skip to main content

Title: Recovering from Biased Data: Can Fairness Constraints Improve Accuracy?
Multiple fairness constraints have been proposed in the literature, motivated by a range of concerns about how demographic groups might be treated unfairly by machine learning classifiers. In this work we consider a different motivation; learning from biased training data. We posit several ways in which training data may be biased, including having a more noisy or negatively biased labeling process on members of a disadvantaged group, or a decreased prevalence of positive or negative examples from the disadvantaged group, or both. Given such biased training data, Empirical Risk Minimization (ERM) may produce a classifier that not only is biased but also has suboptimal accuracy on the true data distribution. We examine the ability of fairness-constrained ERM to correct this problem. In particular, we find that the Equal Opportunity fairness constraint [Hardt et al., 2016] combined with ERM will provably recover the Bayes optimal classifier under a range of bias models. We also consider other recovery methods including re-weighting the training data, Equalized Odds, and Demographic Parity, and Calibration. These theoretical results provide additional motivation for considering fairness interventions even if an actor cares primarily about accuracy.
Award ID(s):
1815011 1733556
Publication Date:
Journal Name:
Symposium on Foundations of Responsible Computing (FORC)
Page Range or eLocation-ID:
Sponsoring Org:
National Science Foundation
More Like this
  1. As machine learning becomes more widely adopted across domains, it is critical that researchers and ML engineers think about the inherent biases in the data that may be perpetuated by the model. Recently, many studies have shown that such biases are also imbibed in Graph Neural Network (GNN) models if the input graph is biased, potentially to the disadvantage of underserved and underrepresented communities. In this work, we aim to mitigate the bias learned by GNNs by jointly optimizing two different loss functions: one for the task of link prediction and one for the task of demographic parity. We further implement three different techniques inspired by graph modification approaches: the Global Fairness Optimization (GFO), Constrained Fairness Optimization (CFO), and Fair Edge Weighting (FEW) models. These techniques mimic the effects of changing underlying graph structures within the GNN and offer a greater degree of interpretability over more integrated neural network methods. Our proposed models emulate microscopic or macroscopic edits to the input graph while training GNNs and learn node embeddings that are both accurate and fair under the context of link recommendations. We demonstrate the effectiveness of our approach on four real world datasets and show that we can improve themore »recommendation fairness by several factors at negligible cost to link prediction accuracy.« less
  2. We study fairness in supervised few-shot meta-learning models that are sensitive to discrimination (or bias) in historical data. A machine learning model trained based on biased data tends to make unfair predictions for users from minority groups. Although this problem has been studied before, existing methods mainly aim to detect and control the dependency effect of the protected variables (e.g. race, gender) on target prediction based on a large amount of training data. These approaches carry two major drawbacks that (1) lacking showing a global cause-effect visualization for all variables; (2) lacking generalization of both accuracy and fairness to unseen tasks. In this work, we first discover discrimination from data using a causal Bayesian knowledge graph which not only demonstrates the dependency of the protected variable on target but also indicates causal effects between all variables. Next, we develop a novel algorithm based on risk difference in order to quantify the discriminatory influence for each protected variable in the graph. Furthermore, to protect prediction from unfairness, a the fast-adapted bias-control approach in meta-learning is proposed, which efficiently mitigates statistical disparity for each task and it thus ensures independence of protected attributes on predictions based on biased and few-shot data samples.more »Distinct from existing meta-learning models, group unfairness of tasks are efficiently reduced by leveraging the mean difference between (un)protected groups for regression problems. Through extensive experiments on both synthetic and real-world data sets, we demonstrate that our proposed unfairness discovery and prevention approaches efficiently detect discrimination and mitigate biases on model output as well as generalize both accuracy and fairness to unseen tasks with a limited amount of training samples.« less
  3. Given significant concerns about fairness and bias in the use of artificial intelligence (AI) and machine learning (ML) for psychological assessment, we provide a conceptual framework for investigating and mitigating machine-learning measurement bias (MLMB) from a psychometric perspective. MLMB is defined as differential functioning of the trained ML model between subgroups. MLMB manifests empirically when a trained ML model produces different predicted score levels for different subgroups (e.g., race, gender) despite them having the same ground-truth levels for the underlying construct of interest (e.g., personality) and/or when the model yields differential predictive accuracies across the subgroups. Because the development of ML models involves both data and algorithms, both biased data and algorithm-training bias are potential sources of MLMB. Data bias can occur in the form of nonequivalence between subgroups in the ground truth, platform-based construct, behavioral expression, and/or feature computing. Algorithm-training bias can occur when algorithms are developed with nonequivalence in the relation between extracted features and ground truth (i.e., algorithm features are differentially used, weighted, or transformed between subgroups). We explain how these potential sources of bias may manifest during ML model development and share initial ideas for mitigating them, including recognizing that new statistical and algorithmic procedures needmore »to be developed. We also discuss how this framework clarifies MLMB but does not reduce the complexity of the issue.« less
  4. Link prediction has been widely applied in social network analysis. Despite its importance, link prediction algorithms can be biased by disfavoring the links between individuals in particular demographic groups. In this paper, we study one particular type of bias, namely, the bias in predicting inter-group links (i.e., links across different demographic groups). First, we formalize the definition of bias in link prediction by providing quantitative measurements of accuracy disparity, which measures the difference in prediction accuracy of inter-group and intra-group links. Second, we unveil the existence of bias in six existing state-of-the-art link prediction algorithms through extensive empirical studies over real world datasets. Third, we identify the imbalanced density across intra-group and inter-group links in training graphs as one of the underlying causes of bias in link prediction. Based on the identified cause, fourth, we design a pre-processing bias mitigation method named FairLP to modify the training graph, aiming to balance the distribution of intra-group and inter-group links while preserving the network characteristics of the graph. FairLP is model-agnostic and thus is compatible with any existing link prediction algorithm. Our experimental results on real-world social network graphs demonstrate that FairLP achieves better trade-off between fairness and prediction accuracy than themore »existing fairness-enhancing link prediction methods.« less
  5. 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 »the same injustices that the policy is infamous for.« less