"For action to be rational, the beliefs on which it is based must themselves be well founded. In turn, this requirement is divided into two parts. On the one hand, the beliefs must be unbiased with respect to the information the agent possesses; on the other hand, he must gather an optimal amount of information."

“The theory of rational action may therefore fail, because it is unable to produce unique prescriptions and predictions. It may also fail if the agents’ behavior does not conform to predictions, whether these are unique or not; that is, if the agents are irrational. There are multiple sources of irrationality, hot or cold.”

“In this reconstruction, the idea of reason comprises three elements: impartiality with regard to persons, temporal impartiality, and rational or well-founded beliefs. No doubt we should add goodwill, in order to exclude impartial malice.”

“The premise of this statement is what one might call the principle of non-indirection. I call “indirection” an indirect operation that makes it possible to arrive at a certain result through two successive actions, the first of which serves only to make the second possible. ... This principle is not limited to rational actions. ... The principle of non-indirection thus expresses a broader idea of coherence that includes the emotional as well as the rational.”

“Acting in conformity with reason, in the singular, and acting for good reasons, in the plural, are two different things insofar as reason is objective, whereas reasons are subjective. From an external point of view, we can evaluate a policy as being in conformity with reason or not. From an internal point of view, one can evaluate an action as being rational or not. From this difference it follows that only rationality can be used for explanatory ends. It is only insofar as the agent has made the demands of reason his own that the latter may give rise to, and possibly explain, specific behaviors. The assessment of the actor and that of the observer need not coincide.”

“In a general way, we can represent the present value of a future good as a function of the time separating the present from this future. In the classical conception, an exponential future discount is stipulated, which implies that the curves corresponding to two distinct future goods, one small and immediate, the other large and more remote, never intersect. According to more recent research, however, it seems that this discount typically takes a hyperbolic form.”

“Let us take first an absurd example: always preferring goods that come on Thursdays to those that come on Wednesdays, solely because of a preference for that particular day of the week. As we shall see, this is not contrary to the principles of rational choice, but it is certainly contrary to reason. The simple preference for Thursdays is a reason, but reason also demands the reason for that reason. And obviously there is none.”