Machine Learning-Based Artificial Intelligence in Recommender Systems to Support Financial Education: A Systematic Review

Abstract

A lack of personal financial planning and knowledge about how to manage money are among the factors that contribute to citizens’ indebtedness. This study investigates how machine learning-based artificial intelligence has been applied in recommender systems to support financial education and financial health. A systematic literature review was conducted using PRISMA-informed procedures, with research questions structured according to Population, Intervention, Comparison, Outcome, and Context. Eligible studies applied machine learning to personalized recommendation, financial-profile classification, or financial decision support for individuals, incorporated at least one interpretability, explainability, or process-transparency mechanism, and were published in English, Portuguese, or Spanish between 2021 and 23 July 2026. Five databases were searched on 23 July 2026, yielding 139 records, of which 11 studies were included after deduplication and screening. Methodological quality was assessed using a nine-item framework addressing objective clarity, dataset and validation description, real-user evaluation, and reproducibility. Findings were narratively synthesized across nine research questions because heterogeneous designs and metrics precluded meta-analysis. Included studies primarily employed tree-based and ensemble models, clustering, and time-series techniques, supported by feature-attribution methods, local explanations, Integrated Gradients, interpretable models, or natural-language justifications. Quality appraisal revealed limited real-user evaluation and reproducibility. No controlled experiments compared identical systems with and without explanations; therefore, current evidence does not establish that explanations alone improve user understanding, trust, decision quality, or financial behavior. Future research should prioritize longitudinal, real-user evaluation of explainable financial recommender systems. The review protocol was not publicly registered.