Trade-offs Between Individual and Group Fairness in Machine Learning: A Comprehensive Review

TL;DR

This survey confronts the joint treatment of group fairness (GF) and individual fairness (IF) in machine learning, highlighting the theoretical incompatibility of strict GF and IF and surveying a range of hybrid approaches that explicitly manage their trade-offs. It introduces a workflow- and mechanism-based taxonomy that covers data transformation, reweighting, representation learning, data augmentation, regularization, metric-based constraints, and adversarial methods, organized by how they implement GF–IF objectives. The authors provide theoretical and empirical syntheses, including benchmark datasets, computational considerations, and notes on reproducibility, to illuminate practical constraints and decision-making when designing hybrid fairness systems. They also discuss open directions, such as more diverse evaluations, causal and normative grounding, and cost-aware deployment, to advance principled, context-aware GF–IF methods with reliable guarantees at both the group and individual levels.

Abstract

Algorithmic fairness has become a central concern in computational decision-making systems, where ensuring equitable outcomes is essential for both ethical and legal reasons. Two dominant notions of fairness have emerged in the literature: Group Fairness (GF), which focuses on mitigating disparities across demographic subpopulations, and Individual Fairness (IF), which emphasizes consistent treatment of similar individuals. These notions have traditionally been studied in isolation. In contrast, this survey examines methods that jointly address GF and IF, integrating both perspectives within unified frameworks and explicitly characterizing the trade-offs between them. We provide a systematic and critical review of hybrid fairness approaches, organizing existing methods according to the fairness mechanisms they employ and the algorithmic and mathematical strategies used to reconcile multiple fairness criteria. For each class of methods, we examine their theoretical foundations, optimization mechanisms, and empirical evaluation practices, and discuss their limitations. Additionally, we discuss the challenges and identify open research directions for developing principled, context-aware hybrid fairness methods. By synthesizing insights across the literature, this survey aims to serve as a comprehensive resource for researchers and practitioners seeking to design hybrid algorithms that provide reliable fairness guarantees at both the individual and group levels.

Figures (3)

• Figure 1: Trade-off between group and individual fairness.Left: observations from two groups in feature space $X$, where edges indicate similar individuals (using $k$-nearest neighbors) who should receive similar decisions (individual fairness). A single global threshold (gray radius) maximizes local consistency but yields different acceptance rates across groups. Group-specific thresholds (dashed radii) enforce Independence (equal acceptance rates) but break consistency across similar neighbors (red edges). Right: the same thresholds shown in model-score space $f(X)$, where differing score distributions make these fairness criteria incompatible.
• Figure 2: Number of papers identified per year through the literature search process. Panel (a) shows results excluding arXiv, while panel (b) includes arXiv preprints.
• Figure 3: Summary statistics of papers addressing the GF vs IF trade-off. (A) Dataset distribution, showing heavy reliance on a few benchmark datasets (e.g., Adult, COMPAS). (B) Distribution of methods by processing stage, with in-processing approaches being the most common. (C) Proportion of papers within each sub-taxonomy, where regularization is the most common approach.