I've been thinking about why some domains have reached more definite, mathematized, and law-like models than others, e.g. the hard sciences of physics and chemistry vs the soft sciences of psychology and political science.
One hypothesis is that it's just much easier to experiment on, say, pendulums than on humans. Experimenting on humans is slow, it's hard to create very controlled conditions, IRB's severely limit what you can do. Theoretically, if I could experiment on a large number of humans for long periods of time without constraints on what I could do (something like Aperture Science?), then I maybe could harden the soft sciences.
A piece of evidence on this would be how good are our mouse models? There are substantially fewer restrictions on mice experiments and resultantly we do a lot of experiments on them (hence the "IN MICE" that ought to be appended to so many scientific result headlines). Maybe translation into humans is poor, and we're generally focused on humans so we feel like there's a lot we don't know, but if you just asked about how well we understood mice, actually we have a lot of really solid knowledge about psychology, metabolism, immune function, political economy, etc.
Can anyone familiar with mouse models in any domain comment on how good the mouse models are vs human models? 1x as good, 5x, 20x? Good = really solid predictions, we feel like we know what's going on, etc.
For that matter, the same question could be asked for drosophila.