good news

Good news. The note on Dancy I posted a link to here, entitled ‘Dancy on Acting for the Right Reason,’  has been accepted by the Journal of Ethics and Social Philosophy. I will leave the version linked to in the previous post up until I make some revisions tomorrow.

knowledge, justification, and assertion

I’m sure you’re all dying to for more about ought, but for now I’m going to let you continue to die. This post is about knowledge and assertion (and I’m betting Clayton has already had some of these thoughts). I want to argue that being committed to the following three positions is a sorry state to be in:

knowledge-assert: One may assert if and only if one knows .

justified-not-knowledge: One can be justified in believing even though one fails to know .

belief-assert: If one is epistemically warranted in believing , then one may (=is epistemically warranted in asserting) .

I take it that most defenders of the Knowledge Account of Assertion (KAA) want to be able at least to accept Knowledge-Assertion and Justified-Not-Knowledge. Williamson makes it clear that Justified-Not-Knowledge is compatible with E=K+his account of justification. Moreover, he seems to think that it would be a bad thing if he was forced to give up Justified-Not-Knowledge. I also think that Hawthorne and Stanley (e.g.) want to hold both Knowledge-Assertion and Justified-Not-Knowledge (although, I should say that I’m not as familiar with those two books as I am with Williamson).

Belief-Assert is also very intuitively plausible. Imagine what would make it false: A case in which it is perfectly within your epistemic rights to believe but not within your epistemic rights to assert (of course, it’s easy to imagine a case where it’s not within your etiquette rights to assert ). Some also argue (e.g. Jon Kvanvig and Kent Bach) that the fight about the norm of assertion is really just a fight over the norm of belief. Obviously Belief-Assert falls out of that.

Here’s why it’s a sorry state to hold all three positions: If you hold Justified-Not-Knowledge, it’s possible for someone to be epistemically warranted in believing but not know . Thus, from Belief-Assert one is epistemically warranted in asserting . But since one doesn’t know , it follows from Knowledge-Assert that one may not assert . #

There are several fairly common moves one might employ to get rid of this. One might make some type of objective/subjective distinction about ‘warranted’ and say that one is objectively warranted iff one knows and one is subjectively warranted iff one is justified (or make the distinction between objective ‘may’ and subjective ‘may’). I’m skeptical that will work, but encourage people to try.

Barring arguing that there is an equivocation, the only other way to get rid of the problem is by giving up one of the three views. I take it that most who hold Knowledge-Assert will want to give up Justified-Not-Knowledge (’Sutton’ it, as it were). But that means that justification=knowledge. That means that that Gettier guy was all wrong! Surprising result.

I say let’s give up Knowledge-Assert.

who reads this blog?

I am often curious exactly who reads this blog. So, a la Leiter, consider this an opportunity introduce yourself. Even if you don’t comment often (or ever), I am interested who the readers are.

Update 8.8: Hmm, there have been ~50 views of this post. And yet, there have only been 6 responses. Why this is is puzzling.

the objective ‘ought’ and the three-envelope problem

We left off with the failure of two state of the art views about the subjective ought to explain why Margaret ought to choose envelope three. The problem with those views is fundamental. The problem is that they tie what one ought to do too closely to what one believes. It doesn’t seem like this is what matters. Instead, it seems that what matters is the information accessible to Margaret–i.e. what matters is what information state Margaret is in.

My first suggestion is that we ought to rethink the traditional distinction between the subjective/objective ought. Instead of thinking that A ought to iff one’s fully-informed self ought to , we should think that an ought is objective just in case it is analyzed in terms of objective reasons. This way of drawing the distinction allows for the possibility of views of the objective ought that can explain why Margaret ought to choose envelope three. I want to argue for such a view.

One thing that we learned from discussing the views about the subjective ought is that what seems important is not what one believes, but what information is in one’s information state. Other cases besides Margaret’s support this. Imagine that Nathan is assembling his son’s stroller. Although he has the instruction manual out, he doesn’t pay much attention to it. He thinks that his handy-man instincts are sufficient for successfully assembling the stroller. As it happens, he fails to to attach an essential screw. When he first takes his son out for a stroll, the stroller breaks and his son is injured. It’s natural to say that Nathan should have paid attention to the instruction manual. This is because, I suggest, Nathan was in a position to know that attaching the screw was part of the assembly process. Generalizing, I want to say that one’s information-state consists in the propositions one is in a position to know.

