A community-based AI-assisted health microlearning program for older adults: Preliminary changes in AI literacy, psychological resilience, and health literacy

Population aging and rapid digitalization underscore the need for community-based education integrating artificial intelligence (AI) learning with critical appraisal of health information among older adults. This preliminary single-group pretest–posttest evaluation examined a 10-week AI-assisted health microlearning program delivered across nine sites in Taipei, Taiwan. The study assessed within-participant changes in four self-report outcomes and explored concurrent associations among posttest measures. Valid records comprised 170 baseline and 162 posttest assessments, with 132 participants providing complete matched data. Participants were predominantly women and adults aged 65 years or older, and reported interest in AI-supported health learning despite limited prior AI use. Under the center-clustered analysis with Bonferroni adjustment (α = .0125), attitudes/perceived value (p = .002) and health literacy (p = .009) met the adjusted criterion, whereas AI literacy (p = .109) and psychological resilience (p = .018) did not. Change in AI literacy was modest (Cohen’s dz = 0.295). At posttest, all four domains were positively correlated, with the strongest bivariate association between psychological resilience and health literacy (r = .71); this pair also showed the strongest conditional association in the saturated model (β = .611). Because posttest variables were measured concurrently and the model was saturated (df = 0), these associations do not establish temporal ordering, mediation, or causality. Findings indicate differentiated questionnaire-score changes and concurrent relationships. Controlled longitudinal studies incorporating performance-based measures are needed. Because longitudinal measurement invariance was not established, pretest–posttest findings should be interpreted as questionnaire-score changes rather than confirmed changes in invariant latent constructs.