# Capuchin: Causal Database Repair for Algorithmic Fairness

arXiv: https://arxiv.org/abs/1902.08283 · HTML: https://mlsanity.com/paper/1902.08283/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 21 Feb 2019.

## Built on by (most cited first, 6 of 6)

- [Explaining Black-Box Algorithms Using Probabilistic Contrastive Counterfactuals](https://mlsanity.com/paper/2103.11972/) (2021)
- [Marrying Fairness and Explainability in Supervised Learning](https://mlsanity.com/paper/2204.02947/) (2022)
- [Learning Counterfactually Invariant Predictors](https://mlsanity.com/paper/2207.09768/) (2022)
- [Consistent Range Approximation for Fair Predictive Modeling](https://mlsanity.com/paper/2212.10839/) (2022)
- [Towards Efficient Pareto-optimal Utility-Fairness between Groups in Repeated Rankings](https://mlsanity.com/paper/2402.14305/) (2024)
- [The Fair Game: Auditing & Debiasing AI Algorithms Over Time](https://mlsanity.com/paper/2508.06443/) (2025)

## Abstract (arXiv)

Fairness is increasingly recognized as a critical component of machine learning systems. However, it is the underlying data on which these systems are trained that often reflect discrimination, suggesting a database repair problem. Existing treatments of fairness rely on statistical correlations that can be fooled by statistical anomalies, such as Simpson's paradox. Proposals for causality-based definitions of fairness can correctly model some of these situations, but they require specification of the underlying causal models. In this paper, we formalize the situation as a database repair problem, proving sufficient conditions for fair classifiers in terms of admissible variables as opposed to a complete causal model. We show that these conditions correctly capture subtle fairness violations. We then use these conditions as the basis for database repair algorithms that provide provable fairness guarantees about classifiers trained on their training labels. We evaluate our algorithms on real data, demonstrating improvement over the state of the art on multiple fairness metrics proposed in the literature while retaining high utility.