Learning Adversarially Fair and Transferable Representations

Abstract

In this paper, we advocate for representation learning as the key to mitigating unfair prediction outcomes downstream. Motivated by a scenario where learned representations are used by third parties with unknown objectives, we propose and explore adversarial representation learning as a natural method of ensuring those parties act fairly. We connect group fairness (demographic parity, equalized odds, and equal opportunity) to different adversarial objectives. Through worst-case theoretical guarantees and experimental validation, we show that the choice of this objective is crucial to fair prediction. Furthermore, we present the first in-depth experimental demonstration of fair transfer learning and demonstrate empirically that our learned representations admit fair predictions on new tasks while maintaining utility, an essential goal of fair representation learning.

References in corpus (10)

• Equality of Opportunity in Supervised Learning
• Semi-Supervised Learning with Generative Adversarial Networks
• Semantic Segmentation using Adversarial Networks
• Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairness
• Data Decisions and Theoretical Implications when Adversarially Learning Fair Representations
• On Fairness and Calibration
• A Kernel Method for the Two-Sample Problem
• Domain-Adversarial Neural Networks
• Optimized Data Pre-Processing for Discrimination Prevention
• Provably Fair Representations

Cited by in corpus (70)

• Fairness in Machine Learning
• A Review on Generative Adversarial Networks: Algorithms, Theory, and Applications
• The Frontiers of Fairness in Machine Learning
• Compositional Fairness Constraints for Graph Embeddings
• A Comprehensive Survey on Trustworthy Graph Neural Networks: Privacy, Robustness, Fairness, and Explainability
• An analysis on the use of autoencoders for representation learning: fundamentals, learning task case studies, explainability and challenges
• Large-Scale Long-Tailed Recognition in an Open World
• Towards Equity and Algorithmic Fairness in Student Grade Prediction
• Training individually fair ML models with Sensitive Subspace Robustness
• Fairness for Robust Log Loss Classification
• FairGAN: Fairness-aware Generative Adversarial Networks
• Unbiased Graph Embedding with Biased Graph Observations
• On Adversarial Bias and the Robustness of Fair Machine Learning
• Conditional Generation of Medical Images via Disentangled Adversarial Inference
• What-is and How-to for Fairness in Machine Learning: A Survey, Reflection, and Perspective
• Hurtful Words: Quantifying Biases in Clinical Contextual Word Embeddings
• Improved Adversarial Learning for Fair Classification
• Putting Fairness Principles into Practice: Challenges, Metrics, and Improvements
• Transfer of Machine Learning Fairness across Domains
• Two Simple Ways to Learn Individual Fairness Metrics from Data
• DeBayes: a Bayesian Method for Debiasing Network Embeddings
• Review of Mathematical frameworks for Fairness in Machine Learning
• Adversarial training approach for local data debiasing
• Tuning Fairness by Balancing Target Labels
• SenSeI: Sensitive Set Invariance for Enforcing Individual Fairness
• Rényi Fair Inference
• Say No to the Discrimination: Learning Fair Graph Neural Networks with Limited Sensitive Attribute Information
• Same-Day Delivery with Fairness
• Layer-wise Characterization of Latent Information Leakage in Federated Learning
• Representation via Representations: Domain Generalization via Adversarially Learned Invariant Representations
• Fundamental Limits and Tradeoffs in Invariant Representation Learning
• Contrastive Examples for Addressing the Tyranny of the Majority
• Ethical Adversaries: Towards Mitigating Unfairness with Adversarial Machine Learning
• On the Fairness of Disentangled Representations
• Causally motivated Shortcut Removal Using Auxiliary Labels
• Learning Fair and Transferable Representations
• On the Privacy Risks of Algorithmic Fairness
• A non-discriminatory approach to ethical deep learning
• Improving Fairness of AI Systems with Lossless De-biasing
• Toward a better trade-off between performance and fairness with kernel-based distribution matching
• Technical Challenges for Training Fair Neural Networks
• Dynamic Modeling and Equilibria in Fair Decision Making
• Understanding and Mitigating Accuracy Disparity in Regression
• Augmented Fairness: An Interpretable Model Augmenting Decision-Makers' Fairness
• InfoFair: Information-Theoretic Intersectional Fairness
• State of the Art in Fair ML: From Moral Philosophy and Legislation to Fair Classifiers
• Distributed generation of privacy preserving data with user customization
• Removing Spurious Features can Hurt Accuracy and Affect Groups Disproportionately
• Fairness in TabNet Model by Disentangled Representation for the Prediction of Hospital No-Show
• FairNN- Conjoint Learning of Fair Representations for Fair Decisions
• Learning Smooth and Fair Representations
• Conditional t-SNE: Complementary t-SNE embeddings through factoring out prior information
• Fair Hate Speech Detection through Evaluation of Social Group Counterfactuals
• Metadata Normalization
• Generating private data with user customization
• Additive Adversarial Learning for Unbiased Authentication
• A Primal-Dual Subgradient Approachfor Fair Meta Learning
• Improving Counterfactual Generation for Fair Hate Speech Detection
• Towards Model-Agnostic Post-Hoc Adjustment for Balancing Ranking Fairness and Algorithm Utility
• Individual Fairness Revisited: Transferring Techniques from Adversarial Robustness
• Event-Related Bias Removal for Real-time Disaster Events
• Fair for All: Best-effort Fairness Guarantees for Classification
• BriarPatches: Pixel-Space Interventions for Inducing Demographic Parity
• Hidden Technical Debts for Fair Machine Learning in Financial Services
• Convex-Concave Min-Max Stackelberg Games
• Deconfounding age effects with fair representation learning when assessing dementia
• Invariant Representations with Stochastically Quantized Neural Networks
• Estimating Principal Components under Adversarial Perturbations
• Constrained Non-Affine Alignment of Embeddings
• Practical Compositional Fairness: Understanding Fairness in Multi-Component Recommender Systems