When one is in the grip of a desire one typically knows it. He who wants a cold beer on a hot day knows what he wants and is likely to deem unhinged anyone with the temerity to deny that there are desires. Anywhere on the scale from velleity to craving, but especially at the craving end, there is a qualitative character to desire that makes it phenomenologically undeniable. If the beer example doesn't move you, think of lust. Lust is an intentional state: one cannot lust unless one lusts after someone or something. But although lust flees itself, voids itself in a rush towards its object — as Sartre might have said — there is nonetheless something 'it is like' (T. Nagel) to be in the state of lust. In this respect, desire is more like the non-intentional state of pain than it is like the intentional state of belief. There is most decidedly something it is like for me to desire X; but what is is like for me to believe that you desire X? Is it like anything? Not so clear.
One cannot replace folk psychology (FP) with a better theory (one couched in neuroscientific terms perhaps) unless it is a theory. What I have been suggesting, however, is that it is not a theory. If this is right, then there is just no chance that in the future we will re-conceptualize that in us which we now folk-psychologically conceptualize as desire in some radically different way that makes no mention of desire or cognate mentalistic notions.
But don't we give folk-psychological explanations, and doesn't the fact that we do this show that folk psychology is a theory? Suppose Sally is angry, and we want to figure out why. Talking to her we find out that she is in dire financial straits and needs a substantial raise in pay. Her boss promised her one, and in expectation of the fulfillment of the promise, Sally ran up some credit card debt. But it turns out that the boss was unable to deliver on his promise. Hence, Sally's anger. This suggests an explanation involving a folk-psychological 'law':
L. If a person P is promised X confidently by a person Q whom P trusts, and P needs/desires X and expects the fulfillment of the promise, and Q is unable to deliver on the promise, then, ceteris paribus, P will become angry at Q.
To get an explanation in the familiar deductive-nomological style, we simply instantiate the variables in (L) with 'Sally,' 'boss,' and 'raise' and add the premise that Sally is promised a raise by the boss.
But is this the only way to explain Sally's angry behavior? Must we assume that FP is a theory? There is also a mental simulation view of FP. Very roughly, on this model we would understand Sally's angry behavior by 'putting ourselves in her shoes,' by imagining ourselves in her situation and then running a simulation 'off-line' as it were. For details click on the above link.
The main point is that eliminativism in the philosophy of mind requires the notion that FP is a theory; so any considerations that speak against this assumption are considerations that speak against eliminativism. If FP is a theory, it is a theory of something which could be theoretically interpreted in some other way. But what could that something be? The fact of the matter is that desires and beliefs are phenomenological givens; they are not theoretic posits. They are not theoretical but 'datanic.'