More on the well of knowledge norms

Dustin Locke has published a response to my Thought article, "Knowledge Norms and Acting Well". My paper (draft here) argued that lots of counterexample-based arguments against knowledge norms of practical reasoning take a problematic form: generating a case where it seems like S knows that p, but where it seems like S is not in a strong enough epistemic position to phi. These verdicts together tell us nothing interesting unless we assume some story about the relationship between p and phi; but defenders of knowledge norms needn't and shouldn't accept many such relationships.

For example, in Jessica Brown's widely-discussed surgeon case, it is thought to be intuitive that before double-checking the charts, (a) the surgeon knows that the left kidney is the one to remove; but (b) the surgeon ought not to operate before double-checking the charts. This is only a problem for the idea that one's reasons are all and only what one knows if the proposition the left kidney is the one to remove would, if held as a reason, be a sufficient reason to justify operating before double-checking the charts. But why should one think that?

Dustin resists my argument at several points. I'm not sure what to say in response to many of them; I think they're helpfully clarifying the sources of disagreement, but they don't make me feel any worse about my point of view. For example, Dustin seems to be happy to rest on certain kinds of very theoretical intuitions, like the intuition that the surgeon isn't justified in using the proposition that the left kidney is the bad one as a reason that counts in favour of removing the left kidney immediately. I don't have this intuition, and I wouldn't want to trust it if I did. I feel pretty good about intuitions about what actions are ok in what circumstances, but deeply theoretical claims like these don't seem to me to be acceptable dialectical starting places.

In what I found to be the most interesting part of his paper, Dustin also constructs a version of Brown's surgeon case where, if one assumes that (a) a Bayesian picture of practical rationality is correct and (b) practical reasons talk translates into the Bayesian talk by letting one conditionalize on one's reasons, we can derive the intuition mentioned above. I think that both of these assumptions are very debatable, but I also think that the case Dustin tries to stipulate is more problematic than he assumes. He offers the following stipulations:

1. The surgeon cares about, and only about, whether the patient lives.
2. The surgeon has credence 1 that exactly one of the patient's kidneys is diseased, and a .99 degree of credence that it is the left kidney.
3. If the surgeon performs the surgery without first checking the chart, she will begin it immediately; if she first checks the patient's chart, she will begin the surgery in one minute.
4. The surgeon has credence 1 that were she to check the chart, she would then remove the correct kidney.
5. If the patient has the correct kidney removed during the operation, then there are the following probabilities that he will live, depending on how soon the surgery begins: (5a) If the surgery begins immediately and the correct kidney is removed, there is a probability of 1 that the patient will live; (5b) If the surgery begins in one minute and the correct kidney is removed, there is a probability of .999 that the patient will live.
6. If the patient has the wrong kidney removed during the operation, then the probability that he will live is 0.

(This list is quoted directly.) I have two worries. First, Dustin also says of the case that "it's quite plausible that the surgeon knows that the left kidney is diseased", and assumes that she does. But this requires a very substantive epistemological and psychological assumption about the relationship between credence and knowledge. It is not at all innocent to assume that knowledge is consistent with non-maximal credence like this. For lottery-related reasons, Dustin is probably committing himself to the denial of multi-premise closure here. (Indeed, for reasons like the ones Maria Lasonen-Aarnio has emphasized, he may very well commit himself to denying single-premise closure.) That's not a completely crazy thing to end up being committed to, but I think it substantially mitigates the rhetorical force of an argument against me here. Similarly, there are probably good reasons to deny that the surgeon outright believes that the left kidney is diseased under these circumstances, either for conceptual/metaphysical reasons (see e.g. Brian Weatherson's "Can we do without pragmatic encroachment" or Roger Clarke's "Belief is credence one (in context)" or for psychological reasons (e.g. Jennifer Nagel's "Epistemic anxiety and adaptive invariantism"). If any of these views are right, the Dustin is committing to knowledge without outright belief.

My second worry concerns stipulation number 1: this is a surgeon who cares only about the life of the patient. From a realistic point of view, this is a very strange surgeon. According to Dustin's stipulations, the surgeon cares nothing at all about any of the following: whether she follows hospital procedure; whether she sets a good example for the students observing; whether she acts only on propositions that she knows; whether she is proceeding rationally. These strong assumptions are not idle; if we allow that she cares about any of these things, the utility calculus will not require her to go without checking, even when she conditionalizes on the content of her knowledge that the left kidney is diseased. (Suppose she cares about whether she acts only on that which she knows, and that she doesn't know whether she knows; then there is a substantial risk of the negative outcome of acting on something she doesn't know.) But these very strange assumptions will make our intuitions harder to trust. When we try to imagine ourselves in her position, we naturally assume she cares about the ordinary things people might care about. Stipulating that she only cares about one thing—not even mentioning the many other things we have to remember to disregard—makes it very hard to get into her mindset. So I'm inclined to mistrust intuitions about so heavily-stipulated a case.

