A B S T R A C T

The purpose of the research – this paper analyze recommender systems in e-commerce, highlight the transparency and trust concerns caused by black-box algorithms, and determine how explainability can improve user satisfaction and confidence.

The methodology of the research – the article reviews recent Explainable Recommender System (XRS) methods and applies them to a synthetic e-commerce dataset of user–item ratings with product attributes (category, brand, price). Both intrinsic interpretable models and post-hoc techniques (e.g., attention mechanisms, LLM-based generation, SHAP analysis) are employed to illustrate how recommendations can be explained.

The practical importance of the research – the findings provide value for e-commerce practitioners and researchers by demonstrating how explainability can enhance transparency, build consumer trust, and improve platform credibility. They also offer guidance for integrating XRS approaches into real-world online retail environments.

The results of the research – the study shows that generating human-understandable rationales for recommendations (e.g., “this item is suggested because it matches your preferred category and is low-priced”) improves user trust and satisfaction. However, explainability also involves a trade-off with predictive accuracy, which must be carefully managed.

The originality and scientific novelty of the research – the paper emphasizes the underexplored role of explainability in e-commerce recommender systems and demonstrates, through synthetic data analysis, how XRS methods can be operationalized to balance accuracy with interpretability. It contributes original insights into bridging transparency gaps in recom­mendation algorithms.

Keywords: explainable recommender systems, e-commerce, transparency, user trust, in­ter­pretable machine learning, explainable AI, synthetic dataset, SHAP.

http://doi.org/10.59610/bbu4.2025.4.11