteria profile, surveillance interval and eligibility. Scheme rules, cycles and governance belong to the programme model.
Dispute, appeal and complaint case model - Carry the appeal reference, its outcome and the hold it places on disposition. Conducting proceedings, managing timelines and recording hearings belong to the case model.

composes

Provenance model aligned to W3C PROV - Attach Entity, Activity and Agent provenance to assessment content: attribution of determinations to agents, derivation of conclusions from evidence, and start and end times of assessment activities.

aligned

Code list, vocabulary and unit registry - Bind outcome vocabularies, scale definitions, severity schemes, unit code systems and language tags at stated versions, and publish crosswalks between them. Vocabulary authoring and deprecation belong to the registry.

neighbor

Measurement / observation model - An assessment consumes measured or observed values as evidence and adds criterion-referenced judgement; a bare measured value with an uncertainty statement is a measurement result, not an assessment. Units, uncertainty evaluation and calibration traceability are referenced from the metrology model.
Criteria / requirement catalogue model - This model carries a criterion reference, its pinned catalogue version, local applicability and interpretation notes only. Criterion text, structure, normative status and catalogue lifecycle belong to the catalogue, exactly as OSCAL separates a control catalogue from assessment results and XCCDF separates a Benchmark from a TestResult.
Attestation / certificate / credential model - ISO/IEC 17000 separates determination and review from decision and from attestation. This model ends at the recorded conclusion; issuing, signing, suspending or revoking a certificate or credential is a downstream act, and in EU product law only a designated notified body may give that act legal effect.
Evaluator or test-execution engine - EARL records the assertor and the test mode as attributes of an assertion, and XCCDF records the checking system as a reference. Naming a tool, its version and its configuration never transfers ownership of running that tool, its scheduling, or its operational logs.
Risk register and treatment model - OSCAL keeps observations, findings and risks distinct and hands risks onward to a POA&M. This model may produce findings that seed risks, but risk characterisation lifecycle, mitigating factors, remediation tracking and deadlines belong to the risk model.
Party, competence and accreditation registry - Assessor identity, competence evidence, accreditation and designation scope are referenced by identifier and version. Whether a body is accredited or notified, and for what scope, is asserted and maintained by the accreditation authority, not by an assessment record.
Provenance and audit-trail models - Content provenance (which agent produced which determination, from which evidence, when) is mixed in from a PROV-aligned model and retained. Platform-level record access and change logging is a separate audit model whose storage and evaluation this model does not own.
Records retention and disposition model - This model declares a retention class, a retention trigger event and a legal-hold flag, and records the referral. Setting periods and executing destruction, transfer or erasure is owned by the adopting Dimension's records policy.
Aggregate rating / reputation model - schema.org AggregateRating rolls many individual ratings into counts and averages across a population. That population-level statistic is a sibling concern; this model bounds one assessment occasion on one declared subject and scope.
Assessment programme / scheme model - A conformity assessment scheme, examination programme or continuous-monitoring programme defines recurring rules, cycles and eligibility. This model references the scheme and carries the parameters it fixed for this occasion; the scheme's own rules and cycle belong to the programme model.

