The Primary Value Problem in Knowledge

Knowledge and true belief both tend to be things we want to have, but all else being equal, we tend to prefer to have knowledge over mere true belief. The Primary Value Problem is the problem of explaining why that should be the case. Many epistemologists think that we should take it as a criterion of adequacy for theories of knowledge that they be able to explain the fact that we prefer knowledge to mere true belief, or at least that they be consistent with a good explanation of why that should be the case.

To illustrate: suppose that Steve believes that the Yankees are a good baseball team, because he thinks that their pinstriped uniforms are so sharp-looking. Steve’s belief is true—the Yankees always field a good team—but he holds his belief for such a terrible reason that we are very reluctant to think of it as an item of knowledge.

Cases like Steve’s motivate the view that knowledge consists of more than just true belief. In order to count as knowledge, a belief has to be well justified in some suitable sense, and it should also meet a suitable Gettier-avoidance condition (see Gettier Problems). But not only do beliefs like Steve’s motivate the view that knowledge consists of more than mere true belief: they also motivate the view that knowledge is better to have than true belief. For suppose that Yolanda knows the Yankees’ stats, and on that basis she believes that the Yankees are a good team. It seems that Yolanda’s belief counts as an item of knowledge. And if we compare Steve and Yolanda, it seems that Yolanda is doing better than Steve; we’d prefer to be in Yolanda’s epistemic position rather than in Steve’s. This seems to indicate that we prefer knowledge over mere true belief.

The challenge of the Primary Value Problem is to explain why that should be the case. Why should we care about whether we have knowledge instead of mere true belief? After all, as is often pointed out, true beliefs seem to bring us the very same practical benefits as knowledge. (Steve would do just as well as Yolanda betting on the Yankees, for example.) Socrates makes this point in the Meno, arguing that if someone wants to get to Larisa, and he has a true belief but not knowledge about which road to take, then he will get to Larisa just as surely as if he had knowledge of which road to take. In response to Socrates’s argument, Meno is moved to wonder why anyone should care about having knowledge instead of mere true belief. (Hence, the Primary Value Problem is sometimes called the Meno Problem.)

So in short, the problem is that mere true beliefs seem to be just as likely as knowledge to guide us well in our actions. But we still seem to have the persistent intuition that any given item of knowledge is more valuable than the corresponding item of mere true belief. The challenge is to explain this intuition. Strategies for addressing this problem can either try to show that knowledge really is always more valuable than corresponding items of mere true belief, or else they can allow that knowledge is sometimes (or even always) no more valuable than mere true belief. If we adopt the latter kind of response to the problem, it is incumbent on us to explain why we should have the intuition that knowledge is more valuable than mere true belief, in cases where it turns out that knowledge isn’t in fact more valuable. Following Pritchard (2008; 2009), we can call strategies of the first kind vindicating, and we can call strategies of the second kind revisionary.

There isn’t a received view among epistemologists about how we ought to respond to the Primary Value Problem, so the most useful thing to do at this point is to consider a number of the more interesting proposals from the literature, and to look at their problems and prospects.

i. Knowledge as Mere True Belief

A very straightforward way to respond to the problem is to deny one of the intuitions on which the problem depends, the intuition that knowledge is distinct from true belief.  Meno toys with this idea in the Meno, though Socrates disabuses him of the idea. (Somewhat more recently, Sartwell (1991; 1992) has defended this approach to knowledge.) If knowledge is identical with true belief, then we can simply reject the value problem as resting on a mistaken view of knowledge. If knowledge is true belief, then there’s no discrepancy in value to explain.

The view that knowledge is just true belief is almost universally rejected, however, and with good reason. Cases where subjects have true beliefs but lack knowledge are so easy to construct and so intuitively obvious that identifying knowledge with true belief represents an extreme departure from what most epistemologists and laypeople think of knowledge. Consider once again Steve’s belief that the Yankees are a good baseball team, which he holds because he thinks their pinstriped uniforms are so sharp. It seems like an abuse of language to call Steve’s belief an item of knowledge. At the very least, we should be hesitant to accept such an extreme view until we’ve exhausted all other theoretical options.

