Section II. Goldman Unlike Quine, Goldman did not begin his career in epistemology by attempting to develop a comprehensive account of how epistemology fits into the study of human behavior. Rather Goldman began by addressing himself to a very local epistemological issue to which Quine never addressed himself. Specifically, Goldman began work in epistemology by trying to solve the Gettier problem, i.e., by trying to explain what must be added to an account of empirical knowledge as justified, true belief in order to offer non-circular conditions for knowledge that are both necessary and sufficient. The Gettier problem can be briefly illustrated by means of the following example (but for more detailed discussion of the problem and the vast variety of proposed solutions, see the chapter from Richard Feldman in this volume):
Jones believes, on the basis of very compelling evidence that his coworker Smith owns a Ford. And so Jones has a justified belief that Smith owns a Ford. But this belief is false: despite all the evidence to the contrary, Smith does not in fact own a Ford. Nonetheless, Jones's coworker Brown owns a Ford. And so, when Jones infers from his belief that Smith owns a Ford to the conclusion that one of his coworkers owns a Ford, Jones comes to believe something true. Furthermore, this latter true belief is justified, since it is deduced (by means of an obviously valid deduction) from something else that Jones justifiably believes. So Jones has a justified and true belief that one of his coworkers owns a Ford. But Jones does not know that one of his coworkers owns a Ford. So justified true belief is not sufficient for knowledge.
It is possible to multiply such examples ad infinitumfor cases of empirical belief, but it is not clear that such cases can be constructed for non-empirical belief. In fact, Goldman took the traditional analysis of knowledge as justified, true belief to be correct for non-empirical knowledge, but he wanted to find an adequate analysis of empirical knowledge.
Roughly, on Goldman's earliest published view (in his 1967 essay "A Causal Theory of Knowing"), S has empirical knowledge that p if and only if S's true belief that p is causally connected in an appropriate way with the fact that p, where what counts as an appropriate causal connection is to be specified simply by appeal to an open list of examples (e.g., perception, memory, testimony from someone who perceives or remembers, and good inference from something perceived or remembered). Gettier cases fail to qualify as knowledge because they do not involve the appropriate sort of causal connection between the fact believed and the believer's belief.
While Goldman gave up this causal account of knowledge nine years later, he never gave up its central contention that, whether or not S knows that p (in the particular sense of "know" that interests epistemologists, Goldman now adds) depends upon the causal origination or causal sustenance of S's true belief that p. In fact, it was Goldman’s insistence upon this point, and his way of developing the point, that put his epistemology in contact with empirical science, and thereby made him a kind of “naturalist” in epistemology. For Goldman, knowing is a matter of having the right kind of causal history. Epistemology is supposed to tell us what very general sort of causal history is the “right” kind for knowing, but cognitive science is supposed to spell out the details, and tell us specifically which of our cognitive faculties or methods of belief-formation provide this right kind of causal history, and under what conditions. Thus, epistemology and cognitive science work together, on Goldman’s view, to provide us with a detailed account of what it is to know something, which is just an account of the causal history that constitutes knowledge.
In his 1976 essay "Discrimination and Perceptual Knowledge", Goldman presented what is generally regarded as a counterexample to his 1967 causal theory of knowledge, and attempted to accommodate the example by appeal to a view that he would eventually call "reliabilism", which is itself a version of the general idea that whether or not S knows that p depends upon the causal formation or causal sustenance of S’s true belief that p. Here is the example that Goldman presented (and which he credits to Carl Ginet): Henry is driving through the countryside looking at a barn. His vision is normal, and he sees the barn clearly, and thereby forms the belief "there's a barn there". Unbeknownst to Henry, however, this is fake barn country, and almost all of the apparent barns in this region happen to be mere barn facades; Henry just happened to be looking at one of the only real barns in the area. While Henry has a justified, true belief that there's a barn there, he does not know that there's a barn there. But notice that, in a normal situation in which Henry is driving through the countryside and sees a barn, he can come thereby to know that there's a barn there, even though the causal relation between his belief and the fact that there's a barn there is no different in the barn facade case than it is in the normal case. So Henry's failure to know, in the barn facade case, that there's a barn there, cannot be due to any lack of an appropriate causal connection between his belief that there's a barn there and the fact that there's a barn there. His failure to know must be due to something else. But what could it be due to? Goldman claims that it is due to the fact that Henry cannot reliably discriminate real barns from barn facades (at least not by looking at them from the distance and angle that he actually occupies), and that the exercise of this discriminatory ability is required for knowledge that there's a barn there when Henry is in barn facade country, but it is not required for such knowledge when Henry is normally situated. In short, knowing that p requires an ability reliably to discriminate between its being the case that p and its being the case that some relevant alternative to p obtains, where which alternatives count as "relevant" varies from context to context.
Goldman leaves it open whether the context that fixes the range of alternatives that are relevant to whether or not someone knows something is the context of the putative knower, or is rather the context of someone who is thinking about the putative knower. But, no matter how this issue is resolved, Goldman’s reliabilism involves the claim that knowing involves the exercise of a discriminative ability. And this is just one way of spelling out the very general claim that, for S to know that p, S’s true belief that p must have the right kind of causal history.
The 1976 view that I just described is a view about the requirements for perceptual knowledge, and it is a view according to which, as Goldman originally presented it, perceptual knowledge does not require that one's perceptual belief be justified. But by the time of his 1979 paper "What is Justified Belief?" Goldman reversed himself on this last point, and claimed that perceptual knowledge, like non-empirical knowledge, does require something that epistemologists call “doxastic justification”: in other words, it requires that the belief that constitutes knowledge is itself a justified belief. (For someone to have a justified belief is different from her having a justification to believe something: you can have a justification to believe something whether or not you actually believe it.) But doxastic justification, on Goldman’s view, itself requires a certain kind of causal reliability. On Goldman's view, for a belief to be justified involves its being formed by a reliable process. A reliable process is one that tends to produce a high ratio of true beliefs. It is empirically obvious that such processes include at least normal forms of perception, memory, inference, and testimony: this is why these processes are sources of doxastic justification, and so of knowledge.
