Using Individualized Treatment Effects to Assess Treatment Effect Heterogeneity

Imagine running a clinical trial and, instead of just asking "Does this drug work on average?", you ask "How much does this drug help this specific type of patient compared to others?" This paper is about math and algorithms that estimate, for each individual patient profile, how much extra benefit (or harm) they get from a treatment versus not taking it, and then using those estimates to understand which subgroups benefit most or least.