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.
I want to say you're a bit too quick to dismiss the importance of something like tracking even in the one-off case (ie, it doesn't *only* matter when it comes to deciding whether to rely on someone in the future). Here's a sketchy way of putting a point that I think Craig is getting at.
It's hard/rare/unrealistic to be in a situation where (1) I'm ignorant as to whether P, (2) I know somebody else has a true belief as to whether P, but (3) I don't have good reason to think they satisfy tracking with respect to P.
Why? Given my ignorance as to whether P, if I'm going to know that somebody else (call him Richard) is right about whether P, it needs to be that Richard is right in all the epistemic possibilities open to me. And since I'm ignorant as to whether P, that will include both P-possibilities and not-P possibilities. Normal ways of getting into that spot--of being reasonably confident that whichever way it turns out, P or not-P, Richard will have gotten it right--will involve knowing that Richard satisfies tracking with respect to P. Normally, the epistemic route to thinking they're getting it right in *this* case is knowing that they *would* get it right across a range of cases; if I'm relatively ignorant about the subject matter, it's easier to first know the latter (e.g., by learning about Richard's methods), and to base my knowledge of the former on it, than to know the former via some other route.
Thanks, Daniel!! I agree with all that. And I think that sort of articulates what Craig is getting wrong? He's mistaking a pretty normal way of establishing a concept applies for the content of the concept itself. I agree that we often do establish the stronger claim that S's belief is modally sensitive to the truth regarding p when what we really need to establish is that S's belief is correct regarding p. But since we only need a concept that covers the weaker condition in order to flag people as testifiers to learn from in a one-off case, why have a concept that covers the stronger condition? We'd be throwing away free knowledge because we wouldn't be flagging people correctly.
I think this is where the squishiness of the conceptual genealogy project can help him. Just like Nietzsche can say the function of moral concepts is to wield ressentiment against the powerful, even though that's not baked into their content, Craig can say the function of the knowledge concept is to identify good informants, even if the content is strictly stronger than it needs to be to play that function.
Why the gap between the content needed to play the function and the actual content? Maybe the concept with stronger modal conditions (K) is a more psychologically natural kind--easier to teach/learn--than the one that packs in only the strictly necessary content (TB).
Yeah, I see how he might go that way, but I think it really reduces the force of his account. It's sort of hard to believe that the concept of being correct about p is trickier to learn that the concept of knowing whether p. I suppose some of the Williamson stuff pushes in the direction of thinking it's somehow the natural concept in the area, but I've never found that stuff convincing.
This seems to tie into ideas that I got from Jennifer Nagel's Locke lectures last year. She pointed out that many animals seem to have something like a concept of knowledge - complete with Gettier condition!
They have this concept because it's useful to know who in your environment is a knower - if you see that there's a treat being placed somewhere behind a wall, and you can see that someone else is on that side of the wall and watching, then you can follow where that person is reaching, if you want to get at the treat yourself. (This is the easy way that you can know that S knows whether p, without yourself knowing p.)
This concept is also often useful for predicting how someone else will behave - if you know that S knows where the treat is, then you also know that S getting an opportunity to go for the treat will give you an opportunity to jump in and take S's nice perch while they're off getting the treat.
Evidence that animals seem to have a Gettier condition is obtained by seeing when animals are surprised. If you see the treat go into a box, while you see the other animal watching, then you treat the other animal as a knower. If something blocks the other animal's view and reach for a minute, but then goes away, then you'll expect them to reach for the box with the treat, and be surprised if they reach for a different box. If the boxes both open to remind you where the treat is while the other animal is blocked, nothing changes. But if the treat comes out of the box and goes back in while the other animal is blocked, you now think the other animal has lost track, and are equally surprised whichever box they reach for.
Of course, she argues that they don't really have a Gettier condition - that's an even more sophisticated concept than belief, and truth, which are themselves concepts that only arise once animals have the cognitive sophistication to represent abstract contents that can be true or false, and believed or not believed. (It seems that 5 year old humans have this concept, but few if any other animals do, or younger humans.) Instead, most of these animals just have a representation of others as knowers or non-knowers, and treat various kinds of impediments as reasons to represent the other as a non-knower, without allowing them to represent the other as a false-believer or a true-but-Gettiered-believer.
