Analogy in decision-making and social interaction and emergent rationality

Colman’s reformulation of rational theory is challenged in two ways. Analogy-making is suggested as a possible candidate for an underlying and unifying cognitive mechanism of decision-making, one which can explain some of the paradoxes of rationality. A broader framework is proposed in which rationality is considered as an emerging property of analogy-based behavior. Rationality has long been shown to fail as a descriptive theory of human decision-making, both at the individual and social levels. In addition, Colman presents strong arguments that rationality also fails as a normative theory for “good” decision-making – “rational” thinking does not produce optimal behavior in social interaction and even acts against the interests of the individual in some cases. Fortunately, human beings often act against the postulates of rationality and achieve better results than prescribed by the theory. Therefore, Colman concludes that “rationality” has to be redefined by extending it with additional criteria for optimization, such as the requirement for maximizing the “collective” payoff, or with additional beliefs about the expected strategies of the coplayers. He does not clarify how and when these additional criteria are triggered or where the common beliefs come from. We are so much attached to the notion of rationality that we are always ready to repair it, but not to abandon it. The theory of rationality is, in fact, a formalization of a naive theory of human thinking. This naive theory makes it possible to predict human behavior in most everyday situations in the same way as naive physics makes it possible to predict natural phenomena in everyday life. However, no one takes naive physics that seriously to claim that it provides “the explanation” of the world. Moreover, even refined and formalized versions of this naive theory, like Newtonian mechanics, are shown not to be valid, and more complicated and counterintuitive theories at the microlevel, like