They’d probably argue that this same model works for that - you just update the manifold. That’s probably fair, the model isn't really powered to be predictive just correlative.

It still embeds some of the same problems, though, mainly how an individual actually interacts with it and why their interpretation is correct.