Learning Adversarially Fair and Transferable Representations

2018/02/17 by David Madras, Elliot Creager, Madras, David +6 · 61 citations
Computer Science · Mathematics · Social Sciences · #Adversarial Robustness in Machine Learning #Ethics and Social Impacts of AI #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.1802.06309

openalex publication_date 2018/02/17 · arxiv created 2018/10/22 · arxiv updated 2018/10/23 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

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.

Citations

• Domain-Adversarial Neural Networks
• Equality of Opportunity in Supervised Learning
• Data Decisions and Theoretical Implications when Adversarially Learning Fair Representations
• A Kernel Method for the Two-Sample Problem
• On Fairness and Calibration

Cited by

• Personalized Pricing in Social Networks with Individual and Group Fairness Considerations
• Representation Invariance and Allocation: When Subgroup Balance Matters
• IFFair: Influence Function-driven Sample Reweighting for Fair Classification
• FairMT: Fairness for Heterogeneous Multi-Task Learning
• Model-Agnostic Fairness Regularization for GNNs with Incomplete Sensitive Information
• Prompt Fairness: Sub-group Disparities in LLMs
• Optimal Fairness under Local Differential Privacy
• Fairness in Multi-modal Medical Diagnosis with Demonstration Selection
• Learning Fair Representations with Kolmogorov-Arnold Networks
• FAST-CAD: A Fairness-Aware Framework for Non-Contact Stroke Diagnosis
• Budgeted Multiple-Expert Deferral
• Contrastive learning enhances fairness in pathology artificial intelligence systems
• Doubly-Regressing Approach for Subgroup Fairness
• Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference
• From Global to Local: Social Bias Transfer in CLIP
• On Optimal Steering to Achieve Exact Fairness
• Philosophy-informed Machine Learning
• DELTA: Variational Disentangled Learning for Privacy-Preserving Data Reprogramming
• Fairness for the People, by the People: Minority Collective Action
• Kernel-based Equalized Odds: A Quantification of Accuracy-Fairness Trade-off in Fair Representation Learning
• SoK: Data Minimization in Machine Learning
• FairDRL-ST: Disentangled Representation Learning for Fair Spatio-Temporal Mobility Prediction
• Algorithmic Fairness amid Social Determinants: Reflection, Characterization, and Approach
• Understanding-informed Bias Mitigation for Fair CMR Segmentation
• FADE: Adversarial Concept Erasure in Flow Models
• Incorporating Fairness Constraints into Archetypal Analysis
• Learning Private Representations through Entropy-based Adversarial Training
• Fair CCA for Fair Representation Learning: An ADNI Study
• Underrepresentation, Label Bias, and Proxies: Towards Data Bias Profiles for the EU AI Act and Beyond
• FairDiffuseVQVAE: Sampling-Time Fairness in Tabular Diffusion via Conditional Refinement of Vector-Quantized Latents
• Benchmarking Stochastic Approximation Algorithms for Fairness-Constrained Training of Deep Neural Networks
• Fairness Overfitting in Machine Learning: An Information-Theoretic Perspective
• Mastering Multiple-Expert Routing: Realizable H-Consistency and Strong Guarantees for Learning to Defer
• SaGIF: Improving Individual Fairness in Graph Neural Networks via Similarity Encoding
• How Many Domains Suffice for Domain Generalization? A Tight Characterization via the Domain Shattering Dimension
• Expressive Score-Based Priors for Distribution Matching with Geometry-Preserving Regularization
• PASS: Private Attributes Protection with Stochastic Data Substitution
• Compositional Fairness Constraints for Graph Embeddings
• The Frontiers of Fairness in Machine Learning
• Evaluating AI for Finance: Is AI Credible at Assessing Investment Risk?
• Conditional t-SNE: Complementary t-SNE embeddings through factoring out prior information
• Embracing Contradiction: Theoretical Inconsistency Will Not Impede the Road of Building Responsible AI Systems
• Learning Representational Disparities
• Training individually fair ML models with Sensitive Subspace Robustness
• On Adversarial Bias and the Robustness of Fair Machine Learning
• Fairness-aware Anomaly Detection via Fair Projection
• FairSHAP: Preprocessing for Fairness Through Attribution-Based Data Augmentation
• Simple and Effective Specialized Representations for Fair Classifiers
• Towards Fair In-Context Learning with Tabular Foundation Models
• Enabling Group Fairness in Graph Unlearning via Bi-level Debiasing
• Hurtful Words: Quantifying Biases in Clinical Contextual Word Embeddings
• Fair Representation Learning for Continuous Sensitive Attributes using Expectation of Integral Probability Metrics
• Learning Smooth and Fair Representations
• Metadata Normalization
• Augmented Fairness: An Interpretable Model Augmenting Decision-Makers' Fairness
• ReLU integral probability metric and its applications
• Mitigating Degree Bias in Graph Representation Learning with Learnable Structural Augmentation and Structural Self-Attention
• Fairness for Robust Log Loss Classification
• Estimating Principal Components under Adversarial Perturbations
• Fairness in TabNet Model by Disentangled Representation for the Prediction of Hospital No-Show
• FAIR-SIGHT: Fairness Assurance in Image Recognition via Simultaneous Conformal Thresholding and Dynamic Output Repair

Related

• Learning Fair and Transferable Representations
• Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference
• Learning Fair Representations with High-Confidence Guarantees
• Generating Fair Universal Representations using Adversarial Models
• Learning fair representation with a parametric integral probability metric