Acting on knowledge under uncertainty

Susan, who is entertaining later this evening, is about to walk to the local market to buy some olives for cocktails. She now faces that famous perennial question whether to take her umbrella. We stipulate:

• Susan does not know whether it will rain during her walk.
• Susan's rational credence that it will rain during her walk is 0.4.
• The nuisance of carrying the umbrella on her walk will cost Susan 10 utils.
• The nuisance of being rained upon without an umbrella is -30 utils. (It is no nuisance at all to be rained on if she has her umbrella.)

It's pretty reasonable to suppose that in this case, Susan ought to take the umbrella; we calculate the expected value pretty straightforwardly. The umbrella costs 10, and gives her a 0.4 chance of saving 30. (30 * 0.4) - 10 = +2. If Susan takes the umbrella in a way sensitive to the positive expected value of doing so, there's a pretty strong intuition to the effect that she's done everything right.

As you know, cases like this one are sometimes thought to be problematic for the knowledge norm of practical reasoning. Alex Jackson's careful paper does a nice job separating different knowledge-norm-like commitments, but he identifies this kind of case as at least a prima facie challenge to the following determination thesis:

What one knows determines what it is rational for one to do (possibly in concert with one’s desires).

I like this determination thesis. (In fact, I like something stronger -- I think we should be after something more like metaphysical grounding, not just determination.) So I need a story about the case of Susan. The Hawthorne & Stanley story is that Susan, if she is acting appropriately, is acting on knowledge about epistemic probabilities. She says to herself: there is a 0.4 chance that it will rain; this thing that she says to herself is what she treats as a reason. And it is something she knows.

I'm pretty uneasy about this line, although I've never been able to put my finger on exactly what I don't like about it. There is something odd, it seems to me, about probabilistic contents playing these kinds of roles. I know that's not an objection; I'm just recording my uneasiness. Here, anyway, is an objection: suppose she doesn't know the relevant probabilistic claim. Suppose that, for all she knows, the chance that it will rain is 0.3.

Remember, this is evidential probability we're talking about; the difference between the chance's being 0.3 and its being 0.4 can't be made by meteorological facts wholly outside Susan's ken. Still, it's not at all implausible that Susan might not know, with that level of precision, whether her evidence probabilifies rain to degree 0.3 or 0.4. Indeed, my own evidence right now, it seems to me, puts the chances of its raining on me as I walk to work tomorrow right around that ballpark; but I have no idea whether it is closer to 0.3 or to 0.4. I hope you agree this is not a very implausible possible situation.

Notice also that if the probability really is only 0.3, then, given the stipulations above, Susan's expected value for taking the umbrella is negative. ((30 * 0.3) - 10 = -1) So under the current stipulations, for all Susan knows, taking the umbrella might have negative expected value. She doesn't know that she should take it. She does, we may allow, know that there is some chance of rain, but this doesn't look like a good enough reason to perform this action.

You might think about trying to gild the bitter pill at this stage, suggesting that if she doesn't know whether it's better to take it, then she really does violate the action norm in taking it, although she does so in an excusable way. This seems to be Hawthorne & Stanley's line. But I don't think we should take it. For it's consistent with our case here that Sarah is exceptionally well-attuned to the evidence. That is, if the chance really were only 0.3, then she wouldn't take the umbrella. This is of course totally consistent with her failure of introspective discrimination.

I think Hawthorne & Stanley were right to look to knowledge with contents other than that it will rain, but wrong to focus in on probabilistic ones, in part for the reason just offered: there's no particular reason to expect them to be known. (And also in part because of that feeling I haven't managed yet to articulate, that these things aren't the right kinds of things to be invoking in one's reasoning.) While there's no reason, it seems to me, to think that Sarah must know the probabilistic content, there is, it seems to me, good reason to think she must have some other relevant knowledge around. If, for example, you think that E=K, then the evidential probability must be probability conditional on some knowledge. Let that knowledge be the reason for action. What is it that is the relevant evidence? I don't know, it depends on how the case is filled out. Maybe something a forecaster said? Maybe the look of the clouds? Whatever it is, I say we understand Susan as acting on the basis of that evidence.

Can you get a case like this involving no such evidence? Alex Jackson tries to give us one. But this blog post is getting long and I'm getting hungry, so maybe I'll leave discussion of that for another day. 

