Mind the Graph When Balancing Data for Fairness or Robustness
This work addresses fairness and robustness failures in machine learning for practitioners, highlighting that data balancing can be ineffective or harmful without considering causal structure.
The paper investigates conditions under which data balancing leads to fair or robust models, finding that balanced distributions often fail to selectively remove undesired dependencies in causal graphs, resulting in failure modes and interference with other techniques like regularization.
Failures of fairness or robustness in machine learning predictive settings can be due to undesired dependencies between covariates, outcomes and auxiliary factors of variation. A common strategy to mitigate these failures is data balancing, which attempts to remove those undesired dependencies. In this work, we define conditions on the training distribution for data balancing to lead to fair or robust models. Our results display that, in many cases, the balanced distribution does not correspond to selectively removing the undesired dependencies in a causal graph of the task, leading to multiple failure modes and even interference with other mitigation techniques such as regularization. Overall, our results highlight the importance of taking the causal graph into account before performing data balancing.