Learning Fair Representations via Rate-Distortion Maximization
This addresses fairness issues in AI for applications like hiring or lending, but it is incremental as it builds on existing debiasing techniques.
The paper tackles the problem of biased decisions from machine learning models due to demographic information encoded in text representations, presenting Fairness-aware Rate Maximization (FaRM) which removes protected information by making representations uncorrelated within attribute classes, achieving state-of-the-art performance and significantly reducing information leakage in evaluations.
Text representations learned by machine learning models often encode undesirable demographic information of the user. Predictive models based on these representations can rely on such information, resulting in biased decisions. We present a novel debiasing technique, Fairness-aware Rate Maximization (FaRM), that removes protected information by making representations of instances belonging to the same protected attribute class uncorrelated, using the rate-distortion function. FaRM is able to debias representations with or without a target task at hand. FaRM can also be adapted to remove information about multiple protected attributes simultaneously. Empirical evaluations show that FaRM achieves state-of-the-art performance on several datasets, and learned representations leak significantly less protected attribute information against an attack by a non-linear probing network.