Correction in "Knowledge Norms and Acting Well"

I was very pleased to have my short discussion on evaluating the knowledge norm of practical reasoning appear in the inaugural issue of Thought. Unfortunately, I've just noticed that there are two errors near the end of the published version of the paper. One, which is entirely my fault, is that I misspelled Mikkel Gerken's name. I'm very sorry, Mikkel!

The second error, which seems to have been introduced in copyediting, is more likely to interfere with comprehension. So I thought I should at least set the record straight here. The penultimate paragraph of my paper was meant to run thus:

The point cuts in both directions: pairs of intuitions like the ones featured above cannot be used to refute the knowledge norm of practical reasoning; neither can cases of that include both knowledge and appropriate action, or both ignorance and apt criticism of action, be used to speak directly in favor of the knowledge norm. The same point applies to attempts to evaluate knowledge norms from the other side: just as one can’t get very far from arguments of the form ‘S knows that p, but oughtn’t to Φ, neither can one get very far from arguments of the form ‘S doesn’t know that p, but it would be correct to Φ.’ Relatedly, pairs of cases that differ with respect to knowledge, but are alike with respect to appropriate action—as is plausible, for instance, with knowers and their Gettierized counterparts—do not bear at all directly on knowledge norms (contra  Gerkin (2011), pp. 535-36; Smithies (2011), p. 5). The knowledge norm identifies knowledge with reasons, but the facts about what reasons one has do not supervene on the facts about what actions are appropriate. (Perhaps there is supervenience in the other direction.)

The penultimate sentence of this paragraph unfortunately became rather mangled. (I regret that I whiffed my chance of catching it in proof corrections.) This is what was printed:

...Relatedly, pairs of cases that differ with respect to knowledge, but are alike with respect to appropriate action—as is plausible, for instance, with knowers and their Gettierized counterparts—do not bear at all directly on knowledge norms (contra Gerkin 2011, pp. 535–536; Smithies 2011, p. 5). The knowledge norm identifies knowledge with reasons, but the facts about what reasons one has to do does not supervene on the facts about what actions are appropriate. (Perhaps, there is supervenience in the other direction.)

The point I was trying to make was that everyone should agree that sometimes, pairs of subjects who have distinct reasons available to them ought nevertheless to perform the same actions—different reasons may point in the same direction. And given the knowledge norm, this is pretty plausible in the case of Gettier subjects and their knowledgable counterparts. Henry in fake barn country and twin-Henry in real barn country do not share all the same reasons: twin-Henry has the proposition that there is a barn in front of him, and Henry does not. Nevertheless, if they're both allergic to barns, they are each reasonable in stepping away from the structure before him. Twin-Henry's action is made reasonable by the reason that there is a barn before him (combined with his allergy and interests); Henry's action is made reasonable by the reason that there is a building that looks just like a barn before him (combined with his allergy and interests).

The Knowledge Norm of Practical Reasoning

For reasons exactly like the ones outlined in the previous post, these two claims are importantly distinct:

(1) If S knows p, then S can appropriately rely on p in practical reasoning.

(2) If S knows p, then p is warranted enough to justify S in phi-ing, for any phi.

I argued a couple of weeks ago that (1) is not strong enough to establish pragmatic encroachment. I suggested then that this was a problem for Fantl and McGrath; the discussion with Jeremy in the comments thread is part of what helped me to see the distinction between (1) and (2) more clearly. If their argument proceeded from (2), rather than (1), my argument doesn't apply. However, in light of the important distinction between these two claims linking knowledge and action -- I won't speak for anybody else, but this is a distinction I certainly hadn't been thinking clearly about before recently -- we should, if relying on claims like (2), proceed carefully, distinguishing arguments for (2) from arguments for (1).

I gave in my most recent post, linked at the top, an argument against a strictly weaker principle than (2). If that argument was right, then (2) is false. Let phi be an action that, for accidental reasons concerning the background environment, is only justified if S has some super-knowledge access to p. The example from that post was, let phi be the assertion that p, and let the background be such that S has promised, in a morally weighty way, not to assert p unless S knows that she knows that she is absolutely certain that p; let p be known, but the higher condition not be met. Then S is not justified in phi-ing, under the circumstances, even though she knows p, and even though, were p better warranted, she would be.

But maybe that argument went a little too quick. For maybe it's question-begging to assume, as I did, that one can know p without being in the super-epistemic position, under the circumstances described. Maybe the act of promising collapses that distinction. If so, then my argument against (2) can be resisted. Indeed, it's sort of the point of (2) that the 'standards' for knowledge raise to as high a level as one might need in any given circumstance.

