Epistemic relativism

2. Two relativistic accounts of cognitive assessment

Though it often goes unnoticed, some of the most popular accounts of how systems of reasoning are to be assessed are, or at least might well turn out to be, relativistic. Here, two such accounts are considered: one based on reflective equilibrium, the other based on a system’s truth-generating capacity.

Nelson Goodman claimed that general principles of inference were justified by their conformity with the particular inferences we make and accept, and that our acceptance of particular inferences was justified by their accord with general inferential principles. This, he noted, looked ‘flagrantly circular’ but, he continued:

John Rawls (1971) introduced the term ‘reflective equilibrium’ to label the endpoint of the process of ‘delicate … mutual adjustments’ that Goodman describes.

Although Goodman did not discuss the matter, other authors have noted that there is no guarantee that everyone who uses the process will end up at the same point. If two people begin with significantly different judgments rejecting or accepting particular inferences, or with different views about which rules they are willing to amend (or both), then it seems entirely possible that they will end up with quite different sets of rules, though each set will be in reflective equilibrium. If, as Goodman insists, the process of mutual adjustment is all that is needed for rules and inferences to be justified, then these people may end up reasoning in very different ways, each of which is justified for the person who reasons in that way.

Reliabilist accounts of how to assess systems of reasoning or belief revision link the assessment to the truth-generating capacity of the system (see Reliabilism). Other things being equal, the better a system is at producing true beliefs and avoiding false one, the more highly a reliabilist will rank it. Though it is not often emphasized by reliabilists, this sort of assessment is quite sensitive to the environment in which people using the system find themselves. Thus it may well turn out that a given system of reasoning does an excellent job for one person and a very poor job for another. Imagine a pair of people who suddenly fall victim to Descartes’ demon, and are from that time provided with systematically misleading or deceptive perceptual data. Suppose that one of the victims has been using cognitive processes quite like our own, and that these have done a good job in generating truths and avoiding falsehoods, while the other victim’s cognitive processes have been (by our lights) quite mad, and have produced far more falsehoods and far fewer truths. In their new demon-infested environment, however, the ‘normal’ system of cognitive processes will yield a growing fabric of false beliefs. The other system, by contrast, may now do a much better job at generating truths and avoiding falsehoods, since what the evil demon is doing is providing his victims with radically misleading evidence – evidence that only a lunatic would take to be evidence for what actually is the case. So on an account of cognitive evaluation in which generating truths and avoiding falsehoods plays a central role, our system would be preferable in one environment, the mad system in another. Which system a person ought to use will depend on which environment the person is in.

Invocation of evil demons to make the point might suggest that this is a very peripheral phenomenon that is hardly worth worrying about. However, the Cartesian demon case is just the very small tip of a very large iceberg. Any reliabilist evaluation of cognitive processes is going to be acutely sensitive to the cultural, technological and epistemic setting in which the processes are to function. The likelihood that one system of cognitive processes will do a better job than another at generating truth, I suspect, will depend on such factors as the existence of a system of writing, the existence and the structure of disciplinary communities, and the relation of those communities to the political and economic arrangements of the wider society. It will also often depend on the level of conceptual, mathematical, scientific and technological sophistication that has been achieved. If these conjectures are right, it follows that reliabilist accounts of cognitive or epistemic evaluation will have a certain post-Hegelian historicist flavour. There will be no one ideal method of inquiry, no cognitive system that excels in all historical settings. Rather, we can expect that the assessment of a cognitive system will vary as its historical setting varies, and that, just as with technologies (and indeed with genes), it will sometimes happen that a successful system will undermine its own success by changing the environment in such a way that competing systems will now be more successful.