disparate impact
In plain wordsWhen a rule or system that treats everyone alike on paper still lands much harder on one group, such as women or older workers; the law can forbid this even if no one intended it.
- 1
Law
In US anti-discrimination law, the doctrine that a facially neutral practice is unlawful if it disproportionately disadvantages a group defined by a protected attributeprotected attribute, n.: A personal trait, such as race, sex, religion, age or disability, that the law says must not be used to treat people unfairly. and the employer or provider cannot show it is justified by business necessity, or if a less discriminatory alternative was available but not adopted. It contrasts with disparate treatment, which requires explicit or intentional differential treatment. EU and UK law address similar conduct as indirect discrimination, permitted only if objectively justified and proportionate.
- 2
Machine learning
A measure of algorithmic biasalgorithmic bias, n.: Unfairness in software that judges or serves people: it consistently produces worse results for some groups, such as women, older people or ethnic minorities, often because it learned from skewed or incomplete records. borrowing the legal concept, usually the ratio of favorable-outcome rates between a disadvantaged and an advantaged group; ratios below 0.8, the four-fifths rule of US employment-selection guidelines, are treated as a warning sign. Because models can reproduce disparities through correlated features even when protected attributes are excluded, disparate impact is the main legal route for challenging unintended algorithmic discrimination.
The screening model's selection ratio for applicants over 40 was 0.62, well below the four-fifths threshold.
Origin
Established by the US Supreme Court in Griggs v. Duke Power Co. (1971), which struck down diploma and aptitude-test requirements that excluded Black workers without predicting job performance, and codified in the Civil Rights Act of 1991. The four-fifths rule comes from the 1978 Uniform Guidelines on Employee Selection Procedures. Solon Barocas and Andrew Selbst's 'Big Data's Disparate Impact' (2016) brought the doctrine into AI debates.
Usage note
As of October 2026 the doctrine's federal standing is contested. Executive Order 14281 (April 2025) directed agencies to deprioritize disparate-impact enforcement, and a June 2026 opinion of the Justice Department's Office of Legal Counsel, binding on federal agencies but not on courts, called the EEOC's disparate-impact approach and the Uniform Guidelines unconstitutional. The statutory provision remains, and private suits and state laws are still available.
Related
- algorithmic biasalgorithmic bias, n.: Unfairness in software that judges or serves people: it consistently produces worse results for some groups, such as women, older people or ethnic minorities, often because it learned from skewed or incomplete records.
- demographic paritydemographic parity, n.: A fairness test that asks whether a system says yes (approves the loan, shortlists the applicant) at the same rate for every group, whatever other differences exist between them.
- protected attributeprotected attribute, n.: A personal trait, such as race, sex, religion, age or disability, that the law says must not be used to treat people unfairly.
- proxy discriminationproxy discrimination, n.: Discrimination that happens when a system cannot see a trait like race directly but relies on something closely linked to it, such as a postcode, and so treats groups unequally anyway.
- algorithmic fairnessalgorithmic fairness, n.: The effort to make automated decisions fair: to say in numbers what fair treatment means, measure whether a system achieves it, and correct the system if it does not.
Sources
- Griggs v. Duke Power Co., 401 U.S. 424 (1971).
- Barocas, S. & Selbst, A. D. (2016). Big Data's Disparate Impact. California Law Review 104, 671–732.
Defined in law or standards
US Civil Rights Act of 1964, Title VII, §703(k) (42 U.S.C. §2000e-2(k)), added by the Civil Rights Act of 1991
Mentioned in
algorithmic biasalgorithmic bias, n.: Unfairness in software that judges or serves people: it consistently produces worse results for some groups, such as women, older people or ethnic minorities, often because it learned from skewed or incomplete records., algorithmic fairnessalgorithmic fairness, n.: The effort to make automated decisions fair: to say in numbers what fair treatment means, measure whether a system achieves it, and correct the system if it does not., Blueprint for an AI Bill of RightsBlueprint for an AI Bill of Rights, prop. n.: A 2022 White House document listing five protections Americans should have when automated systems affect their lives — safety, fairness, privacy, explanation and a human alternative. It was guidance, not law., protected attributeprotected attribute, n.: A personal trait, such as race, sex, religion, age or disability, that the law says must not be used to treat people unfairly., proxy discriminationproxy discrimination, n.: Discrimination that happens when a system cannot see a trait like race directly but relies on something closely linked to it, such as a postcode, and so treats groups unequally anyway., automated employment decision toolautomated employment decision tool, n.: Software that scores, ranks or screens job candidates or employees. New York City requires employers using such tools to have them checked for bias every year and to tell applicants., demographic paritydemographic parity, n.: A fairness test that asks whether a system says yes (approves the loan, shortlists the applicant) at the same rate for every group, whatever other differences exist between them., equal opportunityequal opportunity, n.: A fairness test that asks whether people who truly deserve a yes (who would repay the loan or succeed in the job) get one at the same rate in every group., group fairnessgroup fairness, n.: Measuring fairness by comparing how a system treats whole groups, such as women and men, on average: their approval rates, their error rates or what their scores mean.