Internalism and Epistemology

The Architecture of Reason

Timothy McGrew, Lydia McGrew

• Publisher: Routledge
• Year: 2007
• ISBN:
  9781135985967
• Edition: 1
• Topic: Philosophy
• Subtopic: Philosophy History & Theory

This book is a sustained defence of traditional internalist epistemology. The aim is threefold: to address some key criticisms of internalism and show that they do not hit their mark, to articulate a detailed version of a central objection to externalism, and to illustrate how a consistent internalism can meet the charge that it fares no better in the face of this objection than does externalism itself.

This original work will be recommended reading for scholars with an interest in epistemology.

1 Internalism and the Collapse of the Gettier Problem

Among challenges to internalism, the Gettier problem has pride of place. Four decades of discussion and analysis have produced no consensus on the problem itself, but the externalist response has been (not surprisingly) to tout it as evidence that internalism is seriously flawed. Alvin Plantinga, in particular, has argued that the Gettier problem is not just a minor conundrum requiring us to “tweak” the traditional TJB definition of knowledge but rather a sign that external factors regarding proper function are constitutive of the deepest and most important epistemic concepts.

For their own part, most internalists are understandably unwilling to define knowledge in general as requiring infallibility; to do so would preserve internalism at too high a price. But the introduction of a fourth condition involving causal grounding or reliable production mechanisms serves to weaken the impression that possessing beliefs and relying on them as reasons is of more than “folk psychological” importance.

In the rush to externalism, an alternative possibility for internalists has been largely ignored and, when discussed, has almost invariably been given short shrift. The heart of this solution is a clarifying analysis of what Gettier (and many others) took to be the J condition in the traditional analysis of knowledge. We will argue that the term ‘justification’ must be disambiguated. In one sense, it is equivalent to rationality; in another – the sense in which it is required for inferential knowledge – it involves an extension of the T condition.

We will argue that, while externalists have used Gettier as a motivation for the addition of various external causal requirements for knowledge, the externalist use of the Gettier problem has involved an ambiguity regarding what counts as “accidental knowledge.” We will then consider a variety of attempted counterexamples to the solution we advocate, demonstrating that in each case the example fails and our analysis enables us to pinpoint the reason for its failure. Finally, we will discuss the effects of adding a T condition to crucial premises for the second sense of ‘justification’ and demonstrate that, far from pushing us to an externalist concept of knowledge, questions about the T component of the J condition actually collapse into the traditional epistemic regress argument.

Two senses of ‘justification’ and closure

In his seminal article, Gettier sets out to counterexample the traditional definition of knowledge as true, justified belief by exhibiting two cases in which a putative knower has true, justified belief but fails to have knowledge. ¹ Before laying out the examples, Gettier notes two assumptions that are necessary to make the examples work. First, in the pertinent sense, S may be justified in believing that q without its being the case that q: justification does not entail truth. Second, if S is justified in believing that q, and q entails p, and S deduces p from q and accepts p as a result of this deduction, then S is justified in believing that p: justification is closed under known entailment.

Gettier goes on to set out two attempted counterexamples to the TJB analysis of knowledge, each of which has the following characteristics:

1. An individual S has evidence e which justifies his belief that q
2. S knows that q entails p
3. S forms the belief that p on the basis of his knowledge of the entailment relation from q and for no other reason
4. As it turns out, q is false
5. As it turns out, p is true.

Gettier maintains that p is a justified belief because of 1–3 and the principle that justification is closed under known entailment. But by 5, p is true. Hence p is a true, justified belief for S. Yet it seems intuitively obvious that it cannot count as knowledge; hence, true, justified belief is insufficient for knowledge.

Ironically, both Gettier’s puzzle and the central insight required to resolve the difficulty were indicated by Bertrand Russell half a century prior to the publication of Gettier’s paper.

To take a very trivial instance: If a man believes that the late Prime Minister’s last name began with a B, he believes what is true, since the late Prime Minister was Sir Henry Campbell Bannerman. But if he believes that Mr. Balfour was the late Prime Minister, he will still believe that the late Prime Minister’s last name began with a B. . . . [He] may proceed to draw valid deductions from the true premiss that the late Prime Minister’s name began with a B, but he cannot be said to know the conclusions reached by these deductions.²

As Russell points out, the obvious way to rule these cases out is to amend our definition of knowledge, and as a first step he suggests the following: “Knowledge is what is validly deduced from known premises.”³ Russell goes on to modify this definition in various ways; what he does not change, and sees no reason to change, is the emphasis on the importance of true premises in inferential knowledge.

