The Importance of Automatic Subgroup Analysis in Machine Learning Models

A lot of people talk about AI drift as kind of one of the liabilities for a lot of these models, but on the flip side, it also could be like a fantastic opportunity. There's an interesting parallel here too with the whole real world evidence thing because even when a drug comes to market, you're trying to find is other sub populations that this works much better for. So yeah, I think a big part of work with AI is how to also to the model and monitor human drifts.