AI Racial Identification

In this episode, hosts Prasanth and Mohannad interview @judywawira (Emory University), Chima Okechukwu (GeorgiaTech) @BrandonPriceRad (FSU College of Medicine) and Anthony Celi (MIT laboratory for Computational Physiology) as they discuss a groundbreaking paper they published earlier this year, in which they demonstrate how shockingly accurate machine learning systems can be at determining race. Starting from an accidental discovery to an arduous process of eliminating surrogate indicators or proxies they were startled to see how robust these model predictions were to signals virtually imperceptible to even expert human observers, across modalities as well as anatomic regions. They will also speak about the broader questions of access to healthcare in across minority and underrepresented groups, the importance of careful ML design choices. The discussants address the concept of race as a social construct, a predicter or health outcomes, myriad difficult race-related questions science and society will have to tackle as we embrace AI in Medicine.

Connect with us! You can find our podcast on Apple Podcast, Google Podcasts, Stitcher, or anywhere else you subscribe to podcasts. Please help us out by leaving a review on Apple Podcasts.

You can find us on Twitter: @SIIM_Tweets, and individually at @mohannadhussain, @jaynagels, @AAnandMD

Special Thanks to @RandalSilvey of PodEdit for editing and post processing support.