Abstract / Summary
Artificial intelligence (AI)-based clinical prediction models (CPMs) using electronic health record (EHR) data may support earlier autism detection, yet raise important ethical questions—particularly regarding consent, transparency, equity, and communication. This US-based qualitative study examined perspectives on ethical use of AI-based CPMs among caregivers of young autistic children ( n = 8), clinicians involved in screening and early intervention ( n = 10), and autistic adults ( n = 8). Using semi-structured interviews and focus groups, we conducted thematic analysis to identify concerns, sources of trust, and preferences for communicating autism likelihood within shared decision-making contexts. Participants expressed cautious optimism about AI-based CPMs, noting potential benefits for earlier identification, clearer understanding of developmental concerns, and more timely access to services. At the same time, they highlighted risks such as inequities stemming from incomplete EHR data, limited transparency of “black-box” models, the emotional burden of very early screening, and reduced caregiver involvement when tools operate without caregiver input. Participants felt screening was most valuable when conducted early—but not too early that families receive results without clear next steps—allowing information to be meaningfully actionable. Ethical implementation of AI-based CPMs must extend beyond technical performance to include embedded ethical practices that promote trust, equity, and supportive caregiver-clinician partnerships. Interest-holder perspective can inform the ways that this ethical implementation is achieved.