Fairness Through Computationally-Bounded Awareness

Abstract

We study the problem of fair classification within the versatile framework of Dwork et al. [ITCS '12], which assumes the existence of a metric that measures similarity between pairs of individuals. Unlike earlier work, we do not assume that the entire metric is known to the learning algorithm; instead, the learner can query this arbitrary metric a bounded number of times. We propose a new notion of fairness called metric multifairness and show how to achieve this notion in our setting. Metric multifairness is parameterized by a similarity metric $d$ on pairs of individuals to classify and a rich collection $\mathcal{C}$ of (possibly overlapping) "comparison sets" over pairs of individuals. At a high level, metric multifairness guarantees that similar subpopulations are treated similarly, as long as these subpopulations are identified within the class $\mathcal{C}$.

Appears at NeurIPS 2018

Cited by in corpus (29)

• Prediction-Based Decisions and Fairness: A Catalogue of Choices, Assumptions, and Definitions
• Fairness in Machine Learning: A Survey
• Fairness in Machine Learning
• Aligning AI With Shared Human Values
• Online Learning with an Unknown Fairness Metric
• Machine learning fairness notions: Bridging the gap with real-world applications
• Calibration for the (Computationally-Identifiable) Masses
• Fairness without Demographics through Adversarially Reweighted Learning
• Probably Approximately Metric-Fair Learning
• Noise-tolerant fair classification
• Average Individual Fairness: Algorithms, Generalization and Experiments
• SenSeI: Sensitive Set Invariance for Enforcing Individual Fairness
• Kernel Dependence Regularizers and Gaussian Processes with Applications to Algorithmic Fairness
• Moment Multicalibration for Uncertainty Estimation
• Modeling Techniques for Machine Learning Fairness: A Survey
• Sample Complexity of Uniform Convergence for Multicalibration
• Beyond "Fairness:" Structural (In)justice Lenses on AI for Education
• Preference-Informed Fairness
• The Random Conditional Distribution for Higher-Order Probabilistic Inference
• Online Multivalid Learning: Means, Moments, and Prediction Intervals
• Learning Smooth and Fair Representations
• Defining AI in Policy versus Practice
• Beyond traditional assumptions in fair machine learning
• Individually Fair Gradient Boosting
• On the Identification of Fair Auditors to Evaluate Recommender Systems based on a Novel Non-Comparative Fairness Notion
• Legally grounded fairness objectives
• Outcome Indistinguishability
• A Sociotechnical View of Algorithmic Fairness
• Non-Comparative Fairness for Human-Auditing and Its Relation to Traditional Fairness Notions