Examples
Worked examples
- Is an instance
An independent auditor tests an automated hiring-screening tool for equal selection rates across gender and race categories before an employer may use it, per NYC Local Law 144.
- Is an instance
A university tests an admissions-support model for calibration parity across applicant demographic groups.
Counter-examples
Looks similar, but isn't
- Not an instance
A vendor’s own internal fairness testing, without an independent auditor and without published results, does not satisfy a Local Law 144-style bias-audit requirement, which specifically mandates auditor independence.
Editorial commentary
A bias audit is a model audit scoped specifically to fairness and discrimination criteria rather than the full range of a model audit’s possible scope (performance, robustness, security, privacy). It measures quantitative fairness metrics — equal-opportunity difference, demographic parity, calibration parity, predictive-parity gaps — across pre-specified groups, alongside qualitative failure-mode analysis of how and where errors concentrate.
Why the choice of metric matters
Choosing which fairness metric to audit against is itself an ethically loaded decision, not a neutral technical one. Chouldechova’s impossibility results show that, except in special cases, distinct fairness metrics cannot generally all be satisfied simultaneously by the same model — a system calibrated to equalise false-positive rates across groups will typically not also equalise predictive parity. A credible bias audit states, in advance, which fairness definition it is measuring against and why, rather than reporting whichever metric happens to look favourable.
Regulatory anchor: NYC Local Law 144
New York City’s Local Law 144 (effective 2023) requires employers using an ‘automated employment decision tool’ to have it bias-audited by an independent auditor — not the vendor itself — within one year before use, to publish a summary of the audit results, and to notify candidates that such a tool is in use. Several other US states have since introduced comparable requirements. This is the clearest example of bias-audit obligations moving from voluntary best practice into binding law, and it specifically requires independence: an audit performed by the tool’s own developer does not satisfy the law’s intent.
Relationship to model audit — these stay distinct terms
A general model audit can cover fairness alongside performance, security and robustness in one exercise; a bias audit is the narrower case where fairness is the entire scope, often because a specific law or policy (like Local Law 144) mandates exactly that scope and nothing broader. The two terms are not redundant: ‘model audit’ is the umbrella independent-verification concept, and ‘bias audit’ names the fairness-specific instance of it that increasingly carries its own distinct legal requirements.
References
- Buolamwini, Gebru, ‘Gender Shades’ (FAccT, 2018); Chouldechova, ‘Fair prediction with disparate impact’ (Big Data, 2017); NYC Local Law 144 (2021, effective 2023).
Also known as
fairness audit · algorithmic bias audit
Machine-readable encodings
Use in your systems
<role vocab="credit"
vocab-identifier="https://casrai.org/dictionary/"
vocab-term="Bias audit (model)"
vocab-term-identifier="https://casrai.org/dictionary/term/bias-audit-model" />{
"@context": "https://schema.org",
"@type": "DefinedTerm",
"@id": "https://casrai.org/dictionary/term/bias-audit-model",
"name": "Bias audit (model)",
"identifier": "https://casrai.org/dictionary/term/bias-audit-model",
"description": "A model audit scoped specifically to fairness and discrimination criteria: measuring quantitative fairness metrics and qualitative failure-mode patterns across pre-specified groups, often to satisfy a specific legal requirement.",
"inDefinedTermSet": "https://casrai.org/dictionary/domain/ai-ml-research-outputs#set",
"url": "https://casrai.org/dictionary/term/bias-audit-model",
"sameAs": [
"fairness audit",
"algorithmic bias audit"
],
"license": "https://creativecommons.org/licenses/by/4.0/",
"publisher": {
"@id": "https://casrai.org/#organization"
},
"author": {
"@id": "https://casrai.org/#editorial-team"
},
"datePublished": "2026-05-21T02:22:50",
"dateModified": "2026-08-22T16:26:23",
"inLanguage": "en-GB",
"isAccessibleForFree": true
}
