Thanks for your reply. The discussions of computational modelling and simulations as a new empirical method are interesting. See Robbert Dijkgraaf's contemplations as a particularly good recent example.) It seems a distance from empirical as observations based in reality, but fascinating as a means of exploring the implications of explanations and theories. So I would say, this area is interesting to the degree that they may produce observation statements as hypotheses that could be subsequently evaluated against empirical observations as they’re conventionally understood.

RE: emergence, I don’t know what to do with accounts that don’t make reference to explanations generally. I’m further flummoxed by blanket dismissals of varying modes of explaining. For example, I don’t see a problem in invoking a mode like causality or teleology, if that mode provides a handle to investigate the phenomena more deeply using other modes. (The example I’m holding at the moment is a bacterium “searching for food.” I’m plugging this conversation between David Haig and Sean Carroll as a particularly good discussion of this problem.) That high-level explanation would obviously be unsatisfying if that were the end of it. But it motivates an ambitious research program to understand exactly how and why a bacterium could search for food, which may involve explanations in physics, information theory, chemistry, etc.

I don’t think the matter is just that it would be inefficient and counter-productive to understand phenomena using some blanket mode, but that understanding is this interweaving of explanations. To return to your analogy of ANNs, it’s not only futile to try to understand how the model works as networks of nodes and weights, but that to take such a perspective would forego the opportunity to understand it all.