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.

Counterfactual evaluation provides a promising framework for assessing explanation fidelity in recommender systems, but perturbation metrics adapted from computer vision suffer three key limitations: (1) they conflate explaining and contradictory features, (2) they average over entire user histories instead of prioritizing concise, high-impact explanations, and (3) they use fixed-percentage perturbations, leading to inconsistencies across users.We introduce refined counterfactual metrics that focus on the most relevant explaining features, exclude contradictory elements, and assess fidelity at a fixed explanation length, ensuring a more consistent and interpretable evaluation.Our code is at: https:// github.com/DeltaLabTLV/FidelityMetrics4XRec

The increasing focus on explainability within ethical AI, mandated by frameworks such as GDPR, highlights the critical need for robust explanation mechanisms in Recommender Systems (RS). A fundamental aspect of advancing such methods involves developing reproducible and quantifiable evaluation metrics. Traditional evaluation approaches involving human subjects are inherently non-reproducible, costly, subjective, and context-dependent. Furthermore, the complexity of AI models often transcends human comprehension capabilities, rendering it challenging for evaluators to ascertain the accuracy of explanations. Consequently, there is an urgent need for objective and scalable metrics that can accurately assess explanation methods in RS. Drawing inspiration from established practices in computer vision, this research introduces a counterfactual methodology to evaluate the accuracy of explanations in RS.