Demystifying Sequential Recommendations: Counterfactual Explanations via Genetic Algorithms

2025-08-05

AI Summary

This work introduces the first counterfactual explanation method for sequential recommendation systems (SRSs) that operates directly on discrete interaction sequences, aiming to identify the minimal, feasible modification to a user’s historical interactions that would alter the target recommendation—thereby enhancing model interpretability and user trust. We formally prove that generating an optimal counterfactual sequence is NP-complete, reflecting the inherent computational hardness of the problem. To address this, we propose a tailored genetic algorithm that explicitly respects both the discrete nature of item sequences and semantic plausibility constraints. Extensive evaluation across three representative SRS architectures, three real-world datasets, and four experimental settings (target/non-target, classification/unclassified) demonstrates that our counterfactuals achieve high readability and near-perfect model fidelity (∼1), significantly outperforming existing baselines. Our core contributions are: (i) the first formal problem formulation of counterfactual explanation for SRSs; (ii) establishment of its computational complexity boundary; and (iii) an efficient, scalable solution framework.

Technology Category

• Reasoning under Uncertainty: Sequential Decision Making
• Search and Optimization: Adversarial Search
• Data Mining & Knowledge Management: Recommender Systems

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: Economics and fairness of platforms and recommendation systems

Abstract

Sequential Recommender Systems (SRSs) have demonstrated remarkable effectiveness in capturing users' evolving preferences. However, their inherent complexity as "black box" models poses significant challenges for explainability. This work presents the first counterfactual explanation technique specifically developed for SRSs, introducing a novel approach in this space, addressing the key question: What minimal changes in a user's interaction history would lead to different recommendations? To achieve this, we introduce a specialized genetic algorithm tailored for discrete sequences and show that generating counterfactual explanations for sequential data is an NP-Complete problem. We evaluate these approaches across four experimental settings, varying between targeted-untargeted and categorized-uncategorized scenarios, to comprehensively assess their capability in generating meaningful explanations. Using three different datasets and three models, we are able to demonstrate that our methods successfully generate interpretable counterfactual explanation while maintaining model fidelity close to one. Our findings contribute to the growing field of Explainable AI by providing a framework for understanding sequential recommendation decisions through the lens of "what-if" scenarios, ultimately enhancing user trust and system transparency.

Problem

• Generates counterfactual explanations for sequential recommender systems
• Identifies minimal changes in user history altering recommendations
• Ensures interpretability while maintaining model fidelity

Innovation

• Genetic algorithm for counterfactual explanations
• NP-Complete problem solution for sequences
• Framework for what-if scenario analysis