This is why we do randomized experiments, just FYI. Obviously, there are many situations where we can't do them (e.g., randomizing people to smoking / not smok

This is why we do randomized experiments, just FYI. Obviously, there are many situations where we can't do them (e.g., randomizing people to smoking / not smoking). In those cases, the challenges the article discusses become relevant.