Source use and currentness
The HCD.1/.3/.4 direct-source qualifications remain dated 1 September 2026. The NICE social-anxiety guideline (BEH-05) was separately checked on 31 August 2026; refresh its affected clinical boundary when the recommendation changes.
HCD.1 uses Kraiger and Ford’s workplace-instruction review (2021) (COG-10) for the event-to-transfer distinction, and Dell’Acqua and colleagues’ Navigating the jagged technological frontier (AI-03) for a bounded independent-error-challenge cue. HCD.1 also uses current FPF, qualified target-domain results, and evidence for the person’s later Work.
HCD.3 uses only the bounded diagnostic warning from Melanson and Fahmie’s Functional analysis of problem behavior: a 40-year review (2023; BEH-01), the clinical stop from NICE, and the target-engagement limits from Sheeran and colleagues’ Self-determination theory interventions for health behavior change (2020; MOT-01) and Does increasing autonomous motivation or perceived competence lead to health behavior change? (2021; MOT-02). HCD.3 also uses E.23.CAE when its differential can change the decision, and direct specialist evidence when the question needs it. These health and clinical sources supply no universal human-capability diagnosis or clinical procedure for HCD.
HCD.4 uses Winget and Persky’s A practical review of mastery learning (2022; COG-06) for local thresholds, correction and retest; Gutierrez de Blume’s Calibrating calibration (2022; COG-07) for self-monitoring accuracy; Arieli-Attali and colleagues’ Expanded evidence-centered design (2019; COG-08) for claim/task/evidence/inference traceability; and Abdelrahman and colleagues’ Knowledge tracing: a survey (2023; COG-09) for model-prediction limits. The design framework is not a causal intervention study, and none of these sources establishes P14’s current capability, a universal ladder, or transfer and retention. C.32.MWA, C.11, and direct task evidence supply the remaining architecture, choice, and local evidence conditions.
The HCD.2 and HCD.5–HCD.10 bodies bind their source uses to programme comparison, Method selection, task design, support, provider or environment operation, focused practice, and condition-sensitive progression. Their shared source line is summarized here; the individual bodies carry the decisive limits and reopen conditions.
| Source group | Contribution and limit used by HCD.2 and HCD.5–HCD.10 |
|---|---|
Maintained 4C/ID model and Paas and van Merriënboer (2020; COG-05) | Sufficiently whole complex tasks, integrated information, worked support and expertise-sensitive guidance; no local programme effect, mandatory sequence or universal fading rule. |
Winget and Persky (2022; COG-06) and Wisniewski, Zierer, and Hattie (2020; MOT-05) | Corrective activity, retest and information-bearing feedback under task and receiver conditions; no universal threshold, feedback type or later-work inference. |
Carpenter, Pan, and Butler (2022; COG-01), Brunmair and Richter (2019; COG-03), and Czyż, Wójcik, and Solarská (2024; COG-04) | Spacing, retrieval, interleaving and delayed-transfer candidates whose value depends on material, similarity, setting and horizon; no universal gap, mix, repetition count or professional schedule. |
Kraiger and Ford (2021; COG-10), Matsuo and Tsukube (2020; SOC-02), Brooks and colleagues (2024; SOC-05), and Wiese, Kilty, and Bennett (2018; SOC-03) | Separation of instructional events, learner activity, provider contributions, access, supervision, trust, risk and service pressure; no generic provider credential, capability effect or transplanted clinical governance. |
Tullis, Goldstone, and Hanson (2015; COG-11), Dell’Acqua and colleagues (AI-03), and Bastani and colleagues (2025; AI-04) | Assisted and later unassisted performance, human checking, and AI-support conditions remain distinct; laboratory, consulting-task and school-mathematics results establish neither a universal support schedule nor an adult workplace effect or AI ban. |
Macnamara, Hambrick, and Oswald (2014; BEH-02), Nurse and colleagues (2024; BEH-03), and Sheeran and colleagues (2020; MOT-01, 2024; MOT-03) | Limits on universal deliberate-practice doses, bounded focused-practice ingredients, target engagement, and conditional if–then planning; no fixed hours threshold, cross-domain causality or professional transfer guarantee. |
The added HCD.11–HCD.14 source uses were qualified on 3 September 2026. The bodies bind each source to a specific assessment or revision rule:
| Source | Contribution and limit used here |
|---|---|
| 2014 Testing Standards, Chapter 1 | Intended assessment interpretation and use, task fitness and scoring evidence; design guidance, not local validation or authorization. |
| Arieli-Attali and colleagues, expanded evidence-centred design, 2019 | Claim/task/observation/inference and learning during assessment; no treatment effect is inferred. |
| Kraiger and Ford, workplace instruction, 2021 | Instruction, learner activity, learning and work transfer remain distinct; no particular programme’s effectiveness is established. |
| Carpenter, Pan and Butler, 2022 and Yang and colleagues, 2021 | Retrieval/testing can affect learning under specified conditions; no universal delay, engineer effect size or all-domain transfer guarantee. |
