Diversity Recommendation via Causal Deconfounding of Co-purchase Relations and Counterfactual Exposure

2025-12-19

AI Summary

Existing recommendation methods rely on co-occurrence modeling, making them vulnerable to item popularity bias and user attribute confounding—leading to distorted embeddings, insufficient diversity, and a lack of causal interpretability. To address this, we propose Cadence, a causal deconfounding framework: (1) it constructs an unbiased asymmetric co-purchase graph (UACR) to disentangle popularity effects from genuine user preferences; (2) it introduces a counterfactual exposure mechanism to simulate latent interactions under high-exposure conditions, jointly optimizing recommendation accuracy and diversity. Cadence is the first work to systematically integrate causal inference into diversity-aware recommendation, with theoretically grounded identifiability and intrinsic interpretability. Built upon LightGCN, Cadence consistently outperforms state-of-the-art methods across multiple real-world datasets, achieving simultaneous improvements in Intra-List Diversity (ILD), Coverage, and Recall@K—while maintaining computational efficiency and strong cross-domain transferability.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Data Mining & Knowledge Management: Recommender Systems

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Beyond user-item modeling, item-to-item relationships are increasingly used to enhance recommendation. However, common methods largely rely on co-occurrence, making them prone to item popularity bias and user attributes, which degrades embedding quality and performance. Meanwhile, although diversity is acknowledged as a key aspect of recommendation quality, existing research offers limited attention to it, with a notable lack of causal perspectives and theoretical grounding. To address these challenges, we propose Cadence: Diversity Recommendation via Causal Deconfounding of Co-purchase Relations and Counterfactual Exposure - a plug-and-play framework built upon LightGCN as the backbone, primarily designed to enhance recommendation diversity while preserving accuracy. First, we compute the Unbiased Asymmetric Co-purchase Relationship (UACR) between items - excluding item popularity and user attributes - to construct a deconfounded directed item graph, with an aggregation mechanism to refine embeddings. Second, we leverage UACR to identify diverse categories of items that exhibit strong causal relevance to a user's interacted items but have not yet been engaged with. We then simulate their behavior under high-exposure scenarios, thereby significantly enhancing recommendation diversity while preserving relevance. Extensive experiments on real-world datasets demonstrate that our method consistently outperforms state-of-the-art diversity models in both diversity and accuracy, and further validates its effectiveness, transferability, and efficiency over baselines.

Problem

• Addresses item popularity bias in co-occurrence based recommendation systems
• Enhances recommendation diversity while maintaining accuracy through causal methods
• Proposes deconfounded item relationships to improve embedding quality and performance

Innovation

• Causal deconfounding removes popularity bias from item relationships
• Counterfactual exposure simulates user engagement with diverse items
• Plug-and-play framework enhances diversity while maintaining recommendation accuracy