I want to say you're a bit too quick to dismiss the importance of something like tracking even in the one-off case (ie, it doesn't *only* matter when it comes to deciding whether to rely on someone in the future). Here's a sketchy way of putting a point that I think Craig is getting at.
It's hard/rare/unrealistic to be in a situation where (1) I'm ignorant as to whether P, (2) I know somebody else has a true belief as to whether P, but (3) I don't have good reason to think they satisfy tracking with respect to P.
Why? Given my ignorance as to whether P, if I'm going to know that somebody else (call him Richard) is right about whether P, it needs to be that Richard is right in all the epistemic possibilities open to me. And since I'm ignorant as to whether P, that will include both P-possibilities and not-P possibilities. Normal ways of getting into that spot--of being reasonably confident that whichever way it turns out, P or not-P, Richard will have gotten it right--will involve knowing that Richard satisfies tracking with respect to P. Normally, the epistemic route to thinking they're getting it right in *this* case is knowing that they *would* get it right across a range of cases; if I'm relatively ignorant about the subject matter, it's easier to first know the latter (e.g., by learning about Richard's methods), and to base my knowledge of the former on it, than to know the former via some other route.
Thanks, Daniel!! I agree with all that. And I think that sort of articulates what Craig is getting wrong? He's mistaking a pretty normal way of establishing a concept applies for the content of the concept itself. I agree that we often do establish the stronger claim that S's belief is modally sensitive to the truth regarding p when what we really need to establish is that S's belief is correct regarding p. But since we only need a concept that covers the weaker condition in order to flag people as testifiers to learn from in a one-off case, why have a concept that covers the stronger condition? We'd be throwing away free knowledge because we wouldn't be flagging people correctly.
I think this is where the squishiness of the conceptual genealogy project can help him. Just like Nietzsche can say the function of moral concepts is to wield ressentiment against the powerful, even though that's not baked into their content, Craig can say the function of the knowledge concept is to identify good informants, even if the content is strictly stronger than it needs to be to play that function.
Why the gap between the content needed to play the function and the actual content? Maybe the concept with stronger modal conditions (K) is a more psychologically natural kind--easier to teach/learn--than the one that packs in only the strictly necessary content (TB).
Yeah, I see how he might go that way, but I think it really reduces the force of his account. It's sort of hard to believe that the concept of being correct about p is trickier to learn that the concept of knowing whether p. I suppose some of the Williamson stuff pushes in the direction of thinking it's somehow the natural concept in the area, but I've never found that stuff convincing.
This seems to tie into ideas that I got from Jennifer Nagel's Locke lectures last year. She pointed out that many animals seem to have something like a concept of knowledge - complete with Gettier condition!
They have this concept because it's useful to know who in your environment is a knower - if you see that there's a treat being placed somewhere behind a wall, and you can see that someone else is on that side of the wall and watching, then you can follow where that person is reaching, if you want to get at the treat yourself. (This is the easy way that you can know that S knows whether p, without yourself knowing p.)
This concept is also often useful for predicting how someone else will behave - if you know that S knows where the treat is, then you also know that S getting an opportunity to go for the treat will give you an opportunity to jump in and take S's nice perch while they're off getting the treat.
Evidence that animals seem to have a Gettier condition is obtained by seeing when animals are surprised. If you see the treat go into a box, while you see the other animal watching, then you treat the other animal as a knower. If something blocks the other animal's view and reach for a minute, but then goes away, then you'll expect them to reach for the box with the treat, and be surprised if they reach for a different box. If the boxes both open to remind you where the treat is while the other animal is blocked, nothing changes. But if the treat comes out of the box and goes back in while the other animal is blocked, you now think the other animal has lost track, and are equally surprised whichever box they reach for.
Of course, she argues that they don't really have a Gettier condition - that's an even more sophisticated concept than belief, and truth, which are themselves concepts that only arise once animals have the cognitive sophistication to represent abstract contents that can be true or false, and believed or not believed. (It seems that 5 year old humans have this concept, but few if any other animals do, or younger humans.) Instead, most of these animals just have a representation of others as knowers or non-knowers, and treat various kinds of impediments as reasons to represent the other as a non-knower, without allowing them to represent the other as a false-believer or a true-but-Gettiered-believer.
Oh, brilliant! Yes! I imagine half-remembered bits of that were swirling around in my head, but I hadn't consciously made the connection at all.