Axiomatic methods, dopamine and reward prediction error