Artificial Intelligence (AI) Facts Labels: An Innovative Disclosure Tool Promoting Patient-Centric Transparency in Healthcare AI Systems.

Abstract

In the field of healthcare, artificial intelligence (AI)-assisted solutions can be viewed with anxiety or apprehension, thus transparency and trust-building are essential. AI is often invisible (and potentially undisclosed) to users, violating the ethical principle of transparency -- particularly with respect to informed disclosure. This Brief Technical Report describes the creation of a novel, patient-centric prototype AI transparency tool (AI Facts Label) using the United States Drug Facts Label as a model. The prototype was then populated with lay language in the context of a hypothetical AI-assisted wearable medical device. Using lay language amid a defined graphic template that could be globally harmonized, the AI Facts Label is a communication tool that honors the ethical principle of transparency and could potentially aid readers (patients and healthcare workers) in understanding the use of AI in a specific product. AI Facts Labels have the potential to be used globally (adapted to local language) in clinical research settings as well as commercial settings as a transparency tool that discloses when a system is AI-assisted. AI Facts Labels could also be used as a communication tool in other settings such as general consumer electronics (e.g., televisions, household appliances, mobile phones) and non-healthcare apps.