In the Mind or in the World? Types of Beliefs and the Locality of Evidence

People’s beliefs appear to be divisible into two distinct types. Some beliefs refer directly to the observable world, readily guide behavior, and are easily revised when challenged. On the other hand, ideological, political, and religious beliefs are associated with much more obscure evidence, often do not lead to behavior, and appear immune to counterevidence. To explain this intuitive difference between beliefs, several theories in psychology have proposed that the mind possesses two qualitatively different types of belief with different cognitive properties. Abelson (1986) distinguished testable from distal beliefs, where the former are updated on the evidence while the latter are held for instrumental reasons. Sperber (1997) proposed that intuitive beliefs are automatically formed by the mind, while reflective beliefs are only meta-represented, keeping them sequestered from other thoughts. Van Leeuwen (2023) divides beliefs into factual beliefs and credences, arguing that many ideological beliefs are a cognitive attitude more like imagination than true belief. In this paper, I draw on three neglected theories in psychology from the 1950s and suggest that they can explain apparent differences among beliefs without positing a novel type of belief. In so doing, I critique an assumption underlying types-of-beliefs theories: that an appeal to psychological properties is needed to explain why beliefs appear to differ from each other. In contrast, I suggest that differences among beliefs may be explained by a non-psychological characteristic of the world. This emphasis on studying the environment is inspired by Brunswik (1952, 1956). Relatedly, Bruner, Goodnow, and Austin (1956) suggested that concepts might differ not in the psychological processes used to classify them, but in the cues that contribute to their category-membership. Importing their notion of attribute immediacy to belief, I propose that beliefs differ in the locality of evidence that supports them. When most evidence is local, e.g., in beliefs based on perception, people’s beliefs often converge. However, in domains like politics, much of the relevant evidence is unavailable for scrutiny, allowing disagreements to persist. I develop a continuum of the proportion of evidential locality from deduction – which produces certain conclusions from fully local evidence – to perceptual and highly isotropic (Fodor, 1983) beliefs, which represent decreasing locality and certainty. An important consequence of non-local evidence is March and Simon’s (1958) uncertainty absorption, where conclusions are transmitted without their supporting evidence or reasoning. In domains with low evidential locality, we often need to defer to experts, with the result that we are unaware of the evidence supporting their conclusions. This may help beliefs in such non-local domains avoid counterevidence – beliefs cannot be conclusively refuted if their believers are unaware of their own reasons. I discuss how locality of evidence and uncertainty absorption provide a parsimonious account of a series of phenomena which motivated proposals of types-of-beliefs and suggest future directions for research into non-local evidence.