AI for Surrogate Decision Making?!? Dave Wendler, Jenny Blumenthal-Barby, Teva Brender

[@SolidarityAndLiberation]

I appreciate the conversation you all shared. This was a breadbasket of philosophical and ethical questions about medicine both in the acute care setting and the broader system in which that happens. However, I think it’s best as a thought experiment and don’t see a particular benefit to empirical testing. When there are dozens of questions about how something could go wrong, maybe you ought to focus on eliminating those factors before diving into experiments. Particularly the questions around profit seeking and bias. There are so many ways to bias an AI system for the benefit of the system. Manipulating training data, cherry picking hyper parameters, etc. it doesn’t have to be a conspiracy theory. Small changes to increase profit are the natural incentives.