September 20, 2004

Knowledge, Knives, and Gettier

Let's try and get a bit of philosophical substance up, shall we?

Brian, who believes that knowledge is merely justified true belief, gave the following explanation for why people have Gettier intuitions:

The people who have Gettier intuitions are (on the whole) basically sceptics who have (perhaps) talked themselves out of their most sceptical intuitions but not this one. More carefully, people who have Gettier intuitions are disposed to intuitively apply KNOW in very few cases where possibilities of error are salient. Some of them may have convinced themselves that we KNOW we are not brains in vats, or that a mule is not a very cleverly disguised football official, but the underlying sceptical intuitions are still doing too much work.

(KNOW refers to heavy stress on 'know'--discussed here.)

(Also, to brush aside worries about whose intuitions these are--in this entry when I talk about intuitions I mean the intuitions of the people who found Gettier's examples convincing. Since I'm also talking about intuitions about jusitification, and justification used in this way is something of a philosopher's term, I think it's only philosophers' intuitions that are really relevant here. But anyway, the question pertains just to the community of people who have these intuitions. Although, as we'll see, it may just pertain to me.)

I want to argue that Brian's account doesn't really help explain why people should have Gettier intuitions if the JTB account is true. If the JTB account is true, why should the salience of possibilities of error make a difference? Salient possibilities aren't explicitly mentioned anywhere in the JTB theory.

One answer is: Making a possibility of error salient puts pressure on the justification component. If I fall victim to your skeptical hypotheses, I begin to think along the lines of "I shouldn't be so sure that I'm sitting in my office..."; skepticism here makes the beliefs seem not only not known but not justified.

But that can't be what's going on in the Gettier case, because part of the Gettier intution is to conclude that the subject lacks knowledge even though her belief is justified. We don't think, "Alice shouldn't believe that one of her students owns a Chevy Caprice, since she hasn't excluded the possibility that Mr. Nogot has been deceiving her." The point of the Gettier cases is to conclude that the belief is justified and true, and nevertheless not knowledge.

Here's an analogy--suppose we define a table setting as a knife and fork. We can generate skepticism about table settings by the Crocodile Dundee move: "That's not a knife." If we become convinced that KNIVES have to be very large and sharp, we will decide that many things we had thought of as table settings really aren't. But why would we think that we didn't have a table setting while being convinced that we really do have a knife and fork?

Just so you can keep track of the analogy: "table setting" = knowledge; "knife" = justification; "fork" = truth; "Crocodile Dundee" = skeptic. It's a shame that this was supposed to be a serious post.

Now here I think I have something of an answer. When we talk about a table setting we're creating a thing where previously there was just a separate knife and fork. So you might say--in order for this to be a real TABLE SETTING, the knife and fork have to match.

--That is to say, we think knowledge should be an organic unity, so we think the justification and belief components have to match up in a way that they don't in Gettier examples. This is not unconnected with Edward Craig's views on 'objectivisation' in Knowledge and the State of Nature. (I just noticed that the heavy emphasis on "knife" a couple of paragraphs ago doesn't work with the analogy, because we don't tend to stress 'justified'. Oh well.)

Another point to make is that sometimes it matters whether a belief is Gettierized; Timothy Williamson gives the example of a burglar who keeps ransacking a house because he knows a jewel is hidden inside; if he had a Gettierized belief based on false information that the jewel was under the bed, he'd give up when he looked under the bed. My favored account of knowledge can account for this while preserving the table-setting explanation of Gettier examples.

That account is this: There are many things we care about when evaluating a belief. We always care about truth, usually about justification, sometimes about other things as in the Williamson example. Attributing knowledge is a quick handy way of making several of these attributions at once, allowing us to retrieve the relevant one.

To switch the analogy: Knowledge is like a Swiss Army knife. It's got the knife blade and screwdriver, which you use all the time, and a few other things, like the corkscrew, tweezers, and fingernail board, which won't come up that often. It would be very irritating to carry those five things around with you, just as it would be cumbersome to enunciate separately that a belief was true, justified, etc., or to inquire into that question. But a Swiss Army knife gives you a handy way of carrying them all around, so that you'll have whichever one you need.

Now consider this--if you have a knife and a screwdriver, separately, you may have everything you'll need the Swiss Army knife for, but you don't have a Swiss Army knife. To have a Swiss Army knife, the blades must be connected. The Swiss Army knife consists of all the different blades in an organic whole.

