Search for: All records

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. 

We describe customized synthetic datasets for publishing mobility data. Private companies are providing new transportation modalities, and their data is of high value for integrative transportation research, policy enforcement, and public accountability. However, these companies are disincentivized from sharing data not only to protect the privacy of individuals (drivers and/or passengers), but also to protect their own competitive advantage. Moreover, demographic biases arising from how the services are delivered may be amplified if released data is used in other contexts. We describe a model and algorithm for releasing origin-destination histograms that removes selected biases in the data using causality-based methods. We compute the origin-destination histogram of the original dataset then adjust the counts to remove undesirable causal relationships that can lead to discrimination or violate contractual obligations with data owners. We evaluate the utility of the algorithm on real data from a dockless bike share program in Seattle and taxi data in New York, and show that these adjusted transportation datasets can retain utility while removing bias in the underlying data. 

Data too sensitive to be "open" for analysis and re-purposing typically remains "closed" as proprietary information. This dichotomy undermines efforts to make algorithmic systems more fair, transparent, and accountable. Access to proprietary data in particular is needed by government agencies to enforce policy, researchers to evaluate methods, and the public to hold agencies accountable; all of these needs must be met while preserving individual privacy and firm competitiveness. In this paper, we describe an integrated legal-technical approach provided by a third-party public-private data trust designed to balance these competing interests. Basic membership allows firms and agencies to enable low-risk access to data for compliance reporting and core methods research, while modular data sharing agreements support a wide array of projects and use cases. Unless specifically stated otherwise in an agreement, all data access is initially provided to end users through customized synthetic datasets that offer a) strong privacy guarantees, b) removal of signals that could expose competitive advantage, and c) removal of biases that could reinforce discriminatory policies, all while maintaining fidelity to the original data. We find that using synthetic data in conjunction with strong legal protections over raw data strikes a balance between transparency, proprietorship, privacy, and research objectives. This legal-technical framework can form the basis for data trusts in a variety of contexts.