I now suggest the following as necessary and sufficient conditions to be fulfilled, or to be assumed to be fulfilled, in order appropriately to regard any given behavior pattern as purposive:
There must be, on the part of the behaving entity, i.e. the agent: (a) a desire, whether actually felt or not, for some object, event, or state of affairs as yet future; (b) the belief, whether tacit or explicit, that a given behavioral sequence will be efficacious as a means to the realization of that object, event, or state of affairs; and (c) the behavior pattern in question...
Two objections to this analysis may be anticipated. If, firstly, it is objected that my conception invokes dubious or occult entities, viz., desires and beliefs, I reply that, however unsuccessful science may be in describing or explaining these occurrences, everyone knows perfectly well, in one clear sense, what it is to desire something, and what it is to believe something, and that, to this extent, there is nothing at all dubious or occult about them. And if, secondly, it is objected that these occurrences are not observable, and thus not operationally useful to science, I reply that this simply is not true; they are observable in the same sense that atoms, for example, are, viz., as inferences from what is directly observed, and are directly observable to the agent in whom they occur, in the same sense that pleasure and pain, for example, are.
Taylor, R. Purposeful and non-purposeful behavior: a rejoinder. Phil. Sci. 17, 327-332, 1950.