Abstract Artificial intelligence-enhanced Clinical Decision Support (AI-CDS) is a growing multibillion-dollar industry leveraging a wide range of clinical, genomic, social, geographical, web-based, and wearable device data for improvements in health outcomes broadly circumscribed under the term “precision health.” Powered by Big Data, characterized by volume, velocity, veracity, variety, and value, “big knowledge” in the form of AI-CDS is becoming increasingly ubiquitous (volume), rapidly developing (velocity), available to a wide range of medical fields (variety), based on data from a wide range of sources that reflects the health of individuals and populations (veracity), and focused on lowering costs and promoting better health outcomes (value). Current policy paradigms for CDS, including whether to classify it as a medical device, are not designed for adaptive artificial intelligence technologies. Patients and providers have no reasonable way to discern how these “black box” technologies operate or their accuracy. Innovative policies (e.g. standards in product labeling) that address these concerns are likely to require direct consumer outreach and communications to ensure public trust in the growing AI-CDS field. Indeed, public trust in AI-CDS has been identified as a top priority for the AI- CDS big knowledge ecosystem by the National Academy of Medicine, NIH, FDA, and OMB, among others. Trust is particularly salient given the range of critical ethical and policy considerations related to transparency, privacy, non-maleficence, equity, accountability, and utility of AI-CDS. In Aim 1 of our proposed study, we will measure the public's current trust in AI-CDS for precision health and assess (a) its relationship to the public's expectations and concerns about privacy, equity, non-maleficence, responsibility, and utility and (b) how it may be affected by policies and practices, such as labeling or certification. In Aim 2 we will use deliberative democracy methods and expert interviews, designed to directly inform policy and standards that address perceived risks of AI-CDS and in Aim 3 we propose to develop a product information label that would both increase transparency and accessibility of information about AI-CDS for patients and providers. The continued acceptance and adoption of AI-CDS is predicated on public trust and our proposal provides a research-focused and evidence-based approach to incorporating public participation into emerging national standards.

Project Narrative The goal of the proposed project is to address the gap in current research on public perspectives about ethical best practices and the impact of Artificial intelligence-enhanced Clinical Decision Support (AI-CDS) on the trust of the public. We propose to focus on policy and practice options such as certification, notification, and product labeling related to essential, endemic issues in the AI-CDS ecosystem: transparency, privacy, equity, non- maleficence, accountability, and utility (Aim 1 and Aim 2). For Aim 3, we evaluate key attributes for product labeling and an accompanying multi-dimensional metric of trust reflective of the Food and Drug Administration's system of categorizing software used as a medical device.