I am not opposed to mathematics in the least. I am not even opposed to deductive approaches. I am opposed to making assumptions that are unrealistic only because they make the math work better. The maximization assumption is mistaken but makes it possible to use calculus. I am not impressed. Neither am I impressed with sophisticated deductions from false assumptions. Such deductions can sometimes be useful, but I will bet on other more realistic approaches.

Miyake's comment seems aimed at the tautology charge. The problem lies in the claim that there should be a rational reason for all behavior, not in a simplistic claim that all behavior is simply dictated by preferences. The human mind does not work like a computer, is not good at math. The neural net simulations are much more interesting and the point of neural net approaches is to get a simulation that not only does what the brain does but to do it in the way the brain does. Even Bayesian updating would be more realistic, but the math would not be as elegant.