But -- and here's the main point of this post -- one can retain (1) without collapsing that distinction. That's another respect in which (1) is interestingly different from (2).

Knowledge Norms and Pragmatic Encroachment

I'm thinking a bit more today about the point I made in a post yesterday about the use of intuitions about cases to evaluate knowledge norms. That point was basically that facts about whether S knows p and whether S is well-enough situated epistemically in order appropriately to X don't by themselves say anything about the knowledge norm of practical reasoning; S may know p without X's being appropriate just by virtue of p's not being a good enough reason to X. Yesterday I used this observation to rebut a certain kind of argument against knowledge norms.

I now think that, in addition to that use, this observation undercuts certain implications sometimes drawn from knowledge norms. In particular, I think it points to a lacuna in one of the central arguments of Fantl and McGrath's defense of pragmatic encroachment. I think this is a pretty fair reconstruction of their §4.3:

1. KJ: If you know that p, then p is warranted enough to justify you in phi-ing, for any phi.


2. Consider some low-stakes action, X, for which LOW has q as a sufficient reason; LOW can appropriately perform X. (Their example: Matt knows the train is a local; this justifies his boarding it.)


3. There is a possible counterpart of LOW in the same epistemic position but with higher stakes, HIGH, such that HIGH cannot appropriately perform X. (Their example: Jeremy really needs to get off at a local stop; the stakes are too high for him to risk boarding.)


4. So q is not well-enough warranted to justify X in phi-ing.


5. So HIGH does not know q.


6. So two subjects in the same epistemic position can differ with respect to knowledge of q.


7. So purism is false.

(For maximum faithfulness to F&M's broader intentions, we should understand this entire argument, including its conclusion, as being offered under the conditional assumption that fallibilism is true -- that it's possible to know that p, even though there is some epistemic possibility that not-p.)

I'm concerned with the move from (3) to (4). It has pretty much the same form as the arguments sketched above, except that it's holding different bits fixed. Jessica Brown argued from what she saw as intuitive verdicts about knowledge and appropriate action against the knowledge norm; Fantl and McGrath argue from intuitive verdicts about appropriate action and the principle of the knowledge norm against the knowledge verdict Brown found intuitive. My argument shows the flaw in both of these arguments. (In a sense, they are instances of the same argument, run in different directions.)

The move from (3) to (4) relies, like Brown's surgeon case argument, on the assumption that the knowledge of q, the non-actionability of X, and the knowledge norm are incompatible. But as I've shown, they're not. They're incompatible only on the assumption that q is, if possessed as a reason, a good enough reason to X. And there just aren't obvious intuitive verdicts about facts like these; neither are there clear theories that dictate which claims like these to accept. It's obvious enough whether S knows q, and whether S would be justified in Xing -- but whether q itself is well-enough justified to be among the reasons S has for Xing is an esoteric question on which naive intuitions are silent.

Everybody who accepts (2) and (3) has to think there's some important difference that derives from a change in the stakes that bears on actionability. But you can think this without giving up on KJ, fallibilism, or purism, if you want to. You can say that what propositions are good enough reasons to X depends in part on the stakes. That is: whether p is known is stakes-independent; so too is whether p is warranted enough to be a reason for phi-ing, for any phi. What varies by stakes is whether p, supposing that it is a reason, is by itself a good enough reason to phi. When the stakes are high, p, though still genuinely a reason, isn't a good enough reason. In lower stakes, p is a good enough reason. Insensitive knowledge; insensitive reasons for action; sensitive needs of actions for reasons.

This does look to me like a significant and substantive gap in F&M's argument. I'm generally pretty sympathetic to F&M-style views -- I'm a different sort of contextualist than those who are motivated by 'intellectualism' -- but this does look to me to be a potentially promising avenue for resisting pragmatic encroachment from F&M-style arguments.

It will not obviously help, however, with the knowledge intuitions about bank cases and the like. Lots of pragmatic encroachment people seem to be turning their emphasis away from these cases recently, in favor of broader theoretical arguments. Certainly, this seems to be one of F&M's aims. I rather suspect, though, that the pragmatic encroachment theorist may end up needing to rely on judgements about cases more than they think. (I don't know that that's a problem.)

Knowledge Norms and Intuitions about Cases

Here's a boring thought experiment that doesn't demonstrate anything.

Smith burgled the house last night; Detective Stanley is investigating the crime scene. He acquires evidence sufficient for knowledge that the burglar came in through the window, but finds very little evidence about whether it was Smith or someone else who committed the crime.