| Bastani and colleagues, 2025 | Assisted practice and later unassisted performance can differ; school-mathematics evidence is not adult occupational transfer or long retention. |
| Schwartz and Martin, 2004 | Historical design evidence for distinguishing learning with a new resource from performance without it; not a claim that this programme accelerates learning. |
The HCD.15–HCD.17 applied-learning source uses were qualified on 1 September 2026. The cultural-continuation comparison was checked on 24 August 2026. The new bodies use only claims that change their practitioner move:
| Source group | Contribution and limit used by HCD.15–HCD.17 |
|---|---|
Kraiger and Ford (2021; COG-10); Matsuo and Tsukube (2020; SOC-02); Wiese, Kilty, and Bennett (2018; SOC-03) | HCD.15 separates instructional events, learner activity, learning outcomes, transfer and job performance and may use bounded modeling, coaching, supervision and graduated access. No universal cadence, provider form, transplanted clinical governance or E27 trajectory follows. |
Abedini, Abedin, and Zowghi (2021; SOC-04) and Brooks and colleagues (2024; SOC-05) | Communities may supply participation, exchange, facilitation, platform, coordination, access and professional support. Membership and activity establish neither capability, cultural continuation nor causal effect. |
Brynjolfsson, Li, and Raymond (2025; AI-02); Dell’Acqua and colleagues (2026; AI-03); Bastani and colleagues (2025; AI-04) | HCD.16 separates deployed equipped performance, rare-outage same-job signals, task-frontier error, assisted practice and later unassisted performance. These populations and tasks do not establish an adult release-engineering Method, general AI benefit or harm, other-Work transfer, or long retention. |
Bauer and colleagues (2024; BEH-08); Mertens and colleagues (2022; MOT-06) with Maier and colleagues’ publication-bias critique | HCD.16 can reopen one used mechanism, population, condition or pooled-effect claim and retain further probe; it prescribes neither clinical counterconditioning nor public-policy nudging. |
Singh and colleagues (2021; SOC-06) through current C.36 | HCD.17 distinguishes generation, transmission, receiving use, recognition, selection, memory, retention and loss. The review supplies no HCD intervention recipe, authority or population capability. |
| The CDIO version history, Agile Manifesto history, OMG Essence editions and Google SRE’s account of community-driven software adoption | These histories describe creation, revision, institutional selection, publication and reported use of particular methods or standards. Use each account for the events and participants it describes. To assess whether a learning provider or learner performs a practice, retains it or improves capability, examine that practice and its results in the relevant setting. |
HCD.18/.19 use the material profile’s source-informed comparisons, with their specific source roles, limits and reopens in the bodies. Their task-based evaluation and ordinal anchors are professional adaptations, not a validated universal score. The M0/M1 walkthrough preserves actual prepared-agent responses; its constructed task facts, material revisions, author-side diagnostic replay and missing target-human observations remain explicit.
The programme and L17 engineering exercise supply designed tasks, stipulated domain facts and constructed observations, not an empirical learner dataset. Real engineering, safety or professional claims require current direct-domain sources and actual case evidence.
ImpactReview-17-Domain-Oracle-v1 is constructed case data, not a source-ledger entry or currentness claim. Replacing any stipulated engineering correctness or criticality value with a real-world claim requires a named current target-domain Method or result and qualified safety or specialist source.
Refresh only the affected pattern or case claim when a relied-on FPF distinction, direct source contract, target-domain Method or result, specialist boundary, holder, Work family, configuration, evidence window, critical-error predicate, case truth posture, support arrangement, practice condition, or receiving decision changes. A new publication alone does not reopen unaffected claims.
HCD.20–28 use task-sensitive contributions from the official 4C/ID, TBL, Jigsaw and FROG accounts and bounded research on collaboration, visibility and active observation. Their SoTA-Echoing sections state the adopted operation and its limit. HCD.21 adds explicit composition and actual-contact constructions; HCD.24/.25 distinguish comparative judgement from demonstrated transfer; HCD.26 distinguishes prompting from uptake. Juxtapeer, group-formation and contact studies, Edwards and the later CMU feedback study retain their task, population and observation limits. HCD.7 applies tutor-facing assistance through its existing competence, checking and capacity conditions. The linked diagnostic product, numerical resource comparison and recovery variants are authored constructions; they are not empirical outcomes or evidence that one grouping is generally best.