By 2021, the size of the health AI market is expected to be about 11 times what it was in 2014, growing from $600 million to an estimated $6.6 billion. With this explosion of health AI technology comes an urgent need to better understand how to effectively communicate information about these tools in order to harness benefits and mitigate risks. Clinical decision software enabled by AI represents a particularly important area of focus. Information around appropriate use, benefits, and risks is potentially more complicated to determine and/or disclose for AI-enabled clinical decision software than for traditional medical devices. For example, AI-enabled software that utilizes certain deep learning techniques can make recommendations, but there may be no way for a human to understand how the software came to that recommendation. Additionally, commercial competitiveness concerns may restrict how much information a company may want to release. In the absence of information about “how the software works,” what information do stakeholders want? What information do companies want to share?

To explore these issues and more, the Center for Innovation Policy at Duke Law and the Duke-Margolis Center for Health Policy are holding a public meeting to better understand how to incentivize innovation in this space while communicating necessary information to stakeholders on how to use these products safely and effectively. This session will delve into the role of innovation incentives, such as patents and trade secrecy, over data and model details. It will also address mechanisms by which FDA regulation affects innovation.