Predict Responsibly: Increasing Fairness by Learning to Defer

Anonymous

Nov 03, 2017 (modified: Nov 03, 2017) ICLR 2018 Conference Blind Submission readers: everyone Show Bibtex
  • Abstract: Machine learning systems, which are often used for high-stakes decisions, typically suffer from two mutually reinforcing problems: unfairness and opaqueness. Many popular models, although generally accurate, cannot accurately express uncertainty about their predictions. Even in regimes where a model is fallible, users may trust the model’s predictions too fully, and allow its biases to reinforce the user’s own. In this work, we explore models that learn to defer. In our scheme, a model learns to classify accurately and fairly, but also to defer if necessary, passing judgment to a downstream decision-maker such as a human user. We propose a learning algorithm which accounts for potential biases held by decision-makers later in a pipeline. Experiments on real-world datasets demonstrate that learning to defer can make a model not only more accurate but also less biased. Even when operated by highly biased users, we show that deferring models can still greatly improve the fairness of the entire pipeline.
  • TL;DR: Incorporating the ability to say I-don't-know can improve the fairness of a classifier without sacrificing too much accuracy, and this improvement magnifies when the classifier has insight into downstream decision-making.
  • Keywords: Fairness, IDK, Calibration, Automated decision-making, Transparency, Accountability

Loading