Counterfactual Multi-player Bandits for Explainable Recommendation Diversification

2025-05-27

AI Summary

Recommendation systems often suffer from “filter bubbles,” and existing diversity-enhancement methods face two key bottlenecks: weak interpretability and difficulty optimizing non-differentiable diversity metrics. To address these, we propose the first framework integrating counterfactual causal inference with multi-player multi-armed bandits, unifying diversity attribution interpretability and optimization generality. Our approach employs counterfactual modeling to identify critical diversity-influencing factors, designs diversity-sensitive rewards, and enables online collaborative optimization of multiple players’ policies—supporting arbitrary (including non-differentiable) diversity measures for the first time. Evaluated on three real-world datasets, our method improves diversity metrics—including ILD and Gini+—by 12.6%–28.3%. Crucially, it delivers clear, attributable explanations for diversity-aware decisions, achieving both strong interpretability and cross-metric generalization.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments

Abstract

Existing recommender systems tend to prioritize items closely aligned with users' historical interactions, inevitably trapping users in the dilemma of ``filter bubble''. Recent efforts are dedicated to improving the diversity of recommendations. However, they mainly suffer from two major issues: 1) a lack of explainability, making it difficult for the system designers to understand how diverse recommendations are generated, and 2) limitations to specific metrics, with difficulty in enhancing non-differentiable diversity metrics. To this end, we propose a extbf{C}ounterfactual extbf{M}ulti-player extbf{B}andits (CMB) method to deliver explainable recommendation diversification across a wide range of diversity metrics. Leveraging a counterfactual framework, our method identifies the factors influencing diversity outcomes. Meanwhile, we adopt the multi-player bandits to optimize the counterfactual optimization objective, making it adaptable to both differentiable and non-differentiable diversity metrics. Extensive experiments conducted on three real-world datasets demonstrate the applicability, effectiveness, and explainability of the proposed CMB.

Problem

• Overcoming filter bubble in recommender systems
• Enhancing explainability of diverse recommendations
• Optimizing diverse metrics including non-differentiable ones

Innovation

• Counterfactual framework identifies diversity factors
• Multi-player bandits optimize diversity metrics
• Adaptable to differentiable and non-differentiable metrics