Governance frameworks8

The volume's ethics chapter names ethics washing as addressing ethical concerns superficially, to gain public trust, while making no substantive change to practice. It identifies three forms this takes: ethical statements that are vague or go unenforced, ethics boards constituted with limited authority, and adoption of frameworks that carry no accountability mechanism. The chapter offers this as a taxonomy of forms, not as a measurement of how often each occurs.

Sources: an2026a

Appears on: /pan-lab

The volume's mental-health chapter reads clinical decision support, digital phenotyping and mental-health conversational agents into the high-risk tier of the EU AI Act, which attaches conformity assessment before placing on the market, post-market monitoring afterwards, and documented human-oversight measures throughout. The obligations follow from the tier rather than from any property of a particular model.

Sources: yang2026c, europeanparliamentandcouncil2024

Appears on: /practice/vendor-gate

The volume's governance chapter matches regulatory intensity to risk level rather than applying one standard everywhere, so that a low-stakes use is not governed as though it were a high-stakes one. The social-work-specific tier table it offers is the chapter's own conceptual synthesis rather than a measured classification, and should be read as a proposal.

Sources: huang2026b, europeanparliamentandcouncil2024

Appears on: /practice/oversight-cadence

A review reported in the volume's governance chapter found that none of the nine principal social-work codes of ethics it examined explicitly addresses the use of predictive algorithms in decision-making. The professional obligation exists in the codes; the specific practice does not appear in them.

Sources: huang2026b, reamer2023

Appears on: /practice/oversight-cadence

The volume's sexual and partner violence chapter draws a scope line around predictive risk modelling and restates it twice: such models are used in research contexts to study population-level risk factors, they are not intended for individual-level decision-making in clinical or legal settings, and they are not intended for practitioners to screen or label individuals directly in real-world settings. The chapter treats the distance between that declared scope and a case-level use as a central governance danger rather than as a modelling defect, and reports no measurement of how often the line is crossed.

Sources: fang2026a

Appears on: /pan-lab

The volume's governance chapter places two standing duties on an agency that deploys AI, both distinct from any pre-deployment approval. First, an incident-reporting protocol that enables timely identification and remediation of algorithmic harm - discriminatory treatment, a biased risk assessment, a misdiagnosis, a breach of confidentiality. Second, transparent channels through which both the people served and the practitioners can report concerns or unexpected effects, so the accountability loop closes after deployment rather than ending at approval. The chapter is a conceptual synthesis and is cited as one: its four-tier social-work risk taxonomy is labeled by its own author as an original construction, informed by but not derived from binding regulation. It may be cited as a framework and must never be presented as a regulatory classification of any deployment in this registry.

Sources: huang2026b

Appears on: /domains/behavioral-health-triage, /domains/public-benefits, /domains/child-welfare, /domains/hiring-employment-screening

The volume's ethics chapter names five redress mechanisms as the operational answer to a diffuse responsibility gap: social impact assessment before deployment, audit trails documenting data inputs and decision processes, appeal mechanisms letting a person challenge an AI-driven decision, liability frameworks allocating responsibility by role, and ethics oversight committees seating practitioners, clients and technologists. It is a normative framework proposal, and no effect size is attached to any of the five.

Sources: an2026a

Appears on: /pan-lab

A model card is not a provenance label, and the two govern different objects. A provenance label marks machine-originated content inside the record, so readers and retrieval discount it. The volume's introduction describes a model card or nutrition label as documentation of the tool — how it functions, what its limitations are, and the contexts in which it should not be applied. Keeping the two apart keeps a record-side control and a procurement-side control from being read as one.

Sources: an2026b

Appears on: /practice/provenance-labeling