Recent work in fairness in machine learning has proposed adjusting for fairness by equalizing accuracy metrics across groups and has also studied how datasets affected by historical prejudices may lead to unfair decision policies. We connect these lines of work and study the residual unfairness that arises when a fairness-adjusted predictor is not actually fair on the target population due to systematic censoring of training data by existing biased policies. This scenario is particularly common in the same applications where fairness is a concern. We characterize theoretically the impact of such censoring on standard fairness metrics for binary classifiers and provide criteria for when residual unfairness may or may not appear. We prove that, under certain conditions, fairness-adjusted classifiers will in fact induce residual unfairness that perpetuates the same injustices, against the same groups, that biased the data to begin with, thus showing that even state-of-the-art fair machine learning can have a "bias in, bias out" property. When certain benchmark data is available, we show how sample reweighting can estimate and adjust fairness metrics while accounting for censoring. We use this to study the case of Stop, Question, and Frisk (SQF) and demonstrate that attempting to adjust for fairness perpetuatesmore »
Robust Fairness under Covariate Shift.
Making predictions that are fair with regard to protected attributes (race, gender, age, etc.) has become an important requirement for classification algorithms. Existing techniques derive a fair model from sampled labeled data relying on the assumption that training and testing data are identically and independently drawn (iid) from the same distribution. In practice, distribution shift can and does occur between training and testing datasets as the characteristics of individuals interacting with the machine learning system change. We investigate fairness under covariate shift, a relaxation of the iid assumption in which the inputs or covariates change while the conditional label distribution remains the same. We seek fair decisions under these assumptions on target data with unknown labels. We propose an approach that obtains the predictor that is robust to the worst-case testing performance while satisfying target fairness requirements and matching statistical properties of the source data. We demonstrate the benefits of our approach on benchmark prediction tasks.
- Award ID(s):
- Publication Date:
- NSF-PAR ID:
- Journal Name:
- Proceedings of the AAAI Conference on Artificial Intelligence
- Sponsoring Org:
- National Science Foundation
More Like this
Graph is a ubiquitous type of data that appears in many real-world applications, including social network analysis, recommendations and financial security. Important as it is, decades of research have developed plentiful computational models to mine graphs. Despite its prosperity, concerns with respect to the potential algorithmic discrimination have been grown recently. Algorithmic fairness on graphs, which aims to mitigate bias introduced or amplified during the graph mining process, is an attractive yet challenging research topic. The first challenge corresponds to the theoretical challenge, where the non-IID nature of graph data may not only invalidate the basic assumption behind many existing studies in fair machine learning, but also introduce new fairness definition(s) based on the inter-correlation between nodes rather than the existing fairness definition(s) in fair machine learning. The second challenge regarding its algorithmic aspect aims to understand how to balance the trade-off between model accuracy and fairness. This tutorial aims to (1) comprehensively review the state-of-the-art techniques to enforce algorithmic fairness on graphs and (2) enlighten the open challenges and future directions. We believe this tutorial could benefit researchers and practitioners from the areas of data mining, artificial intelligence and social science.
The assumption that training and testing samples are generated from the same distribution does not always hold for real-world machine-learning applications. The procedure of tackling this discrepancy between the training (source) and testing (target) domains is known as domain adaptation. We propose an unsupervised version of domain adaptation that considers the presence of only unlabelled data in the target domain. Our approach centres on finding correspondences between samples of each domain. The correspondences are obtained by treating the source and target samples as graphs and using a convex criterion to match them. The criteria used are first-order and second-order similarities between the graphs as well as a class-based regularization. We have also developed a computationally efficient routine for the convex optimization, thus allowing the proposed method to be used widely. To verify the effectiveness of the proposed method, computer simulations were conducted on synthetic, image classification and sentiment classification datasets. Results validated that the proposed local sample-to- sample matching method out-performs traditional moment-matching methods and is competitive with respect to current local domain-adaptation methods.
In today’s increasingly connected world, graph mining plays a piv- otal role in many real-world application domains, including social network analysis, recommendations, marketing and financial secu- rity. Tremendous efforts have been made to develop a wide range of computational models. However, recent studies have revealed that many widely-applied graph mining models could suffer from potential discrimination. Fairness on graph mining aims to develop strategies in order to mitigate bias introduced/amplified during the mining process. The unique challenges of enforcing fairness on graph mining include (1) theoretical challenge on non-IID nature of graph data, which may invalidate the basic assumption behind many existing studies in fair machine learning, and (2) algorith- mic challenge on the dilemma of balancing model accuracy and fairness. This tutorial aims to (1) present a comprehensive review of state-of-the-art techniques in fairness on graph mining and (2) identify the open challenges and future trends. In particular, we start with reviewing the background, problem definitions, unique challenges and related problems; then we will focus on an in-depth overview of (1) recent techniques in enforcing group fairness, indi- vidual fairness and other fairness notions in the context of graph mining, and (2) future directions in studying algorithmic fairness onmore »
Motivated by concerns surrounding the fairness effects of sharing and transferring fair machine learning tools, we propose two algorithms: Fairness Warnings and Fair-MAML. The first is a model-agnostic algorithm that provides interpretable boundary conditions for when a fairly trained model may not behave fairly on similar but slightly different tasks within a given domain. The second is a fair meta-learning approach to train models that can be quickly fine-tuned to specific tasks from only a few number of sample instances while balancing fairness and accuracy. We demonstrate experimentally the individual utility of each model using relevant baselines and provide the first experiment to our knowledge of K-shot fairness, i.e. training a fair model on a new task with only K data points. Then, we illustrate the usefulness of both algorithms as a combined method for training models from a few data points on new tasks while using Fairness Warnings as interpretable boundary conditions under which the newly trained model may not be fair.