Of course it could still be the case that knowledge is no more valuable than mere true belief, even though knowledge is not identical with true belief. But, as we’ve seen, there is a widespread and resilient intuition that knowledge is more valuable than mere true belief (recall, for instance, that we tend to think that Yolanda’s epistemic state is better than Steve’s). If knowledge were identical with true belief, then we would have to take that intuition to be mistaken; but, since we can see that knowledge is more than mere true belief, we can continue looking for an acceptable account which would explain why knowledge is more valuable than mere true belief.

ii. Stability

Most attempts to explain why knowledge is more valuable than mere true belief proceed by identifying some condition which must be added to true belief in order to yield knowledge, and then explaining why that further condition is valuable. Socrates’s own view, at least as presented in the Meno, is that knowledge is true opinion plus an account of why the opinion is true (where the account of why it is true is itself already present in the soul; it must only be recalled from memory). So, Socrates proposes, a known true belief will be more stable than a mere true belief, because having an account of why a belief is true helps to keep us from losing it. If you don’t have an account of why a proposition is true, you might easily forget it, or abandon your belief in it when you come across some reason for doubting it. But if you do have an account of why a proposition is true, you likely have a greater chance of remembering it, and if you come across some reason for doubting it, you’ll have a reason available to you for continuing to believe it.

A worry for this solution is that it seems to be entirely possible for a subject S to have some entirely unsupported beliefs, which do not count as knowledge, but where S clings to these beliefs dogmatically, even in the face of good counterevidence. S’s belief in a case like this can be just as stable as many items of knowledge—indeed, dogmatically held beliefs can even be more stable than knowledge. For if you know that p, then presumably your belief is a response to some sort of good reason for believing that p. But if your belief is a response to good reasons, then you’d likely be inclined to revise your belief that p, if you were to come across some good evidence for thinking that p is false, or for thinking that you didn’t have any good reason for believing that p in the first place. On the other hand, if p is something you cling to dogmatically (contrary evidence be damned), then you’ll likely retain p even when you get good reason for doubting it. So, even though having stable true beliefs is no doubt a good thing, knowledge isn’t always more stable than mere true belief, and an appeal to stability does not seem to give us an adequate explanation of the extra value of knowledge over mere true belief.

One way to defend the stability response to the value problem is to hold that knowledge is more stable than mere true beliefs, but only for people whose cognitive faculties are in good working order, and to deny that the cognitive faculties of people who cling dogmatically to evidentially unsupported beliefs are in good working order (Williamson 2000). This solution invites the objection, however, that our cognitive faculties are not all geared to the production of true beliefs. Some cognitive faculties are geared towards ensuring our survival, and the outputs of these latter faculties might be held very firmly even if they are not well supported by evidence. For example, there could be subjects with cognitive mechanisms which take as input sudden sounds and generate as output the belief that there’s a predator nearby. Mechanisms like these might very well generate a strong conviction that there’s a predator nearby. Such mechanisms would likely yield many more false positive predator-identifications than they would yield correct identifications, but their poor true-to-false output-ratio doesn’t prevent mechanisms of this kind from having a very high survival value, as long as they do correctly identify predators when they are present. So it’s not really clear that knowledge is more stable than mere true beliefs, even for mere true beliefs which have been produced by cognitive systems which are in good working order, because it’s possible for beliefs to be evidentially unsupported, and very stable, and produced by properly functioning cognitive faculties, all at the same time. (See Kvanvig 2003, ch1. for a critical discussion of Williamson’s appeal to stability.)

iii. Virtues

Virtue epistemologists are, roughly, those who think that knowledge is true belief which is the product of intellectual virtues. They seem to have a plausible solution to the Primary (and, as we’ll see, to the Secondary) Value Problem.

