Fair Prediction with Disparate Impact¶
Chouldechova, A. (2017). Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments. Big Data, 5(2), 153-163.
Cited by¶
2 citations across 2 artifacts.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Axiomatic Incompatibility
- In fairness, the Chouldechova and Kleinberg-Mullainathan-Raghavan results show calibration and the two balance conditions cannot all hold under unequal base rates.
This sourceShows calibration and the two error-rate balance conditions cannot all hold when base rates differ across groups.
SupportedVerified against the work's full text
Chouldechova derives that a test-fair (well-calibrated) score cannot equalise false positive and false negative rates across groups whose prevalence differs.
“when the recidivism prevalence differs between two groups, a test-fair score Sc cannot have equal false positive and negative rates across those groups.”
- In fairness, the Chouldechova and Kleinberg-Mullainathan-Raghavan results show calibration and the two balance conditions cannot all hold under unequal base rates.
- Equity
- Chouldechova (2017) shows the algorithmic analogue: fairness criteria appearing transparent and well-articulated can mask incompatibilities that surface only through case-level disparate impact analysis.
This sourceShows that calibration and error-rate balance (equal false-positive/false-negative rates) cannot be jointly satisfied when base rates differ across groups — an impossibility result.
Supported in partVerified against the work's full text
Chouldechova proves several fairness criteria are jointly unsatisfiable under unequal base rates, and shows disparate impact can follow; she does not frame this as criteria 'masking' flaws found only case by case.
“We demonstrate that the criteria cannot all be simultaneously satisfied when recidivism prevalence differs across groups.”
- Chouldechova (2017) shows the algorithmic analogue: fairness criteria appearing transparent and well-articulated can mask incompatibilities that surface only through case-level disparate impact analysis.
Verification¶
Does it exist? Confirmed. This work's DOI resolves to a registered record, which fixes its identity. That is all it fixes.
Does it back the claim? Read against the text for 2 of 2 citations: 1 supported, 1 supported in part. Each verdict is shown under its citation below, with what in the work backs the sentence.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Registry ID ref:979b51460706 · see in the full table