The Russellian solution to the Gettier problem seems to face a difficulty right at the start. For clearly there is something epistemically significant about reasoning from beliefs for which one has reasons, even if it turns out that those beliefs are false. Indeed, to abandon that intuition would be to abandon internalism altogether. Russell’s solution, with its stress on true beliefs, might appear to downplay the importance of rationality, and in that case it would be a non-starter.

But the objection rests on the unspoken premise that supporting beliefs play an important role only insofar as they produce knowledge. To undermine this assumption, we need to mark a distinction between two senses of ‘justification.’ Premises that provide justification₁ to p are (roughly) propositions on the basis of which S believes that p, which provide adequate grounds for rationally believing that p, and which are themselves held with a rational degree of credibility. The crucial premises of an argument that provides justification₂ are premises in this sense, but they must also be true. (We shall discuss the concept of a crucial premise in detail later in the chapter.) Transferring this analysis to Gettier’s cases, S possesses justification₁ but not justification₂ for believing that p, since he grounds it crucially in q, which is false. And therefore, contra Gettier, S’s belief that p does not satisfy the definition of knowledge.

Rather surprisingly, this solution has not gained wide acceptance, despite the fact that, in putting it forward, Russell solved the Gettier problem before Gettier. Besides Russell’s anticipatory treatment, two articles in the mid-1970s advanced the core of this solution after Gettier. Robert Meyers and Kenneth Stern put forward a version of this solution (with explicit acknowledgment given to Russell) in which they mark the distinction by introducing “well-takenness” to indicate what we are calling “justification₁.”⁴ John Dreher, in a brief piece, distinguishes being “justified in believing” from “believing on good evidence” – where his being “justified in believing” refers to justification₁ and “believing on good evidence” indicates justification₂.⁵ But in the face of a flurry of putative counterexamples the Russellian position has received no further public support, and Robert Shope’s treatise on the Gettier problem gives it a fairly cursory dismissal.⁶

One consequence of the Russellian position that does not change despite varying terminology among its proponents is that the justification required for inferential knowledge (justification₂) is not closed under known entailment. Gettier explicitly invokes the premise that justification is so closed,⁷ and it may be that the apparent counterintuitiveness of denying closure has prevented acknowledgment of the force of the solution. But as Meyers and Stern note, Gettier’s article does not distinguish between the two senses of justification. When these two concepts are disentangled, we can argue that justification₁ is hereditary, i.e. closed under entailment known with certainty, although justification₂ is not.⁸ When we realize that justification₁ isolates the distinctively epistemic nature of justification – i.e. rationality – it is clear that it should be closed under certainly known entailment, since entailment preserves rational confidence. (See Chapter 2.) But since both senses of justification leave open the possibility that a proposition could be justified but false, it is obvious that justification₂ is not closed. One may rationally believe a proposition on the basis of a falsehood, if the falsehood itself is rationally believed, but the relation involved in justification₂ requires that the crucial premises of one’s argument be truths as well.

We should note briefly here that, for the strong (incorrigibilist) foundationalist, it is not possible for a foundational belief to be rationally held but false, and (of course) foundational beliefs are exactly those that do not require any premises. Hence, a strong foundationalist will take the distinction between these two types of justification to be relevant only to inferred beliefs. Foundations, on this theory, have all the justification required for knowledge in themselves and are not subject to the possibility of Gettier undermining. We can think of this in various ways, and as long as we avoid confusion it does not much matter which we choose. We can say that justification₂ is required only for inferential knowledge, or we can say that justification₂ requires only that the crucial premises for a belief, if there are any such, must themselves be known. If one expresses the requirement in the latter terms the strong foundationalist would say that the foundations do satisfy the justification₂ condition because there are no rationally required premises for them. It is in part to avoid foreclosing the possibility of strong foundationalism that we have explicated the two senses of justification above in terms of what sort of justification premises can provide to a belief inferred from them, not in terms that might seem to apply to all beliefs a subject could hold. And as we shall see in the end, the Russellian solution to the Gettier problem does point to a need for foundations.

The disparity between the rationality of the subject in believing and the falsity of at least one important premise gives Gettier cases their distinctive characteristics: the subject’s belief is rational while the conclusion is “true for the wrong reason” and therefore “adventitiously true.” The requirement of true crucial premises for inferential knowledge together with the conferral of justification₁ upon the subject’s belief addresses both aspects of Gettier scenarios.