parent

WM-ACT-009

What else AI and robots need to interact with it Filled

Sources Filled

ISO/IEC 17000:2020 Conformity assessment — Vocabulary and general principles - ISO/IEC (Joint Technical Committee)
OSCAL Assessment Results Model (Assessment Layer) - National Institute of Standards and Technology (NIST)
NIST SP 800-53A Rev. 5 — Assessing Security and Privacy Controls in Information Systems and Organizations - National Institute of Standards and Technology (NIST)
NISTIR 7275 Rev. 4 — Specification for the Extensible Configuration Checklist Description Format (XCCDF) Version 1.2 - National Institute of Standards and Technology (NIST)
Evaluation and Report Language (EARL) 1.0 Schema - World Wide Web Consortium (W3C)
PROV-O: The PROV Ontology - World Wide Web Consortium (W3C)
FHIR R5 RiskAssessment resource - Health Level Seven International (HL7)
Question and Test Interoperability (QTI) 3.0 — Implementation and Best Practices - 1EdTech Consortium (formerly IMS Global)
Common Vulnerability Scoring System v4.0 Specification Document - Forum of Incident Response and Security Teams (FIRST)
Common Criteria for Information Technology Security Evaluation, CC:2022 Release 1, and CEM:2022 - Common Criteria Recognition Arrangement (CCRA)
ILAC-G8:09/2019 Guidelines on Decision Rules and Statements of Conformity (ILAC Guidance Series) - International Laboratory Accreditation Cooperation (ILAC)
Conformity assessment (Single Market — Goods, building blocks) - European Commission
Website Accessibility Conformance Evaluation Methodology (WCAG-EM) 2.0 - World Wide Web Consortium (W3C)
schema.org Rating, Review and AggregateRating types - Schema.org Community Group (W3C)
JCGM publications, including JCGM 106:2012 The role of measurement uncertainty in conformity assessment and JCGM 100:2008 (GUM) - Joint Committee for Guides in Metrology (BIPM)
Standards for Educational and Psychological Testing (open-access edition) - AERA, APA and NCME

Open questions

Ratify the composition links and the parent link to WM-ACT-009 once the relationship contract is populated, obtaining WM-ACT-009's own rationale so the parent claim can be compared rather than inherited from a registry field.
Resolve the supersession-chain question: decide whether an assessment series is a distinct identity that owns the effective-record flag, or whether the flag must be derivable from version metadata on each record.
Establish the source basis and a boundary note for the access-control model that the checklist refers enforcement to, then re-derive the access dimension from actual findings and questions rather than prose.
Source jurisdiction-specific retention periods, appeal windows, disclosure duties and erasure rights from records-management and legal instruments, none of which were consulted in this research round.
Re-source the EU legal-effect claims to primary instruments — Decision 768/2008/EC, Regulation (EU) 2019/1020 and the Blue Guide (2022) — instead of the Commission portal summary currently marked primary_source.
Obtain independent second-provider or qualified human expert review of the aggregate boundary, artifact-ownership rules and outcome-vocabulary crosswalks when the Grok waiver is lifted or an alternative reviewer is authorised.
Specify the artifact-provenance semantics of the serial flag, then re-test whether af-instrument-definition, af-technical-manual, af-assessment-procedure and af-crosswalk-table are owned artifacts or referenced instance-invariant assets.
Domain-specific criteria semantics — clinical, educational, financial, environmental, safety — are referenced but not modelled; a Dimension adopting this model in a regulated domain will need a sibling criteria model.
Sampling statistics beyond declaration: no power analysis, confidence-interval construction or acceptance-sampling plan is modelled.
Adaptive and computerised-adaptive testing item-selection algorithms, item exposure control and equating are outside the instrument finding.
Benchmark harness specifics for evaluating machine-learning systems — dataset splits, contamination and leakage controls, held-out set governance — are only partly reached by the reproducibility finding.
Panel deliberation and consensus protocols (Delphi, expert elicitation, jury scoring) are summarised as roles and agreement statistics rather than modelled as procedures.
Cross-border mutual recognition of assessment results, and the conditions under which one jurisdiction accepts another's conclusion, are not modelled.
Effort, cost and scheduling of assessment engagements are excluded.
Continuous and streaming assessment is only partly covered: surveillance obligations are declared, but the monitoring cadence and its state machine belong to the programme model.
Aggregation of many assessments into population statistics, league tables or reputation scores is excluded and left to a sibling model.
Digital signature profiles and long-term validation of signed assessment reports are referenced through integrity digests but not specified.

Machine files

Card as JSON Specification AGENTS.md Source Card backlog vercy.card/1

Provenance

world-models research · reviewable-draft

Built from: models/wm-act-034-assessment-evaluation/spec.yaml, ver-cy/world-models/card-supplements/wm-act-034-assessment-evaluation.json