Events
It’s a model eat model world: Implications of the real-world use of machine learning models | Institute of Artificial Intelligence for Digital Health
There has been massive interest in demonstrating the utility of machine learning / artificial intelligence models in healthcare settings, accompanied by limited real-world deployment and measurement of downstream effects. Little attention has been paid to the fact that since clinical data is sourced from and is captured back into the EHR, normal / expected model operation can create feedback loops that can disrupt the function of currently existing and future models. The work discussed in this talk explores the implications and extent of this effect through simulations of model deployment in various scenarios.
Speaker Bio
Akhil Vaid, MD is a physician-scientist whose work deals with the application of multi-modal machine learning to clinical questions. He’s currently an instructor in the Division of Data Driven and Digital Medicine (D3M) / Department of Medicine at the Icahn School of Medicine at Mount Sinai.