According to a prominent strand of virtue epistemology, knowledge is true belief for which we give the subject credit (Greco 2003), or true belief which is a cognitive success because of the subject’s exercise of her relevant cognitive ability (Greco 2008; Sosa 2007). For example (to adapt Sosa’s analogy): an archer, in firing at a target, might shoot well or poorly. If she shoots poorly but hits the target anyway (say, she takes aim very poorly but sneezes at the moment of firing, and luckily happens to hit the target), her shot doesn’t display skill, and her hitting the target doesn’t reflect well on her. If she shoots well, on the other hand, then she might hit the target or miss the target. If she shoots well and misses the target, we will still credit her with having made a good shot, because her shot manifests skill. If she shoots well and hits the target, then we will credit her success to her having made a good shot—unless there were intervening factors which made it the case that the shot hit the mark just as a matter of luck. For example: if a trickster moves the target while the arrow is in mid-flight, but a sudden gust of wind moves the arrow to the target’s new location, then in spite of the fact that the archer makes a good shot, and she hits the target, she doesn’t hit the target because she made a good shot. She was just lucky, even though she was skillful. But when strange factors don’t intervene, and the archer hits the target because she made a good shot, we give her credit for having hit the target, since we think that performances which succeed because they are competent are the best kind of performances. And, similarly, when it comes to belief-formation, we give people credit for getting things right as a result of the exercise of their intellectual virtues: we think it’s an achievement to get things right as the result of one’s cognitive competence, and so we tend to think that there’s a sense in which people who get things right because of their intellectual competence deserve credit for getting things right.

According to another strand of virtue epistemology (Zagzebski 2003), we don’t think of knowledge as true belief which meets some further condition. Rather, we should think of knowledge as a state which a subject can be in, which involves having the propositional attitude of belief, but which also includes the motivations for which the subject has the belief. Virtuous motivations might include things like diligence, integrity, and a love of truth. And, just as we think that, in ethics, virtuous motives make actions better (saving a drowning child because you don’t want children to suffer and die is better than saving a drowning child because you don’t want to have to give testimony to the police, for example), we should also think that the state of believing because of a virtuous motive is better than believing for some other reason.

Some concerns have been raised for both strands of virtue epistemology, however. Briefly, a worry for the Sosa/Greco type of virtue epistemology is that (as we’ll see in section 3) knowledge might not after all in general be an achievement—it might be something we can come by in a relatively easy or even lazy fashion. A worry for Zagzebski’s type of virtue epistemology is that there seem to be possible cases where subjects can acquire knowledge even though they lack virtuous intellectual motives. Indeed, it seems possible to acquire knowledge even if one has only the darkest of motives: if a torturer is motivated by the desire to waterboard people until they go insane, for example, he can thereby gain knowledge of how long it takes to break a person by waterboarding.

Still, the idea that knowledge can be analyzed as true belief which is somehow virtuously produced and creditable to the agent seems to be worth pursuing. Because the virtue-approach seems to be able to handle most of the Gettier-style problems which plague previous analyses of knowledge, and because it can provide what is on the face of it a plausible solution to the Primary Value Problem, virtue epistemology represents a promising research program, and its problems and prospects deserve careful exploration.

iv. Reliabilism

The Primary Value Problem is sometimes thought to be especially bad for reliabilists about knowledge. Reliabilism in its simplest form is the view that beliefs are justified if and only if they’re produced by reliable processes, and they count as knowledge if and only if they’re produced by reliable processes and they’re not Gettiered. (See, for example, Goldman and Olsson (2009, p. 29 ) The apparent trouble for reliabilism is that reliability only seems to be valuable as a means to truth—so, in any given case where we have a true belief, it’s not clear that the reliability of the process which produced the belief is able to add anything to the value that the belief already has in virtue of being true. The value which true9 beliefs have in virtue of being true completely “swamps” the value of the reliability of their source, if reliability is only valuable as a means to truth. (Hence the Primary Value Problem for reliabilism has often been called the “swamping problem.”)

To illustrate with an example borrowed from Zagzebski (2003): the value of a cup of coffee seems to be a matter of how good the coffee tastes. And we value reliable coffeemakers because we value good cups of coffee. But when it comes to the value of any particular cup of coffee, its value is just a matter of how good it tastes; whether the coffee was produced by a reliable coffeemaker doesn’t add to or detract from the value of the cup of coffee. Similarly, we value true beliefs, and we value reliable belief-forming processes because we care about getting true beliefs. So we have reason to prefer reliable processes over unreliable ones. But whether a particular belief was reliably or unreliably produced doesn’t seem to add to or detract from the value of the belief itself.

