Perceived Safety and Continued Use in AI-Enabled Ride-Sharing: A Protection Motivation Perspective Among Uber Riders in Bangladesh

Background: Ride-sharing platforms convert a traditionally interpersonal transport transaction into a digitally mediated interaction in which users must evaluate both physical safety and the platform’s ability to support protective action. In Bangladesh, existing ride-sharing research has largely emphasized service quality, satisfaction, convenience, price and general adoption, leaving a narrower question underdeveloped: whether safety appraisal translates into continued platform use once safety perceptions themselves are considered. Objective: This study develops and empirically audits a Protection Motivation Theory (PMT)-informed safety pathway linking threat/coping appraisal (TCA), perceived safety (PS), and continued use intention (CUI) among Uber riders in Bangladesh. Method: The analysis uses the Group-4 master survey workbook supplied for Summer 2026. Of 400 preallocated respondent IDs, 385 cases contain complete survey variables and pass the workbook’s attention check; 15 rows are blank. Construct diagnostics, correlations, bootstrap path regressions, mediation analysis, effect sizes, and out-of-sample cross-validation were conducted. Results: TCA is positively associated with PS (β = 0.343, p < .001; R² = 0.116) and positively associated with CUI in the total-effect model (β = 0.226, p < .001). When PS is entered simultaneously, the TCA→CUI path remains significant (β = 0.194, p < .001), whereas PS→CUI is not significant at the 5% level (β = 0.092, p = .095). The indirect effect through PS is small and its bootstrap 95% interval includes zero. Measurement diagnostics are weaker than conventional publication thresholds: Cronbach’s alpha ranges from .384 to .462 and AVE from .448 to .482 across the three focal constructs. Conclusion: The dataset indicates a meaningful safety-appraisal signal but does not support a simple mediated claim that perceived safety itself explains continued use. The distinctive implication is a safety-continuance disconnect: riders can translate threat/coping appraisal into stronger safety perceptions without that perception independently becoming a robust continuance driver. The result motivates future models incorporating trust, structural assurance, incident experience, alternative availability, price, and platform-response quality while improving the measurement instrument.