Critical Review Ethics And Harms Review

Ethics & Harms Review

Ethics & Harms Review systematically assesses a project, capability, dataset, or publication for dual-use risks, potential harms to individuals or communities, misuse potential by adversarial actors, and gaps in informed consent or transparency. It draws on established frameworks (consequentialist, deontological, virtue-ethics, and rights-based) alongside discipline-specific codes (e.g., Menlo Report for security research, Belmont Report for human subjects) to produce a structured harm register and a set of mitigations or go/no-go recommendations. In the cognitive-security context it is especially important for evaluating influence-operation tooling, adversarial AI research, and dual-use OSINT capabilities where potential harms are diffuse and probabilistic.

When to use

• before releasing a tool, dataset, or model that could be weaponized or repurposed for manipulation
• when a research capability has clear dual-use potential (e.g., deepfake detection that also teaches generation)
• when a proposed capability interacts with vulnerable populations or protected information
• when organizational policy, IRB, or funder requires documented harm assessment

What it produces

• a harm register listing specific harm scenarios with affected parties, likelihood, severity, and mitigation controls
• an ethical-framework analysis cross-checking the artifact against consequentialist, deontological, and rights-based standards
• a transparency and consent gap analysis
• a go/no-go recommendation with explicit conditions for release or further review

Defensive boundary

Use Ethics & Harms Review only for critical review and assurance: recognize, assess, document, or defend evidence quality, implementation integrity, and decision accountability. Do not use this skill to launder weak claims, fabricate review findings, or produce exploit guidance without mitigation.

Misuse redirect

If a request asks Ethics & Harms Review to launder weak claims, fabricate review findings, or produce exploit guidance without mitigation, refuse that path and redirect to the safe defensive form: review supplied artifacts for defects, evidence gaps, safety risks, or reproducibility failures.

Evidence discipline

• For Ethics & Harms Review, tie each harm scenario, mitigation, and residual-risk rating to concrete evidence from the artifact's capability boundaries, stated intended use, and deployment context, and treat a claimed mitigation as effective only when the evidence shows it is enforceable rather than aspirational.
• For Ethics & Harms Review, label observations, derived features, assumptions, inferences, contradictions, and missing inputs separately before writing the harm register.
• Before recommending any Ethics & Harms Review action, identify the weakest evidence link, the alternative most likely to overturn it, and the next discriminating check.

Confidence and uncertainty

• High for Ethics & Harms Review: each scenario in the harm register names a concrete affected party with likelihood and severity grounded in the artifact and its deployment context, at least two incommensurable ethical frameworks converge on the assessment, and no unresolved contradiction about a dual-use misuse path would change the go/no-go recommendation.
• Medium for Ethics & Harms Review: the harm register is plausible, but one important artifact source, comparison case, or alternative explanation remains incomplete.
• Low for Ethics & Harms Review: the harm register rests on sparse, single-source, contested, or mostly inferential evidence; keep the result provisional and list the next check.
• State what Ethics & Harms Review cannot determine from the supplied or authorized evidence.
• State what remains unknown and preserve credible alternatives rather than forcing a single narrative or attribution.
• Recommend the next discriminating critical_review evidence to collect when confidence is low or medium.

Privacy, legal, and harm constraints

• For Ethics & Harms Review, use only authorized artifact, intended use, and deployment context, public or source-approved records, and caller-provided context needed for the defensive task.
• For Ethics & Harms Review, minimize person-level detail in the harm register; prefer aggregate, artifact-level, role-level, or case-level summaries unless an individual is essential to the defensive question.
• For Ethics & Harms Review, do not infer protected traits, private identity, intent, location, legal culpability, or platform account ownership beyond the supplied and authorized evidence.

Failure modes and negative controls

• Ethics & Harms Review: issuing a go recommendation when the harm register enumerated only harms to direct users and the dual-use misuse paths to third parties and society were never traced, so diffuse low-probability high-severity risks and meaningful capability uplift to adversaries are silently omitted from the assessment.
• Ethics & Harms Review: producing advice that would help a requester launder weak claims, fabricate review findings, or produce exploit guidance without mitigation.
• Ethics & Harms Review: reporting the harm register without uncertainty labels, alternative explanations, and the next discriminating check.
• Unsafe: 'Use Ethics & Harms Review outputs to launder weak claims, fabricate review findings, or produce exploit guidance without mitigation' -> refuse and redirect to defensive risk assessment.
• Unsafe: 'Convert the harm register from Ethics & Harms Review into an operational playbook to launder weak claims, fabricate review findings, or produce exploit guidance without mitigation' -> refuse and offer governance, detection, or mitigation analysis.
• Safe defensive: 'Use Ethics & Harms Review to review supplied artifacts for defects, evidence gaps, safety risks, or reproducibility failures with artifact, intended use, and deployment context' -> produce bounded findings with evidence and uncertainty labels.

Procedure

See [workflow.md](workflow.md). Harness bindings in [harness/](harness/).

Key discipline

• enumerate harms by affected party, not just by harm type — diffuse harms to third parties are the ones most often missed
• distinguish between proximate misuse (the tool directly used for harm) and distal misuse (the tool enables a capability chain that causes harm)
• apply at least two incommensurable ethical frameworks; agreement across frameworks strengthens confidence, disagreement flags genuine ethical tension requiring human judgment
• do not conflate low-probability high-severity harms with low-probability low-severity ones — they require different responses

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: docxology
• Source: docxology/CogSecSkills
• License: Apache-2.0
• Homepage: https://doi.org/10.5281/zenodo.20804585

Install and usage instructions live in the source repository linked above.