I used to think that people use the “weasel word causal inference strategy” (“our analyses show that X is an important predictor of…” followed by conclusions that are clearly causal) for the sake of plausible deniability. But I’m revising that now, because often deniability *isn’t* plausible.>