Although behavioral economists have criticized the empirical accuracy of the neoclassical rationality assumption, they support the normative view that welfare-increasing choice presupposes stable, context-independent preferences. We argue that this position neglects important features of decision-making that were essential to the thinking of Frank H. Knight and James M. Buchanan (Old Chicago economics). Knight and Buchanan embrace the idea that well-integrated preferences are an intellectual construct that cannot be the normative basis for welfare assessments. They argue that the only trait that is stable in real time and across individuals is the urge to have some vaguely defined better preferences. In other words, exploring, learning, and being contradictory and ambivalent are all part of the process both before and after a temporary maximization. We illustrate their view, present some empirical evidence, and discuss the normative implications especially for the character of a liberal social order such as Richard Epstein has advocated.