> the simplest explanation is simply that women prefer other fields
I either don't find this particularly simple or a particularly explanatory.
On one reading it's just banal/tautologous; most of this discussion (innate inalienable preference, versus systemic bias, versus cultural pressure) concerns preference in some sense - people choosing to do something rather than being directly forced or prevented. My understanding of this discussion is that it is about what influences preference, and whether we should try and alter those influences across the life of a person.
Furthermore it's only really explanatory if you have some a priori expectation that the class of 'woman' ought to possess some group characteristic in this respect. So to me again this seems like a phenomenological observation ('people called Harry are less scared of heights') rather than much of an explanation.
The only way I can make sense of it as an explanation is if we are linking some innate biological preference in sexual females with some intrinsic quality inherent in STEM work. I am not ideologically opposed to either biological biases or intrinsic qualities of work from a cognitive standpoint. But the casual chains I have seen spun out from one to the other seem highly inconclusive, partial, or dubious (e.g. some notion of 'mechanical reasoning' to 'preference across all of STEM'). If the argument based on this simple explanation is intended to dismiss or substantially dismiss theories of systemic biases or structural inequality then I would need to be persuaded that there are not too many nooks and crannies for these cultural mechanisms to work. But it seems to me there are. At the start of this chain there are potential biases in early education (even in the neurological bootstrapping of infancy), and at the end of the chain the link between intrinsic quality of work versus how the work is done within a cultural context.
To try and be more concrete and hypothetical, what of the effect of the teaching of discursive comprehension based subjects to boys that may make it less likely to be their best subject, what about the types of things given to babies to play with (I see this regularly with a 6 month year old daughter), what about the differential effect on working hours across pregnancy, what about the way in which daily tasks in STEM are defined, etc. Seems to me like there are many degrees of freedom here.
To me the cognitive difference between women and men seem like minor cognitive perturbations - so it is surprising that this could cause large differences in distributions in some higher order effects many steps removed like profession.