One worry for simple belief-desire psychology is that it potentially conflates several importantly different kinds of motivation or pro-attitude:
(1) Affect - i.e. positive emotional valence, or pleasure.
(2) Drive - sheer behavioural/motivational force, e.g. compulsion.
(3) Evaluation - degree of reflective endorsement.
(Any others?) It seems at least logically possible that these could come apart. For example, one might compulsively act in a way which feels emotionally neutral to the actor, but which they rationally judge to be bad. Does the agent "desire" to so act? It depends which of the three meanings we have in mind.
It seems clear that mere behavioural drive has no normative significance. Affect is more promising, since we generally like to have positive emotions. But in that case, it is arguably deriving its value from #3: evaluation.
I wonder whether these distinctions can help us to make sense of so-called 'conditional desires', e.g. the desire for ice-cream, which seem in some sense to be conditional on their own persistence. (Shieva has a neat paper [doc] explaining why the notion is so problematic.) We feel a transient affective pull towards ice-cream, and we positively evaluate the goal of obtaining ice-cream while the feeling persists. This allows us to avoid self-reference, as the evaluation is instead conditional on the persistence of affect. Or something like that.