Mental content and functional role
Propositional attitudes have an attitude type (belief, desire, etc.), and a content. A popular idea in the literature on intentionality is that attitude type is determined by functional role and content in some other way. One can find this view, for example, in Fodor (1987, 17), Dretske (1995, 83), or Loewer (2017, 716). I don't see how it could be correct.
Attitude type and content are obviously independent insofar as one can bear more or less any attitude to more or less any content. But it doesn't follow that the type and content of an attitude are determined by independent factors.
More concretely, a state's functional role seems to severely constrains its content. Take a desire for coffee. To a first approximation, the functional role of this state is to cause behaviour of which the subject believes that it will get them coffee. The state has the typical functional role of a desire, but that role isn't neutral about the content! Most desires have a different functional role.
Fodor et al seem to think that a state's functional role entails nothing about their content. The content, they suggest, might be determined by causal connections between relevant concepts and objects in the environment, or by the evolutionary purpose of the relevant representations, or whatever.
The fatal problem I see in all such accounts – in every one of the accounts surveyed by Loewer (2017), for example – is that they make it a mystery why intentional states play their typical behavioural/functional role.
Suppose a state plays the behaviour/functional role of a desire for coffee. According to Fodor et al, this leaves open the state's content. So let's assume that its content (as determined by causal relations or whatever) is that horses can fly. What kind of state is this supposed to be?
It's not a desire that horses can fly, as it doesn't play the right functional role. For the same reason, it's not a belief that horses can fly, or a hope that horses can fly, or any other recognizable attitude directed at horses and flying. Remember the state plays the typical role of a desire for coffee. But it's not a desire for coffee either, as it doesn't have the right content: its content is that horses can fly.
There is no such state! There is no state with the functional role of a desire for coffee and the content horses can fly.
Why is there no such state? The only explanation I can think of is that content (not just attitude type) is constrained by functional role. And then content can't be solely a matter of causal connections etc.
Stalnaker speaks to this issue in a curious passage on pp.17f. of Stalnaker (1984). He discusses the case of a person, Mary, who plays cello at 3 am in order to annoy her neighbour Fred. We'd like to say that Mary wants to annoy Fred and believes that playing the cello will annoy Fred. But Mary's behavioural dispositions don't settle that her attitudes are directed at Fred. Mary's twin on Twin Earth, for example, has the same behavioural dispositions, but her attitudes are directed at Twin Fred. From this, Stalnaker infers that behavioural dispositions underdetermine the content of attitudes. Then he claims that the point generalizes: that Mary's behavioural dispositions entail nothing about the content of her attitudes. ("The content of belief and desire cancels out".)
Stalnaker doesn't really argue for the generalization. It seems obviously false. Take, on the one hand, an agent's total behavioural dispositions. Take, on the other hand, a complete hypothesis about their beliefs and desires. Most pairings don't "fit": most hypotheses about what the agent believes and desires don't make sense of their behavioural dispositions. Most don't even make sense of their actual present behaviour. My current beliefs and desires, for example, would make no sense of Mary's cello playing.
Just how strongly do behavioural dispositions constrain the content of one's attitudes? Lewis briefly looked at this question in Lewis (1983, 50f.). A more careful examination is Williams (2016). Lewis and Williams both conclude that behavioural dispositions underdetermine content in a problematic way – if all we assume is that the agent is disposed to satisfy their desires in light of their beliefs. It's not clear how much underdetermination remains if we adopt additional (substantive, non-formal) rationality constraints. At any rate, I see no reason to think that an agent's behavioural dispositions would entail nothing about their beliefs and desires.
The "bubble puzzle" doesn't seem too different from various supernatural beliefs that are still accepted by some metaphysicians - eg my desire for coffee is part of a fight between guardian angels and devils. Indeed my avoidance of coffee might arise from an intellectual acceptance of Pascal's Wager.