All models are essentially lower-dimension approximations of preexisting human understandings. Without the understanding, you can’t build — or test — the model. As Norbert Wiener (the “father” of modern operations research) informed us decades ago, all human behavioral outcomes are over-determined. Multiple explanations of cause may explain every observed effect and none is uniquely accurate.
There is no safe way to isolate or test for individual factors operating in the holographic internal structures of the brain. The whole notion of “brain circuit” is meaningless word noise. What we have instead is a distributed self-adaptive synapse network operating in a massively parallel biological processor whose electrical architecture is mediated by multiple competing external inputs. Science doesn’t presently know how to model such a combination — and it quite possibly never will. We can’t model the machine from inside it.