Hence, when we use knowledge attributions as an economic way of saying five other things, we are doing more than just conjoining the other things--we're making them into an organic whole. And, even if we'll only be ultimately concerned with truth and justification of beliefs (most of the time), when we ask "Does she know?" we're asking about that organic whole too. If the truth and justification aren't connected we don't have knowledge--just as, if the knife and screwdriver aren't connected you don't have a Swiss Army knife, even if you have everything you need for your construction project.

Posted by Matt Weiner at September 20, 2004 04:38 PM

I don't mean to say that philosophers get suckered by Gettier cases because they are basically JTB theorists who get this kind of case wrong. I meant to say they have basically sceptical intuitions (i.e. no (salient) possibilities of error) but they have trained themselves not to trust them in certain cases. Good thing too, because scepticism is a crazy doctrine. But they haven't completely talked themselves out of scepticism, and it comes out in certain anti-JTB cases. That's the picture at least. It relies on there being a large gap between what knowledge really is and what proto-theory of knowledge most philosophers carry around in their heads, but I take it that's quite plausible. The stuff about stress is because somehow (I have _no_ idea how) more stress makes more possibilities of error salient, which leads the implicit proto theory to go haywire.

The 'organic unity' view has several attractions, but it seems to make it too hard to get knowledge in many cases. I'm sure I've used this case before somewhere, but I might as well try again. Jack decides he'll have a nap before going out to dinner. He's due at the restaurant at 9.30. He notices that the clock on his phone has stopped, so he sets it by the clock on his cable box. Sadly that's also stopped, but happily it's stopped at the right time. So Jack's phone is now a reliable time-keeping device, though not because it's connected to the time in anything like the right way. Anyway, Jack falls asleep and wakes up at 10. He looks at his phone, sees that it is 10, and immediately forms the belief (knowledge) that he is late for dinner. In that case he's got the knife and he's got the screwdriver and though he got them from very different sources, he's got enough for knowledge. Note that in the case the fact that it's dark outside is not an independent source of evidence for Jack. For one thing it would be dark even if it were 9, and Jack were _not_ late for dinner. The only relevant evidence is the phone, and it's good evidence, even though it's not organic evidence.

Or to put the point another way, I can see why we want true beliefs, and I can see the value in justified beliefs, but what's the added value in having the two connected by some bit of red plastic with a white cross on it?

Posted by: Brian Weatherson at September 20, 2004 05:29 PM

One way to read your account is that philosophers start with skepticism, they gradually abandon it because it's a crazy view, and once you start whittling back the only logical stopping point is JTB. I guess that's the view that I'd derive from your paper. In which case it seems to me that my whole post is a red herring--there's no need to connect Gettier intuitions with the JTB account per se.

As for the last question--no value at all in this case. But I don't think knowledge always has value, either. That's part of the reason that we usually do say in that case that Jack knows it's 10 o'clock. If we're using "knowledge" to mean "a true belief that's good enough in the ways we care about," Jack does know. And I think that is the main idea behind the concept of knowledge.

So I was probably overcommitting when I said "[i]f the truth and justification aren't connected we don't have knowledge"; certainly I think that it can be perfectly appropriate to attribute knowledge to Jack. [Though perhaps this is akin to the way that it can be appropriate to say "Yes" when someone says "Did you bring the Swiss Army knife?" if you know that all they care about is that you have a knife and screwdriver; "No, but I brought a separate knife and screwdriver" would be pedantic. Wouldn't want to be pedantic.]

What the Swiss Army knife bit is supposed to do is to answer this question: If knowledge attributions are just meant to encapsulate whatever it is that we care about, then why are we so eager to deny knowledge in Gettier cases, given that the missing element in Gettier cases is so rarely important? (I think that Williamson's cases show it's sometimes important--but not often.) The organic unity account is meant to explain that.

Part of the underlying thought here is that knowledge is a somewhat confused concept, though good enough for the work we put it to. For instance, our knowledge attributions are governed by the idea that knowledge is an organic unity--but it isn't really an organic unity, it's a bunch of other things stuck together (and not always the same things--perhaps I could stretch the analogy to cover the many different models of Swiss Army knives). This thought is something I'm trying to work out.

Posted by: Matt Weiner at September 21, 2004 07:59 AM