Goldman elaborates the view just mentioned in much greater detail in his 1986 book Epistemology and Cognition. That book contains a detailed account of knowledge, as well as of doxastic justification. And the book also elaborates the relationship that Goldman takes epistemology to have to the empirical sciences of cognition. I'll start by explaining Goldman's account of knowledge, then proceed to his account of justification, and finally say something about how he conceives of the relation between epistemology and cognitive science.
Goldman's 1986 account of knowledge and of justification, like his earlier accounts, is a causal reliabilist account. According to Goldman’s 1986 account:
(a) S's believing p at t is permitted by a right system of J-rules, and
(b) this permission is not undermined by S's cognitive state at t." (Goldman 1986, 63)
This requires a bit of gloss: J-rules are rules according to which some beliefs are permitted and others are not. A system of J-rules is "right" just in case its dictates concerning which beliefs are permitted and which are not are correct (and such correctness may, for all we say here, be independent of whether or not anyone takes such dictates to be correct). Finally, there are a couple of ways in which a permission may be "undermined" by the believer's cognitive state. One way in which such undermining can occur is for the believer to have reason to believe that her belief that p is not justified. Another way in which such undermining can occur is for the believer simply to believe that her belief that p is not justified. In either case, S's belief that p is not justified.
Goldman fleshes out this schematic account of doxastic justification by imposing conditions on the rightness of a system of J-rules. A right system of J-rules must make the permissibility of a belief depend upon the process by which the belief is formed or sustained. More specifically, a right system of J-rules must make the permissibility of a belief depend upon the extent to which the processes by which the belief is formed or sustained tend to result in valuable consequences. And more specifically still, a right system of J-rules must make the permissibility of a belief depend upon the extent to which the processes by which the belief is formed or sustained tend to result in a high ratio of true to false beliefs. While a number of different accounts of J-rule rightness satisfy these conditions, Goldman favors the following schematic account:
"A J-rule system R is right if and only if
R permits certain (basic) psychological processes, and the instantiation of these processes would result in a truth ratio of beliefs that meets some specified high threshold (greater than .50)." (Epistemology and Cognition, 106)
When Goldman speaks of a "basic" psychological process, he means a process that is not acquired as a result of learning: it is either unacquired, or else acquired simply through maturation. And when Goldman speaks of the truth ratio that would result from the instantiation of such a process, he means the truth ratio that would result in normal worlds, i.e., in worlds that work in roughly the way that we take the actual world to work (whether or not the actual world does, in fact, work that way). Finally, the normal worlds in question, whether or not they include the actual world, will include a wide range of counterfactual worlds.
This, then, is Goldman's 1986 account of doxastic justification. What is his account of knowledge? For Goldman, knowledge involves true belief that is doxastically justified, and is formed by a process that is locally reliable, i.e., reliable specifically in the kind of setting in which the belief is actually formed, whether that kind of setting obtains in a world in which the belief is true, or in a world in which some relevant alternative to the belief is true.
Since knowledge and doxastic justification both involve reliability of one kind or another, and questions concerning the identity and reliability of our psychological processes are empirical questions, Goldman sees a close relationship between epistemology, on the one hand, and empirical questions concerning the identity and reliability of our psychological processes, on the other. What we know and what we’re justified in believing depends upon how we form our beliefs and how reliable those belief-forming processes are.
While Goldman has continued to refine a few details of his views concerning knowledge and justification, much of his work over the past quarter century has involved applications of his reliabilist account of knowledge and of justification to questions concerning how to organize social institutions so as to improve the production and dissemination of knowledge. Thus, in his 1999 book Knowledge in a Social World, Goldman discusses how the institutions of science, law, politics, education, and communication can be organized so as to maximize our acquisition of true belief, and minimize our acquisition of false belief. This book was seminal in creating the now very professionally active field of social epistemology.
Clearly, Quine and Goldman conceive of epistemology, and its connection to the empirical findings of psychology and cognitive science, very differently. Goldman devotes much of his epistemological work to addressing questions such as “what is knowledge?” or “what is it for a belief to be justified?”; these are questions that he takes to be answerable simply by a priori reflection. He also devotes much of his epistemological work to addressing questions such as “what are our sources of knowledge?” or “which sorts of beliefs are justified?”; these are questions that he takes to be answerable empirically, largely by means of cognitive science. Quine, in contrast, has nothing to say about any of these questions. For Quine, epistemology is simply the study of evidential support: what evidence do we have? How does evidence support theory? What principles guide proper theory choice given a body of evidence? These are the epistemological questions to which Quine devotes all of his attention. Goldman has very little to say about any of these questions. In fact, perhaps the only place in his corpus in which Goldman discusses any of these questions about evidence is in his 2010 paper “Williamson on Knowledge and Evidence”, in which he proposes that a person’s evidence set at a given time t consists of all and only those propositions that the person is non-inferentially propositionally justified in believing at t. But this proposal is offered not on the basis of any empirical considerations, but solely on the basis of its plausibility to reflection. In short, the epistemological question that Quine took to be an empirical question is one that Goldman barely treats at all, and when he does treat it, he treats it as an a priori question. And the epistemological questions that Goldman took to be empirical questions are questions that Quine simply never addressed.