Patient Rights Safeguards and Conditional Acceptance of AI Assisted Diagnosis in Bangladesh

Abstract Acceptance of artificial intelligence in diagnosis raises questions about what people consider useful and which patient safeguards they endorse. This study examines their relationship in 385 adult, AI-aware questionnaire records supplied as empirical data from Bangladesh. Five diagnosis-and-rights statements are analysed separately because their content and low block reliability do not support a validated common acceptance scale. Response distributions, paired endorsement contrasts, eight combinations of rights preferences, Spearman correlations and exploratory logistic models distinguish professional usefulness, information, questioning, human judgement and conditional acceptance. The statement that patient-rights safeguards increase acceptance received 81.6% endorsement, compared with 56.9% for usefulness when AI supports qualified professionals. Information about material AI involvement received 81.0%, opportunities to ask questions 57.7%, and continuing human clinical judgement 63.9%. All three specific safeguards were jointly endorsed by 32.7%. Information preference had a small positive rank association with conditional acceptance after Holm correction. Adjusted information and questioning coefficients were positive but uncertain after correction and sensitivity checks. The findings identify differentiated endorsement of safeguards rather than an observed effect of rights protection on service uptake. They suggest evaluating communication as an opportunity for participation, not merely notification, while keeping normative rights distinct from an adoption strategy. Mixed respondent status, absent verification of patient experience, undocumented recruitment and fieldwork dates, and unresolved Codebook provenance labels restrict generalisation. The study provides an item-specific basis for further research connecting patient communication, professional oversight and conditional acceptance without claiming clinical effectiveness or tested legal knowledge.