Abstract
Reinforcement learning (RL) offers a mechanism for transforming recommender systems from static preference-prediction tools into adaptive decision-support systems that learn from sequential user and environmental feedback. However, evidence on how RL is integrated and evaluated within recommender systems relevant to personalized financial decision support remains fragmented across application domains. This study systematically maps empirical RL-related approaches according to their RL approach, integration role, decision formulation, and evaluation practices. A systematic search of Scopus, ACM Digital Library, ScienceDirect, SpringerLink, and IEEE Xplore covering peer-reviewed studies published between 2015 and 2025 yielded 7,531 records. Following deduplication, screening, eligibility assessment, and quality appraisal, 50 studies were retained, of which 30 empirically implemented RL-related adaptive techniques and constituted the mapped evidence set. Multi-armed and contextual bandits were the most prevalent approaches (43.3%), followed by Q-learning and value-based RL (26.7%), actor–critic and policy-optimization methods (16.7%), feedback-aligned RL (6.7%), and model-based planning or tree-search approaches (6.7%). Across these approaches, RL-related methods supported direct recommendation and policy selection, recommender or strategy selection, model and representation adaptation, and preference elicitation and planning. Reward formulations ranged from behavioral and recommendation responses to utility, revenue, cost, profit, and feedback-derived objectives. However, direct financial applications were comparatively limited, and evaluation relied substantially on offline and experimental settings, with limited longitudinal and real-world validation against sustained financial outcomes. The findings demonstrate that effective RL-enabled financial recommendation requires alignment among the decision context, state and action representation, reward function, feedback mechanism, constraints, and evaluation outcomes. The study provides a structured mapping of the field and identifies financially meaningful reward design and longitudinal validation as key priorities for advancing personalized savings and investment decision support.