Here are two intuitive verdicts that aren't in any tension at all:

1. Stanley knows that the burglar came in through the window.


2. Stanley would need to have more evidence in order for it to be appropriate for him to arrest Smith.

Everybody can accept these obvious claims. In particular, these obvious claims are in no tension with the knowledge norm of practical reasoning, which claims that p can be an appropriate reason for action for S if and only if S knows that p. It would be an anemic objection to the knowledge norm to point out that Stanley knows that the burglar used the window, but needs more evidence in order for it to be appropriate to arrest Smith. That the burglar used the window just isn't a strong enough reason to arrest Smith, so this case doesn't tell us anything about what is and is not a reason. So it doesn't tell us anything about knowledge norms.

The moral of the story is that claims about who knows what, and about what actions are inappropriate, are in general insufficient to refute the knowledge norm of practical reasoning. (So, mutatis mutandis, for the knowledge norm of assertion.)

When you look at the case given above, this moral is really obvious. But sometimes, I think, it's neglected. Jessica Brown, for instance, argues against the knowledge norm of practical reasoning by citing this case:

A student is spending the day shadowing a surgeon. In the morning he observes her in clinic examining patient A who has a diseased left kidney. The decision is taken to remove it that afternoon. Later, the student observes the surgeon in theatre where patient A is lying anaesthetised on the operating table. The operation hasn’t started as the surgeon is consulting the patient’s notes. The student is puzzled and asks one of the nurses what’s going on:

Student: I don’t understand. Why is she looking at the patient’s records? She was in clinic with the patient this morning. Doesn’t she even know which kidney it is? Nurse: Of course, she knows which kidney it is. But, imagine what it would be like if she removed the wrong kidney. She shouldn't operate before checking the patient’s records.

We have, as before, a pair of intuitive verdicts: one attributing knowledge, and another denying appropriateness of action. Brown considers this to be a counterexample to the knowledge norm of practical reasoning, but the case of the burglar shows that this cannot not enough. Just as the burglar argument was transparently invalid, because the burglar's use of the window wouldn't be sufficient reason for arresting Smith, Brown's argument is valid only on the assumption that the disease in the left kidney would be a sufficient reason for operating without checking the charts. But Brown has given us no reason to think that is so. It's entirely open to the defender of the knowledge norm to argue that knowledge of p is sufficient for p to be a reason, but that in this case, p isn't a good enough reason for action.

This strategy is always available. I think this shows that trading in intuitions about who knows what, and who ought to do what, is not a helpful strategy for evaluating knowledge norms.

Knowledge norm of practical reasoning

In chapter 1 of Knowledge and Lotteries, John Hawthorne introduces the knowledge norm of practical reasoning: "At a rough first pass, one ought only to use that which one knows as a premise in one's deliberations." (p.30) He then immediately qualifies this principle in two ways with this footnote (fn.77):

Qualification 1: "In a situation where I have no clue what is going on, I may take certain things for granted in order to prevent paralysis, especially when I need to act quickly."

Qualification 2: "If I am in a situation where the difference between 'Probably p' and 'p' is irrelevant to the case at hand, I may use 'p' as a basis on which to act even though I only know that probably p."

I don't see why either of these are true in a sense that demands qualification of the rough first pass given above. With regard to qualification 1, let's suppose I'm in a situation where I have no clue what is going on. This isn't literally plausible, of course; in any situation in which my actions are rationally evaluable, I'll have some clue what's going on. So I suppose this must be understood as a kind of exaggeration. Perhaps, for instance, I suddenly find myself being charged at by a rhinoceros, and have no idea how I got there. It's clear, however, that if I just stand pat I will be very shortly gored to death. There's a button nearby with no label; I don't really have any clue what it is or what it does. But it may well be rational to press it, since that's the only real option I have and it's clear that if I do nothing, I will die. Maybe the button will open a trap door for rhino or for me—who knows? It's worth a try.

The thing is, all the premises that I'm using, if I so reason, are things I know. I know that there's a rhino; I know that I'll die if I do nothing; I know that there's a button; I know that maybe if I press the button I'll survive. So I don't see what pressure cases in which I have to act under extreme uncertainty put on the rough principle stated.

Similarly, if I'm in a situation where the difference between 'probably p' and 'p' is irrelevant to the case at hand, why think that I'm using 'p' as a basis to act? We've just stipulated that 'probably p' will do just as well—why not say I'm acting on that known proposition?

I don't really see what Hawthorne is up to in this footnote. (I suspect these issues may be developed in the later paper with Jason Stanley -- I read that a couple of years ago, but need to have another look.)