According to Phillips et al., the capacity to represent someone as knowing something is more basic than the capacity to represent someone as believing something. Their evidence for this claim comes from an impressive variety of sources. They observe, for example, that nonhuman primates attribute knowledge, but not belief, and that the capacity to represent what others know emerges earlier in human development than the capacity to represent what they believe. The claim that knowledge is more basic than belief runs counter to a long-standing tradition in philosophy, cognitive science, and social science that emphasizes belief as the primary representational mental state. Think of the false-belief task in psychology, or of the decision theorist's emphasis on beliefs (subjective probabilities) and desires (utilities), or of the traditional epistemologist's attempt to decompose knowledge into belief plus truth plus something else. On the contrary, Phillips et al.'s emphasis on knowledge over belief has an important philosophical counterpart in the recent “knowledge-first” program in epistemology (Williamson, Reference Williamson2000), a connection that will be pursued below.

Let us call the capacity for representing belief the belief capacity, and the capacity for representing knowledge the knowledge capacity. The view the authors reject has two parts: It says that the belief capacity is more basic than the knowledge capacity, and that the knowledge capacity depends on the belief capacity. In arguing that the knowledge capacity is more basic than the belief capacity, the authors reject both parts of this view. But they appear to remain neutral on the further question of whether the converse dependency claim holds (sect. 3.2). Does the belief capacity depend on the knowledge capacity? Or are these two capacities simply independent of each other?

The hypothesis that the belief capacity depends on the knowledge capacity might help to explain some of the generalizations the authors discuss. For example, if you weren't able to represent beliefs without being able to represent knowledge, that would explain why we haven't come across any creatures that can represent belief but not knowledge – we haven't come across any because there couldn't be any. Similarly, the hypothesis that the belief capacity depends on the knowledge capacity would also explain why the belief capacity does not emerge earlier in human development than the knowledge capacity – it doesn't emerge earlier because it couldn't. Of course, other explanations of these facts are possible, but they are likely to be more complex; that provides us with some prima facie motivation for exploring the idea that the belief capacity depends on the knowledge capacity.

Let us suppose, for the sake of argument, that this so. How should we understand this dependence? In traditional epistemology, knowledge depends on belief in the sense that knowledge is belief plus truth plus something else. But belief is almost certainly not knowledge plus anything, because one can believe things one does not know (falsehoods being the prime example). Could belief be knowledge minus truth? Perhaps, but it is not immediately clear what this would mean (although see Yablo, Reference Yablo2014).

An alternative picture emerges if we focus on certain commonalities between knowledge and belief. As the authors observe, knowledge attributions are sometimes deployed to predict behavior. If you represent Maxi as knowing that the chocolate is in the drawer and represent him as wanting to retrieve the chocolate, you will no doubt expect him to look in the drawer. But, of course, belief attributions can also be deployed in this way: If you represent Maxi as (merely) believing that the chocolate is in the drawer and represent him as wanting to retrieve the chocolate, you will still expect him to look there – and this is so even if you know that the chocolate has actually been moved to the cupboard. One thing this suggests is that when we represent someone as (merely) believing p, we expect them to act as they would if they had known p. When Maxi falsely believes that the chocolate is in the drawer, he acts just as he would if he had known the chocolate was in the drawer. (The “if x had known p” locution here is to be understood in a way that doesn't presuppose that p is in fact true.)

One possibility, then, is that representing someone as believing p involves representing them as acting as if they knew p, or perhaps, as being disposed to act as they normally would if they had known p (see also Williamson, Reference Williamson2000, pp. 46–47). If that is (part of) what it is to represent someone as believing something, then it is no wonder that one cannot represent belief if one is unable to represent knowledge. Note also that, according to this proposal, a representation of belief would appear to involve a counterfactual conditional; this might help to account for the apparent link between the ability to engage in (complex) counterfactual reasoning and the ability to pass the false-belief task (e.g., German & Nichols, Reference German and Nichols2003; Riggs, Peterson, Robinson, & Mitchell, Reference Riggs, Peterson, Robinson and Mitchell1998).

Phillips et al. provide some evidence that people are sometimes willing to say that x knows p whereas at the same denying that x believes p (sect. 5.2). This is still possible if belief is understood in the manner suggested above, for one may know p without acting as if one knows p or without being disposed to act as one normally does when one knows p. Indeed, this seems an apt description of the “unconfident examinee” (Myers-Schulz & Schwitzgebel, Reference Myers-Schulz and Schwitzgebel2013; Radford, Reference Radford1966) – she knows the answer, but does not act as if she knows it.

If knowledge is more basic than belief, then it is tempting to think that the belief capacity depends on the knowledge capacity. One form this dependency could take is this: Representing someone as believing involves representing them as acting as if they know. If this proposal is incomplete or entirely wrong-headed, perhaps it will nevertheless serve to provoke others to provide a better account of how belief might depend on knowledge.