Pretty sure postmodern data science is a contradiction in terms. At least the whole SSK movement has seemed sublimely indifferent to the constraints of data- reality quite generally in fact - on theory.

But there's a very long way from positivism to post-modernism. Luckily we can recognize the limitations and distortions of narrow inductivism without having to go full social constructivist and deny the existence of reality external to our perception of it (qua Sokal).

Your comment seems rather to support my point in the article that the postmodern social madness that Quine and Kuhn unwittingly kicked off (much to their despair) drove scientists back into the arms of positivism.

Don't get me wrong; I deeply understand this drive. Positivism is powerful and - as you say - generates a lot of good empirically observable results. Pure postmodern social constructivism produces absolutely nothing, almost by definition. Infinitely rather the former than the latter.

But I think we can do better than positivism. We mustn't be seduced by the purported empirical triumphs of an empirically flawed framework and the damage is real. Statistical analysis in particular would benefit immeasurably from freeing itself from the logical positivist miracle of model-free analysis.