Promoting regular physical exercise is crucial for healthy ageing, yet many older adults remain physically inactive. Artificial intelligence (AI), evolving from rule-based to data-driven systems, offers a promising tool to address this challenge. This scoping review systematically maps existing research on AI in older adult exercise, focusing on intervention characteristics, types of assessed exercise motivation, physical function, mental health outcomes, and underlying theoretical frameworks. Adhering to Arksey & O’Malley’s framework, this review systematically searched Scopus, MEDLINE, Web of Science, PubMed, Google Scholar, and IEEE Xplore for English-language studies on data-driven AI in older adult exercise. Four research questions guided the review. Two independent reviewers screened 5760 articles, with 18 included based on PICOS criteria. Data extraction utilized a structured coding scheme and descriptive analysis, achieving high inter-rater reliability (Kappa ≥ 0.897). Academic interest in AI-based exercise for older adults is rapidly increasing, particularly in East and Southeast Asia. AI-based tools commonly featured sensing/tracking (n = 10), real-time feedback (n = 12), and personalized guidance (n = 12). 55.6% of studies reported positive findings regarding exercise motivation, and 72.2% reported positive changes in balance, gait, muscle strength, and movement accuracy. However, findings regarding grip strength, activities of daily living, lower limb function, range of motion and mental health (44.4% of studies) were inconsistent. Common limitations included small sample sizes, short-term follow-ups, lack of control groups, and external validity issues. Theoretical frameworks were often used for evaluation rather than core AI design. Studies included in this scoping review suggest that AI may have the potential to support exercise engagement among older adults. Reported outcomes included favorable exercise motivation and specific physical function changes. A conceptual model categorizing AI capabilities and their influence on outcomes is proposed. Future studies with robust designs, long-term follow-up, user-centered AI development, and integration of psychological needs (autonomy, competence, relatedness) are needed to support sustainable exercise behavior and healthy ageing.
Ye et al. (Tue,) studied this question.