Knowledge, concept of

6. Externalism

Partly in response to the difficulties with foundationalism and coherentism even as supplemented by the defeasibility theory, epistemologists have developed a variety of alternative accounts of warrant. They have been called ‘externalistic’ because their accounts of warrant focus on features of the world other than the knower’s reasons for belief. Two important ones are the causal theory and the reliabilist theory (see Knowledge, causal theory of; Reliabilism).

In their purest forms, these accounts begin with the view that knowledge, and hence warrant, does not require justification. The foundationalists had already conceded that there are no reasons for properly basic beliefs. This seemed to create a problem for foundationalism only because it was assumed that all beliefs needed to be justified and the ‘basic’ reasons appeared to be arbitrary. But drop the requirement that beliefs need to be justified in order to be warranted, and this problem immediately disappears.

Roughly, the causal theory of warrant holds that a belief is warranted if and only if the state of affairs represented in the belief is appropriately causally related to the belief. For example, suppose I come to believe that there is a bird in a tree as a causal consequence of seeing the bird in the tree. Sometimes the causal connection is more complex; but this direct type of causal connection between the belief and what it represents will suffice for our purposes.

This theory is initially appealing because it appears to satisfy the basic requirement that a warranted belief be non-accidentally true since the state of affairs represented in the belief is a cause of my belief. However, it is easily seen to be both too weak and too strong; and there seem to be some deep problems with it as a general account of warrant. It is too weak because it would count some true beliefs as warranted that clearly are not known. Recall the Grabit case. My belief that I see Tom stealing the book is caused by Tom’s stealing the book, but if he has an identical twin, I do not know that Tom stole the book. It is too strong because there seem to be many beliefs that count as knowledge which can not be appropriately causally related to what they represent. Suppose I know that there is no elephant smaller than a kitten: what possible causal connection could there be between there being no elephant smaller than a kitten and my belief? In addition, potential difficulties arise about knowledge of a priori propositions (such as 2 + 2 = 4) and counterfactuals (such as, if it were raining today, we would have called off the picnic). It looks as though there is no possible way to produce a causal connection between my belief and what is represented in the belief – at least as ‘cause’ is usually understood (see A priori).

Nevertheless, a basic tenet of the causal theory might still be correct: Not all beliefs need to be based on reasons in order to count as knowledge. The reliabilist theory of warrant can be seen as the successor of the causal theory. Instead of requiring an appropriate causal connection between the states of affairs represented in the belief and the belief itself, a typical form of reliabilism holds that a belief is warranted just in case the process resulting in the belief produces true beliefs sufficiently often.

Thus, the non-accidental nature of the true belief receives a very straightforward analysis. The belief is non-accidentally true because the process that produces the belief produces true beliefs sufficiently often. This view has many advantages over the causal theory. My belief that elephants are larger than kittens need not be caused by that state of affairs. All that is required is that the process by which I come to believe that proposition typically (often enough) results in true beliefs. A priori or counterfactual propositions present no problem since there could be reliable processes that produce those beliefs.

Nevertheless, there are problems confronting this view. Suppose that you require that the process should produce true beliefs on 100 per cent of the occasions on which it arises. That is a very stringent condition; but it is not stringent enough! For if the belief that Tom Grabit stole a book arises only once in the history of the world – the time I saw him stealing the book – the actual process produced a true belief 100 per cent of the times it arose; but it is not knowledge. The obvious move for the reliabilist is not only to include the actual occasions when the particular belief is produced but rather to consider whether the type of process that produced this belief would produce true beliefs of this type sufficiently often. But correctly characterizing those types has not proved easy. Is the type of belief one in which Tom is involved? Or identical twins? Or libraries? That seems too narrow. Is the type of process one in which there is first a perception and then some inferences? That seems too broad. It remains an open question whether reliabilism can produce an acceptable account of the types of processes and the types of beliefs.

Finally, there is one further objection that some epistemologists have brought against reliabilism. Perhaps it is best illustrated in a case presented by Keith Lehrer (1990: 163) that can be summarized as follows: a certain Mr Truetemp has a thermometer-with-temperature-belief-generator implanted in his head so that within certain ranges of temperatures he has perfectly reliable temperature beliefs. When it is 50 degrees, he comes to believe that it is 50 degrees. When it is not 50 degrees, he does not come to believe that it is 50 degrees. He holds these beliefs without knowing why he does.

Such beliefs would satisfy all of the requirements suggested by the reliabilists, but many epistemologists would hold that although Mr Truetemp has true beliefs and they are not accidentally true because his thermometer-with-temperature-belief-generator is reliable, they are accidentally true from the cognitive point of view, as he has no reasons at all for his beliefs. Indeed, some would say that what Mr Truetemp possesses is a skill (of telling the temperature) and not propositional knowledge at all.

Here we can detect a fundamental clash of intuitions. The reliabilists would hold that Mr Truetemp does know; the traditional normativists would hold that he does not. There appears to be no way to satisfy both. But this much seems clear: There are some situations in which the steps in the process that brings about a belief include the holding of reasons. In those cases in which there is no automatic true-belief-generator (as in the Truetemp case) and in which we must rely upon our reasoning to arrive at a belief, the questions asked by the traditional normativists are crucial: what must the structure of our reasons be so as to make a true belief acceptable? Are there foundational reasons? Can mutually supporting reasons be offered without begging the question? (Could reasons be infinite in number?) And need those reasons be such that they are not undermined by the truth, as the defeasibility theorists would hold? At least in some cases, it seems that normative standards for belief-acquisition apply and their satisfaction will determine whether a belief ought to be accepted. Thus, it appears that an evaluation of the conditions under which beliefs ought to be accepted, denied or suspended is inescapable (see Internalism and externalism in epistemology; Justification, epistemic).