Responses have been offered on behalf of reliabilism. Brogaard (2006) points out that critics of reliabilism seem to have been presupposing a Moorean conception of value, according to which the value of an object (or state, condition, and so forth) is entirely a function of the internal properties of the object. (The value of the cup of coffee is determined entirely by its internal properties, not by the reliability of its production, or by the fineness of a particular morning when you enjoy your coffee.) But this is a mistaken view about value in general. External features can add value to objects. We value a genuine Picasso painting more than a flawless counterfeit, for example. If that’s correct, then extra value can be conferred on an object, if it has a valuable source, and perhaps the value of reliable processes can transfer to the beliefs which they produce.

(That is a negative response to the value problem for reliabilism, in the sense that its aim is to show that critics of reliabilism haven’t shown that reliabilists can’t account for the value of knowledge.)

Goldman and Olsson (2009) offer two further responses on behalf of reliabilism. Their first response is that we can hold that true belief is always valuable, and that reliability is only valuable as a means to true belief, but that it is still more valuable to have knowledge (understood as reliabilists understand knowledge, that is, as reliably-produced and unGettiered true belief) than a mere true belief. For if S knows that p in circumstances C, then S has formed the belief that pthrough some reliable process in C. So S has some reliable process available to her, and it generated a belief in C. This makes it more likely that S will have a reliable process available to her in future similar circumstances, than it would be if S had an unreliably produced true belief in C. So, when we’re thinking about how valuable it is to be in circumstances C, it seems to be better for S to be in C if S has knowledge in C than if she has mere true belief in C, because having knowledge in C makes it likelier that she’ll get more true beliefs in future similar circumstances.

This response, Goldman and Olsson think, accounts for the extra value which knowledge has in many cases. But there will still be cases where S’s having knowledge in C doesn’t make it likelier that she’ll get more true beliefs in the future. For example, C might be a unique set of circumstances which is unlikely to come up again. Or S might be employing a reliable process which is available to her in C, but which is likely to become unavailable to her very soon. Or S might be on her deathbed. So this response isn’t a completely validating solution to the value problem, and it’s incumbent on Goldman and Olsson to explain why we should tend to think that knowledge is more valuable than mere true belief in those cases when it’s not.

So Goldman and Olsson offer a second response to the Primary Value Problem: when it comes to our intuitions about the value of knowledge, they argue, it’s plausible that these intuitions began long ago with the recognition that true belief is always valuable in some sense to have, and that knowledge is usually valuable because it involves both true belief and the probability of getting more true beliefs; and then, over time, we have come to simply think that knowledge is valuable, even in cases when having knowledge doesn’t make it more probable that the subject will get more true beliefs in the future.

v. Contingent Features of Knowledge

An approach similar to Goldman and Olsson’s is to consider the values of contingent features of knowledge, rather than the value of its necessary and/or sufficient conditions. Although we might think that the natural way to account for the value of some state or condition S1, which is composed of other states or conditions S2-Sn, is in terms of the values of S2-Sn, perhaps S1 can be valuable in virtue of some other conditions which typically (but not always) accompany S1, or in terms of some valuable result which S1 is typically (but not always) able to get us. For example: it’s normal to think that air travel is valuable, because it typically enables people to cover great distances safely and quickly. Of course, sometimes airplanes are diverted, and slow travellers down, and sometimes airplanes crash. But even so, we might continue to think, air travel is typically a valuable thing, because in ordinary cases, it gets us something good.

Similarly, we might think that knowledge is valuable because we need to rely on the information which people give us in order to accomplish just about anything in this life, and being able to identify people as having knowledge means being able to rely on them as informants. And we also might think that there’s value in being able to track whether our own beliefs are held on the basis of good reasons, and we typically have good reasons available to us for believing p when we know that p. Of course we aren’t always in a position to identify when other people have knowledge, and if externalists about knowledge are right, then we don’t always have good reasons available to us when we have knowledge ourselves. Nevertheless, we can typically identify people as knowers, and we can typically identify good reasons for the things we know. These things are valuable, so they make typical cases of knowledge valuable, too. (See Craig (1990) for an account of the value of knowledge in terms of the characteristic function of knowledge-attribution. Jones (1997) further develops the view.)

