Edward Craig on the social role of knowledge ascriptions

In a previous post, I asked of the concept of epistemic rationality the same question that Edward Craig asks of the concept of knowledge in Knowledge and the State of Nature—that is, I asked why we ascribe epistemic rationality to individuals or to their beliefs and credences. In passing, I noted that I think Craig can’t be quite right about the social role of knowledge ascriptions. But I think my discussion was too brief and was open to misunderstanding. So here I will try to state it more clearly.

Craig’s contention is that “the concept of knowledge is used to flag approved sources of information” (10). As he puts it: “Human beings need true beliefs about their environment […] That being so, they need sources of information that will lead them to believe truths” (10). We have sensory apparatus to help us learn directly from the environment and build up our own stock of beliefs, but, as Craig notes, “It will be highly advantageous to them if they can also tap the primary stocks of their fellows.” (10). So we need to be able to identify from whom we can learn these true beliefs we need. And to do that, we use ascriptions of knowledge.

Meadow brown butterfly, Shapwick Heath, Somerset

Of course, as Craig is well aware, that suggestion immediately raises a query. Let’s imagine the chain of events: I learn that S knows p; learning that flags S as an approved source of information about p; then I learn that S believes p; and from that and the fact that S knows p, I learn that p is true. But of course this chain contains a lot of redundant steps. Since (nearly) all agree that S knows p only if p is true, when I learn that S knows p, I automatically learn p is true. There was no need to flag S as an approved source of information. Indeed, there was no need to involve S at all. If this is how Craig imagines our social learning proceeds, we have no need for the concept of knowledge—the concept of truth will do all we need. The above chain might be truncated as follows: I learn that p is true.

As I said, Craig is quite aware of this. And this is why his primary target is not ascriptions of the form ‘S knows p’. Instead, he’s interested primarily in ascriptions of the form ‘S knows whether p’. The idea is this: I am inquiring into whether p, and I want to know where my inquiry should lead me next. I learn that S knows whether p. Then, if I think S is honest, accessible, and generous enough to share their beliefs with me, my next order of business should be to ask S about p.

However, as Craig is also aware, a similar worry might be raised about this proposal. After all, if the purpose of ascriptions of knowledge-whether is to flag individuals as approved sources of information from which we might learn whether p, then surely knowledge-whether requires only true belief concerning p. That is, in order to know that I can learn whether p from S, I need only know that the following condition is satisfied:

(Truth) Either (i) S believes p and p is true or (ii) S believes not-p and p is false.

But knowledge-whether is usually taken to require something more than this. It is usually taken also to require a modal version of Truth, such as for instance Nozick’s tracking condition:

(Tracking) (i) In nearby worlds at which p is true, S believes p, and (ii) in nearby worlds where p is false, S believes not-p.

But note that there’s no need to know that S satisfies Tracking if one wishes to learn about p from S. One need only know that Truth holds. Craig seems aware of this concern, and he responds as follows—I quote at length as much as anything to assure myself there isn’t something I’m missing from Craig’s response:

Why should our inquirer be interested in what is the case in possible worlds? After all, he wants to be told the truth in this world, the actual world, so whence the interest in other, and merely possible, worlds, however ‘close’ they may be to this one? […] But this is too quick: there is a line of thought which shows that the inquirer cannot help being interested in the contents of possible worlds as well as those of the actual.

We have to remember that the inquirer's knowledge of the actual world is bound to be highly incomplete. It is not only that he doesn't yet know whether p; there will be all sorts of things about himself, the environment and the potential informant of which he is ignorant. There are, in other words, enormously many propositions such that he does not know whether A or not‐A, whether B or not‐B, and so on. So if we think of a world as defined by the totality of what is true in it, there are indefinitely many different possible worlds any one of which, so far as he knows, might be the actual world. His concern with getting the right information in the actual world will therefore lead him to hope for an informant who will give him the truth about p whichever of all these possibilities is realised. Which is to say, if you like the jargon, that he wants an informant who will give him the right answer in a range of possible worlds. (19-20)

But I think this just gets things wrong. To learn whether p from S it doesn’t matter whether S knows p—all that matters is whether Truth holds. To see this in a bit more detail, let’s get a bit more precise. Firstly, let’s ask what attitude you want to end up with at the end of your inquiry into p: is it just true belief concerning p (i.e. belief in p if p is true, and belief in not-p if p is false)? Or is it knowledge about p? Secondly, let’s ask what condition do you learn that S satisfies by learning they know whether p: is it only Truth or is it Truth + Tracking (or something similar)? And thirdly, let’s ask what attitude you get towards this condition when you learn S satisfies it: true belief that S satisfies it, or knowledge that S satisfies it?