The externalist use of the Gettier problem

Before responding to attempted counterexamples, we need to understand how and why the Gettier problem has been taken to lead to externalism. That it has had the historical effect of motivating a move to various naturalist and externalist epistemologies is beyond doubt. Once the TJB definition of knowledge was called into question, the idea arose that the much-sought “fourth condition” that must be added had something to do with the circumstances in the real world under which the subject formed his belief, perhaps even circumstances about which the subject had no beliefs or evidence at all. While it is possible to add a requirement that these real-world circumstances be “right” on top of an ordinary internalist requirement for a good argument (or for internally accessible acquaintance with foundations), it was disturbing to traditional philosophers to think that some set of causes or circumstances – especially physical circumstances quite separate from the evidence available to the subject – might be indispensable to the subject’s epistemic state. Traditional internalism involves a (sometimes unexpressed) requirement that at least any empirical truth that must obtain for the subject to have knowledge must be a truth the subject has some reason to believe.

William Alston is emphatic about the impact of Gettier upon the internalist-externalist debate. He implies that the Gettier problem made it very difficult to define “access internalism” because, he says, even self-styled internalists have now acknowledged that there is something other than truth that is necessary for knowledge to which we do not have even “relatively direct” access.

Contemporary internalists who think that justification and truth are required for a belief’s counting as knowledge have been sufficiently impressed by Gettier to recognize that they are not sufficient. In addition, something must be there that will obviate Gettier problems. And, again, internalists accept, for good reasons, that we lack direct access to the satisfaction of these anti-Gettier conditions, e.g., the absence of any fact that when added to the justifying conditions would result in the belief’s not being sufficiently justified. Hence we can’t simply say: an epistemologist is an access internalist about knowledge provided he holds that anything that contributes to knowledge status other than truth is something to which we have relatively direct access. And so we still lack an acceptable general formulation.⁹

Philip Kitcher describes Gettier’s argument as, historically, a quiet but important harbinger of the rise of naturalism that was to follow, important in part because it “called into question” the traditional enterprise of explicating justification.

Psychology reentered epistemology quietly. A central problem in the analysis of knowledge takes for granted a conception of knowledge as justified true belief and seeks to provide an account of justification (foundationalist and coherence theories of justification being the main rivals). In 1963, a short article by Edmund Gettier called this enterprise into question by describing instances in which people have justified true belief but do not seem to have knowledge. Initial responses to Gettier’s problem usually followed the apsychologistic orthodoxy, attempting to impose logical conditions on the subject’s belief that would rule out the problematic examples as cases of knowledge. In the late 1960’s, however, a number of authors proposed that a solution to Gettier’s puzzling cases must lie in differentiating the causal processes that generate and sustain belief on those occasions where the subject knows. These generic approaches were articulated with the same kind of attention to detail that distinguished apsychologistic attacks on the Gettier problem. Yet, from a naturalistic perspective, their primary significance was their break with the apsychologistic tradition. Analyses of the concept of knowledge (and, later, of justification) . . . could take into account the processes . . . that causally generate states of belief.¹⁰

One of these early responses to Gettier was by Alvin Goldman, who in his seminal article, “A Causal Theory of Knowing,” proposes that Gettier problems can be solved satisfactorily only by requiring an appropriate causal connection between the subject’s belief that p and the fact that p. Goldman does argue in this article that it is not necessary for the subject to have any knowledge of this causal connection in the case of non-inferential beliefs (in which category he includes both perceptual and memory beliefs) and that the explication of the relevant causes is a job for the sciences. However, he retains a more traditional approach to inferential knowledge, envisaging in those cases a subject who believes that a particular causal process has taken place, which belief would be false if the situation were Gettierized. Indeed, he argues that the subject must “correctly reconstruct” at least the “important links” in the relevant causal chain to have knowledge based on inference.¹¹

In a more radical move, Peter Unger advanced a theory on which justification is not even necessary for knowledge. Unger’s substitute for justification, intended to cover Gettier cases as well as other putative counterexamples to the TJB analysis, is that it be “not at all accidental” that the subject’s belief is true. Unger applies this notion of accident directly to questions of statistical reasoning and imp...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Introduction
• 1 Internalism and the Collapse of the Gettier Problem
• 2 The Connection to Truth
• 3 Internalism, Externalism, and the Metaregress
• 4 What’s Wrong with Epistemic Circularity
• 5 Analytic a priori Knowledge
• 6 The Problem of Deduction
• 7 The Ground of Induction
• Notes
• Bibliography