Hi Adam, really enjoyed your article (and series). I’m an aspiring data scientist, and reading your post gave me a lot more context on why “throwing data into an algorithmic black box” doesn’t work. I found it particularly enlightening how understanding causal relationships can contribute to building more accurate regressors.
I was wondering how causal diagrams could be applied to other techniques besides regression. This includes common classification techniques such as decision trees or SVMs, and why abstractions like deep nets can be so effective (from Pearl’s perspective). What are your thoughts on this?