A Causal Bayesian Networks Viewpoint on Fairness
This work addresses fairness in machine learning models, particularly for risk assessment tools like COMPAS, by providing a causal framework to measure and mitigate unfairness, though it is incremental in applying existing causal methods to fairness.
The paper tackles the problem of unfairness in datasets by interpreting it as unfair causal paths in causal Bayesian networks, and demonstrates this approach by revisiting the COMPAS debate to show that fairness evaluation requires careful consideration of underlying unfairness patterns.
We offer a graphical interpretation of unfairness in a dataset as the presence of an unfair causal path in the causal Bayesian network representing the data-generation mechanism. We use this viewpoint to revisit the recent debate surrounding the COMPAS pretrial risk assessment tool and, more generally, to point out that fairness evaluation on a model requires careful considerations on the patterns of unfairness underlying the training data. We show that causal Bayesian networks provide us with a powerful tool to measure unfairness in a dataset and to design fair models in complex unfairness scenarios.