Here are three claims:

(1) If I believe truly that S satisfies Truth, and I learn S believes p, then I thereby come to believe truly that p.

(2) If I know S satisfies Truth concerning p, and then I learn S believes p, then I thereby come to know p.

(2) If I believe truly that S satisfies Truth + Tracking, and then I learn S believes p, then I don’t necessarily thereby come to know p.

(1) is obvious. (2) is a consequence of the fact that knowledge transmits across known implication, for Truth entails that, if S believes p, then p. (3) is perhaps a little more controversial. The idea is that, if you believe that S satisfies Truth + Tracking just because of wishful thinking or because of some stereotype bias or because you’re high and believe everyone knows every truth in the universe, and you somehow luck out and they actually do satisfy Truth + Tracking, then you acquire a true belief that p when you learn that S believes p, but you can’t thereby acquire knowledge that p. Perhaps you might deny this, but let’s grant it for the moment.

Then if you care about getting mere true belief about p, then getting true belief about Truth will suffice. If getting knowledge about p, then knowledge of Truth will suffice—you don’t need knowledge of Truth + Tracking. And if you care about getting knowledge of p, then merely believing Truth + Tracking won’t suffice—you need knowledge of Truth + Tracking. So, in the end, Tracking is superfluous. When you’ve got knowledge the relevant condition, you don’t need it; and when you’ve got mere true belief, it doesn’t help you.

Caterpillar, Shapwick Heath, Somerset

You might think that, in order to come to know that S has a true belief about p, without knowing p itself, you’d have to learn it from someone who genuinely knows p. Someone would have to say to you: ‘Look, I know you’re interested in whether p. The person over there…the one in the pillbox hat…they’re right about p. Go ask them about it!’ And for you to come to know that the pillbox hat wearer has a true belief about p, the person who told you that would have to know that. So, at some point in the process, you have to ascribe genuine knowledge, which demands both Truth and Tracking.

But that isn’t so. We often learn that someone is right about something—that is, someone satisfies Truth with respect to a proposition—without learning it from anyone else. I learn that Jay was right about the future direction of the stocks they hold because I see their bank balance go up; I learn that Tara was right about which direction was quickest to get to our shared destination, because I see her there when we arrive; I learn that Bev was right about each of the questions on the online multiple choice test because I see he scored full marks when the computer spits out the answer. In each of these cases, I learn the person satisfies Truth with respect to p without learning whether p is true or false; and I don’t learn that they satisfy Truth + Tracking—Jay’s bank balance would have gone up whether he was right about p because of sound sober reasoning or because by pure chance he heard it uttered to him in a dream. Indeed, I come to know the person satisfies Truth. And so I acquire all I need to learn from them and thereby acquire knowledge myself.

So I think Craig’s account doesn’t work. But I think there is an alternative nearby that does explain the social practice of ascribing knowledge. As is emphasised in much of the literature on the analysis of knowledge, saying that S knows p tells us first of all that S believes p, but it also tells us something about the way in which S came to acquire that belief and the stock of evidence on which it was based, as well as something about how that belief-forming method relates to the world. We thereby learn about some of S’s cognitive machinery. Of course, this is no guarantee S will ever form beliefs in the same way again, but combined with a bit of background knowledge about how human’s work, it is pretty strong evidence they will. And if what we learn about this belief-forming mechanism and the stock of evidence on which it worked is that it tracks the truths in the world in some way, we gain some reason to learn from this person in the future. So, for instance, if I learn that Dani not only truly believes a certain fact about oceanography, but also knows it, I learn she has a body of evidence and a belief-forming mechanism such that when the latter is applied to the former, she gets knowledge. And this gives me more reason to think in the future that her opinions about similar topics will be correct than if I learned only that she has a true belief about that fact from oceanography.

An advantage of this account is that it explains why we need the modal condition on knowledge in a way that Craig’s account doesn’t. But a disadvantage is that it seems to require quite a strong modal or anti-luck condition. After all, if S just happened to use a reliable or truth-tracking belief-forming method to arrive at their belief that p, and could easily have used a really unreliable one, then learning that they know p doesn’t give you much confidence in their future pronouncements. So you need the concept of knowledge to rule this out if you’re going to use it in the way I’m suggesting. Maybe our actual concept of knowledge does require this stronger modal or anti-luck condition—I don’t feel I’m in a position to adjudicate this.