A preference is a type of desire in belief-desire psychology. A desire is an attitude to make or keep something true and a preference along with urges, drives, needs, wants, intentions, plans, goals are all types of desire.
A preference can also mean a lexical ordering of desires. That is given two desires , a desire that P and a desire that Q, two agents could express different preferences over these desires. Agent A might prefer fulfilling desire P over Q, granted they could only fulfil one of these two desires. Similarly agent B might prefer fulfilling desire Q over P.
In this second sense of preference, the clash between fulfilling desires of an agent given limited resources, information, time and so on is already solved - for any individual agent. However it is not solved for clashes between agents. Here economists and others (such as Dworkin) propose excluding external preferences. An external preference is a preference of an agent that is dependent on other's preferences being fulfilled or thwarted. By excluding external preferences the problem of an inter-personal clash between individuals preferences (in the second sense), is supposedly avoided.
There are two problems with this solution. The first is that it excludes any possible benefits to others - one could call these "positive" external preferences - as well as "negative" external preferences. A negative external preference in this light would be items such as supporting and voting for Proposition 8 in the recent California elections. An example of a positive external preference would be giving money or volunteering time to a charitable organisation.
A deeper issue with external preferences is that it is ad hoc because it is ill-defined and indeterminate. If one takes two extreme opposite caricatures of say Randian Objectivist and a Green Socialist - not that either must use such preference terminology - but still they would disagree over the use of a petrol-guzzling 4x4 to take one's children to school. The Objectivist would argue that no external preference is being expressed in this act and it is perfectly within their rights to do so, whereas the Green Socialist would argue that there are external costs - pollution, traffic safety and congestion, inefficient petrol consumption and so on - and so this is an external preference that is being expressed. Which is right? Does it depend in differing ideology or is there a proper empirical basis to decide this?
A careful reading of Informed- and Rational-desire theorists such as Griffin, Railton and Brandt tacitly implies (I can find no explicit references on this, if you know of any please supply it in the comments) that a fully informed agent - say one who would have only true beliefs (Brandt) or who would also know whether fulfilling their desire is really worthwhile (Griffin) - would have not have many such problematic external preferences. For example, their preferences would not be based on double standards such as those leading to actions (such as votes) based on prejudice and bigotry, as this cannot survive rational examination such as the standard of rational consistency. Whilst this is a more natural way of excluding problematic external preferences it is not guaranteed to work and, indeed, Railton needs to further specify - by adding additional conditions - as to what counts as moral desires after the filtering already done to produce only informed desires. This is less ad hoc but still relies on some idealisation - ideal agents -which does not exist in reality and is not guaranteed to work anyway.
Other more sophisticated preference satisfaction theorists of the ethical not economic variety are more versed in this issue and do not invoke external preferences. However they, inexplicably, leave this key issue to be puzzled out on a case by case basis rather than extract and emphasize an overall pattern applicable to all cases. The only writer I can find close to noting this general pattern is an obscure comment from Hare:
this is done by showing how a wise educator, seeking to maximise the satisfaction of all preferences indiscriminately, weight for weight, would try to cultivate some and discourage others. He would cultivate those whose satisfaction is compatible with, and discourage those whose satisfaction militates against, the satisfaction of preferences as a whole. Thus he will discourage sadistic desires, cultivating the disposition to think intuitively that they are evil (as they are).[Hare: Utilitarianism and Double Standards (1992) pg.311]
Quite independently, Alonzo Fyfe's Desire Utilitarianism makes this point not obscurely but quite explicitly and centrally to his underlying preference satisfaction model. A desire is morally good to the extent that it fulfills or tends to fulfil or does not thwart or tends not to thwart all other desires - without exception or bias. So no ad hoc and indeterminate external preference construct need be applied nor is there the need for any hypthetical ideal agent.
Given the choice here over dealing with the issue of external preferences surely the Desire Utilitarian approach is to be preferred as it is not ad hoc, does not rely on hypthetical ideals and is the least indeterminate and most robust solution? Are there any solutions better than this?