MODERATE HUMEANISM

The maximisation of expected utility hypothesis brings together two separate claims. The first concerns what rationality requires of the relation between the agent's preferences between different prospects and her beliefs and desires. Stripped of mathematical baggage, the claim can be expressed as follows:

The Rationality Hypothesis is generally taken to express nothing more than a consistency requirement on the agent's preferences, akin to the requirements that logic places on her beliefs. Consistency requirements are purely formal in nature and place no substantial constraints on the content of any preference, belief or desire taken in isolation. Moreover, the constraints that they place on sets of such preferences, beliefs and desires are not such as to rule out many that we might be inclined to regard as defective in some way; for instance, because they are immoral, self-destructive or just plain ill-considered.

In ordinary talk we tend to be more demanding and speak of beliefs as irrational, even if they are consistent, because they fail to meet some or other standard of adequacy. For instance, we might be inclined to criticise someone for not taking into account all available evidence or for failing to give the long-term consequences of their choices sufficient weight. Such talk, it seems to me, runs together two types of requirements that are best kept separate. One is the requirement that we recognise all the available evidence and that we give appropriate weight to all the possible consequences of our actions; the other that our beliefs be consistent with all the evidence that we recognise and that our preferences for actions be consistent with the weight that we give to each of their possible consequences. The former is a requirement that our judgements respond in an adequate way to the world as it is presented to us; the latter that they fit together in a coherent way.