Showing 1–1 of 1 results for author: Sabri, N

Consistent Range Approximation for Fair Predictive Modeling

Authors: Jiongli Zhu, Sainyam Galhotra, Nazanin Sabri, Babak Salimi

Abstract: This paper proposes a novel framework for certifying the fairness of predictive models trained on biased data. It draws from query answering for incomplete and inconsistent databases to formulate the problem of consistent range approximation (CRA) of fairness queries for a predictive model on a target population. The framework employs background knowledge of the data collection process and biased… ▽ More This paper proposes a novel framework for certifying the fairness of predictive models trained on biased data. It draws from query answering for incomplete and inconsistent databases to formulate the problem of consistent range approximation (CRA) of fairness queries for a predictive model on a target population. The framework employs background knowledge of the data collection process and biased data, working with or without limited statistics about the target population, to compute a range of answers for fairness queries. Using CRA, the framework builds predictive models that are certifiably fair on the target population, regardless of the availability of external data during training. The framework's efficacy is demonstrated through evaluations on real data, showing substantial improvement over existing state-of-the-art methods. △ Less

Submitted 28 July, 2023; v1 submitted 21 December, 2022; originally announced December 2022.