Like Goldman and Olsson’s responses, this strategy for responding to the value problem doesn’t give us an account of why knowledge is always more valuable than mere true belief. For those who think that knowledge is always preferable to mere true belief, and who therefore seek a validating solution to the Primary Value Problem, this strategy will not be satisfactory. But for those who are willing to accept a somewhat revisionist response, according to which knowledge is only usually or characteristically preferable to mere true belief, this strategy seems promising.

b. The Secondary Value Problem

Suppose you’ve applied for a new position in your company, but your boss tells you that your co-worker Jones is going to get the job. Frustrated, you glance over at Jones, and see that he has ten coins on his desk, and you then watch him put the coins in his pocket. So you form the belief that the person who will get the job has at least ten coins in his or her pocket (call this belief “B”). But it turns out that your boss was just toying with you; he just wanted to see how you would react to bad news. He’s going to give you the job. And it turns out that you also have at least ten coins in your pocket.

So you have a justified true belief, B, which has been Gettiered. In cases like this, once you’ve found out that you were Gettiered, it’s natural to react with annoyance or intellectual embarrassment: even though you got things right (about the coins, though not about who would get the job), and even though you had good reason to think you had things right, you were just lucky in getting things right.

If this is correct—if we do tend to prefer to have knowledge over Gettiered justified true beliefs—then this suggests that there’s a second value problem to be addressed. We seem to prefer having knowledge over having any proper subset of the parts of knowledge. But why should that be the case? What value is added to justified true beliefs, when they meet a suitable anti-Gettier condition?

i. No Extra Value

An initial response is to deny that knowledge is more valuable than mere justified true belief. If we’ve got true beliefs, and good reasons for them, of course we might be Gettiered, if for some reason it turns out that we’re just lucky in having true beliefs. When we inquire into whether p, we want to get to the truth regarding p, and we want to do so in a rationally defensible way. If it turns out that we get to the truth in a rationally defensible way, but strange factors of the case undermine our claim to knowing the truth about p, perhaps it just doesn’t matter that we don’t have knowledge.

Few epistemologists have defended this view, however (though Kaplan (1985) is an exception). We do after all find it irritating when we find out that we’ve been Gettiered; and when we are considering corresponding cases of knowledge and of Gettiered justified true belief, we tend to think that the subject who has knowledge is better off than the subject who is Gettiered. Of course we might be mistaken; there might be nothing better in knowledge than in mere justified true belief. But the presumption seems to be that knowledge is more valuable, and we should try to explain why that is so. Skepticism about the extra value of knowledge over mere justified true belief might be acceptable if we fail to find an adequate explanation, but we shouldn’t accept skepticism before searching for a good explanation.

ii. Virtues

We saw above that some virtue epistemologists think of knowledge in terms of the achievement of true beliefs as a result of the exercise of cognitive skills or virtues. And we do generally seem to value success that results from our efforts and skills (that is, we value success that’s been achievedrather than stumbled into (for example, Sosa (2003; 2007) and Pritchard (2009)). So, because we have a cognitive aim of getting to the truth, and we can achieve that aim either as a result of luck or as a result of our skillful cognitive performance, it seems that the value of achieving our aims as a result of a skillful performance can help explain why knowledge is more valuable than mere true belief.

That line of thought works just as well as a response to the Secondary Value Problem as to the Primary Value Problem. For in a Gettier case, the subject has a justified true belief, but it’s just as a result of luck that she arrived at a true belief rather than a false one. By contrast, when a subject arrives at a true belief because she has exercised a cognitive virtue, it’s plausible to think that it’s not just lucky that she’s arrived at a true belief; she gets credit for succeeding in the aim of getting to the truth as a result of her skillful performance. So cases of knowledge do, but Gettier cases do not, exemplify the value of succeeding in achieving our aims as a result of a skillful performance.