Fidelity-Aware Recommendation Explanations via Stochastic Path Integration

2025-11-22

AI Summary

Insufficient explanation fidelity in recommender systems—existing methods fail to accurately reflect the model’s true reasoning process, particularly exhibiting instability under sparse implicit feedback. This paper proposes SPINRec, a fidelity-oriented explanation framework grounded in stochastic path integration. Its core innovations include: (i) an empirical-distribution-driven stochastic baseline sampling strategy that replaces fixed baselines, enabling personalized and robust explanation paths; and (ii) the first unified fidelity evaluation protocol across multiple models (MF, VAE, NCF) and datasets, integrating AUC-based perturbation curves with fixed-length diagnostic metrics. Extensive experiments demonstrate that SPINRec significantly outperforms state-of-the-art methods on three benchmark datasets, establishing a new standard for recommendation explainability. The implementation is publicly available.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Data Mining & Knowledge Management: Recommender Systems
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Explanation fidelity, which measures how accurately an explanation reflects a model's true reasoning, remains critically underexplored in recommender systems. We introduce SPINRec (Stochastic Path Integration for Neural Recommender Explanations), a model-agnostic approach that adapts path-integration techniques to the sparse and implicit nature of recommendation data. To overcome the limitations of prior methods, SPINRec employs stochastic baseline sampling: instead of integrating from a fixed or unrealistic baseline, it samples multiple plausible user profiles from the empirical data distribution and selects the most faithful attribution path. This design captures the influence of both observed and unobserved interactions, yielding more stable and personalized explanations. We conduct the most comprehensive fidelity evaluation to date across three models (MF, VAE, NCF), three datasets (ML1M, Yahoo! Music, Pinterest), and a suite of counterfactual metrics, including AUC-based perturbation curves and fixed-length diagnostics. SPINRec consistently outperforms all baselines, establishing a new benchmark for faithful explainability in recommendation. Code and evaluation tools are publicly available at https://github.com/DeltaLabTLV/SPINRec.

Problem

• Improving explanation fidelity in recommender systems to reflect true model reasoning
• Adapting path integration techniques for sparse implicit recommendation data
• Overcoming limitations of fixed baselines with stochastic baseline sampling

Innovation

• Stochastic baseline sampling from empirical data distribution
• Model-agnostic path integration for recommendation systems
• Captures influence of observed and unobserved interactions