six-lemmas-narrative
I hear a lot about "imposter syndrome" these days. It refers to feelings of inadequacy experienced by people who are, in fact, very competent. Some cases of imposter syndrome are presumably false positives — not a syndrome, just a regular old imposter. In any event, I'm experiencing sensations of imposter syndrome on behalf of my field and I'm naturally wondering if it's a false positive or not. I work on machine learning for causal inference (causal ML) and, quite frankly, it hasn't proven as useful as we all had hoped, at least not yet. But why? Does machine learning for causal inference feel unsuccessful to me simply because it remains underused at present, or is it little used because it has failed to deliver on its promise on those occasions that it has been used so far? Would further attempts redeem it? Discussing this issue requires understanding some of the inherent limitations of the method as well as some common unjustified biases against ML for causal inference. Here, I am g