Traditional debates about the reality of beliefs, especially those arising from the controversies around eliminative materialism, were centred on the question of the existence of beliefs. This ontological way of framing the issue of reality of beliefs has, in my view, tilted the debate against the anti-realist camp: denying the existence of beliefs was widely thought to be self-refuting at worst and counter-intuitive at best. But even if we admit that beliefs do exist (at least in a deflationary sense), there is still a way of questioning their reality: one might question whether belief properties are natural. In this way, a moderate anti-realist view about beliefs becomes the idea that although beliefs exist, they are not natural properties. The issue of naturalness of beliefs can be usefully translated into three more concrete subquestions: of whether beliefs constitute natural kinds, whether they are causally relevant, and whether their content can be naturalised. In all the three cases we have good reasons, even if of negative nature, to prefer the anti-realist option.

The moderate anti-realist position has important implications. First, it supports the autonomy view in the debate about the relationship between folk psychology and cognitive science. If beliefs are best conceptualised as non-natural properties, then it is reasonable to suppose that descriptions in folk psychological vocabulary do not provide us with the knowledge of the causal structure of the human mind. Rather they offer us a constituting a certain vision of what it is to be human. Causal explanations of human mind, on the other hand, are the domain of middle-level posits of cognitive science. In effect, there should be no place for competition between cognitive science and folk psychology.

The second consequence of moderate anti-realism is its pragmatic approach to metaphilosophy of beliefs. In philosophy of mind, we often deal with issues which can be labelled as the question of the ‘boundaries of belief'. We might ask which systems are capable of being genuine believers; e.g. can AIs or non-human animals have them? More often, the question is whether certain mental states of humans (delusions, religious convictions, and implicit biases) should be understood as being cases of beliefs. Adopting a moderately anti-realist view provides basis for the claim that such questions lack objective answers; there are no boundaries of belief which are determined by the world itself. Rather, we should see the concept of belief as plastic: i.e. amenable to changes. This leads to a pragmatic methodology for tackling these questions. We should not be asking questions of the type: is state X really a case of belief, but rather: should we think of state X as being a case of belief? What are the pragmatic benefits/downsides of, e.g. admitting that corporations have beliefs or that religious credences are cases of belief? Moderate anti-realism about beliefs provides us with tools to see many questions in philosophy of mind in a pragmatic fashion.