Society, Ethics & EconomyMachine LearningMathematics & Statistics

counterfactual fairness

nounAlso: causal fairness

In plain wordsA fairness test that asks: had this person been of another race or sex, with everything that would have followed from that, would the system have decided the same way?

A causal fairness criterion requiring that a prediction about an individual would not change if, in an assumed structural causal model, their 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. were counterfactually set to another value while the background (exogenous) factors U inferred for that individual stayed fixed. Because the counterfactual change propagates to every causal descendant of the attribute, a counterfactually fair predictor may use only features not influenced by it, or adjusted versions of them. The criterion makes fairness assumptions explicit, but it depends on a causal graph that is rarely known with confidence, and some scholars argue that counterfactuals about race or sex are ill-defined.
Under counterfactual fairness, the admissions score cannot rely on test results that are themselves shaped by the applicant's race.

P\big(\hat{Y}_{A \leftarrow a}(U) = y \mid X = x, A = a\big) = P\big(\hat{Y}_{A \leftarrow a'}(U) = y \mid X = x, A = a\big)

Origin

Introduced by Matt Kusner, Joshua Loftus, Chris Russell and Ricardo Silva in 'Counterfactual Fairness' (NeurIPS 2017), building on Judea Pearl's structural causal models. Related causal criteria appeared the same year, including path-specific approaches by Razieh Nabi and Ilya Shpitser and by Niki Kilbertus and colleagues.

Sources

Kusner, M. J., Loftus, J., Russell, C. & Silva, R. (2017). Counterfactual Fairness. Advances in Neural Information Processing Systems 30.

Mentioned in

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., causal inferencecausal inference, n.: Working out whether one thing actually causes another, rather than merely occurring alongside it — for example, whether a new feature made users stay longer or just happened at the same time., counterfactualcounterfactual, n.: A statement about what would have happened if something had been different — 'had I left earlier, I would have caught the train'. AI uses counterfactuals to explain decisions and to measure causes., counterfactual explanationcounterfactual explanation, n.: An explanation that tells someone the smallest change that would have produced a different outcome — for example, that the application would have been accepted if it had listed two more years of work experience., fairness through unawarenessfairness through unawareness, n.: The simple idea of making a system fair by hiding sensitive traits like race or sex from it; it usually fails, because other information gives those traits away., individual fairnessindividual fairness, n.: The idea that a fair system should treat similar people similarly: two applicants who are alike in every way that matters should get similar outcomes., structural causal modelstructural causal model, n.: A precise description of what causes what in a system — a diagram of arrows plus equations — that predicts what would happen if you intervened, not just what tends to occur together.