Improving Group Fairness in Tensor Completion via Imbalance Mitigating Entity Augmentation

2025-07-28

AI Summary

This paper addresses group fairness in tensor completion, where data sparsity and bias across sensitive attributes (e.g., gender, age) degrade equitable performance across demographic groups. To mitigate this, we propose STAFF—a Structured Tensor Augmentation Framework for Fairness—which leverages well-observed entity pairs to structurally enrich sparse tensors, enabling the first joint optimization of fairness and accuracy. STAFF is model-agnostic, seamlessly integrating with both classical and deep learning–based tensor completion methods, and supports end-to-end fair-aware completion. Experiments on multiple real-world datasets demonstrate that STAFF reduces MSE by up to 36% and MADE by up to 59% compared to the best prior baseline, achieving superior trade-offs between prediction accuracy and inter-group error parity. It significantly outperforms existing fair tensor completion approaches in both overall reconstruction quality and group-level fairness guarantees.

Technology Category

• Machine Learning: Matrix & Tensor Methods
• Computer Vision: Bias, Fairness & Privacy
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness and ethical considerations in crowd work and in human-in-the-loop AI systems
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

Group fairness is important to consider in tensor decomposition to prevent discrimination based on social grounds such as gender or age. Although few works have studied group fairness in tensor decomposition, they suffer from performance degradation. To address this, we propose STAFF(Sparse Tensor Augmentation For Fairness) to improve group fairness by minimizing the gap in completion errors of different groups while reducing the overall tensor completion error. Our main idea is to augment a tensor with augmented entities including sufficient observed entries to mitigate imbalance and group bias in the sparse tensor. We evaluate method on tensor completion with various datasets under conventional and deep learning-based tensor models. STAFF consistently shows the best trade-off between completion error and group fairness; at most, it yields 36% lower MSE and 59% lower MADE than the second-best baseline.

Problem

• Mitigating group bias in tensor completion
• Reducing completion error gaps between groups
• Balancing fairness and accuracy in sparse tensors

Innovation

• Augment tensor with entities to mitigate imbalance
• Minimize completion error gaps between groups
• Improve fairness and reduce overall error