Another thing we learn from reflecting on the cases is that what is important isn’t any old proposition that one is in a position to know. It seems that what matters are the reasons one is in a position to know. The proposition that there is $1000 in envelope three is one of the propositions in Margaret’s information state that matters w.r.t. what Margaret ought to do. Similarly, the proposition that the screw is essential for the stroller’s safe assembly (or something along these lines) is the proposition that matters w.r.t. what Nathan ought to do.

These thoughts lead naturally to the following proposal, which we’ll call Objective O:

objective o: Necessarily, one objectively ought to iff the set S of objective reasons that A is in a position to know has the following property: S decisively supports -ing.

Objective O gives a nice explanation of why Margaret ought to choose envelope three. The relevant reasons are that there is $1500 in envelope one and that there is $1000 in envelope three. Margaret isn’t in a position to know the former reason and is in a position to know the latter. Thus, the latter is in S and the former isn’t. Thus, it’s obvious that S decisively supports choosing three. Objective O thus entails that Margaret ought to choose envelope three. Moreover, Objective O doesn’t have the problems of that the accounts of the subjective ought do.

Unfortunately, there is a problem. Imagine that instead of there being $1500 in envelope one, there is $15,000,000. As in the original case, imagine that Margaret is told that there $15,000,000 in either envelope one or envelope two and that there is $1000 in envelope three. Under normal circumstances, Margaret would be slightly crazy for choosing envelope three. It’s clear, that is, that Margaret is not obligated to choose envelope three in this case.

The problem for Objective O is that Margaret is in a position to know the same reasons in this case that she is in the original case. Specifically, she is not in a position to know the reason to choose envelope one (that there is $15,000,000 in envelope one). Thus, it seems as if Objective O entails that Margaret ought to choose envelope three in this case as well. This is intuitively false.

I don’t think that this is a problem for Objective O. Instead, I think it’s a problem for an assumption implicit in the dialectic so far–viz. that the following proposition isn’t an objective reason: that there is $15,000,000 in envelope one or envelope two. Call this proposition Disjunction. The thought is that since envelope two is empty, there cannot be a reason supporting the act of choosing envelope two.

Why should we think that Disjunction is a reason? I won’t fully defend the claim here, but gesture at my more complete argument. There are a few reasons to think that Disjunction is an objective reason. One, it’s plausible that Margaret ought to take Disjunction into account in deliberation. To see this, imagine that Susan is in a similar choice situation as Margaret. Unfortunately for Susan, she doesn’t know anything about what’s in envelope one or two. Suppose Margaret and Susan encounter each other, and Margaret tells her what she knows about the first two envelopes. Susan should take Disjunction into account in her deliberation about what she ought to do. Moreover, it’s not simply because Susan believes Disjunction that she ought to take it into account in deliberation. Like Nathan and Margaret’s reasons, Disjunction seems like it’s something Susan ought to take into account in deliberation regardless of whether she believes it.

A second reason to think Disjunction is a reason is that it seems to have weight. We seem to take this as obvious when we make a judgment about the revised case (the $15,000,000 case). Call that there is $1500 in either envelope one or two Disjunction-1. Disjunction seems like a weightier reason than Disjunction-1. A fortiori, Disjunction is a reason. Disjunction can also have less weight than other reasons. Imagine that there is a fourth envelope, and that Margaret is told that there is $15,000,000 in either envelope one or envelope two and that there is $16,000,000 in either envelope three or envelope four. Call that there is 16,000,00 in either envelope three or envelope four Disjunction-2. Disjunction-2 is weightier than Disjunction. But that can be true only if both are reasons.

I take it that these considerations are independent reasons to reject the assumption that there is no reason to choose envelope two. Disjunction is a reason to choose either one or two. Similarly, that there is $1500 in either envelope one or two (call it Disjunction-3) is a reason to choose envelope one or two.

In the revised case, Margaret is in a position to know Disjunction; thus, it’s in S. Moreover, it’s plausible that Disjunction decisively outweighs Margaret’s reason to choose envelope three. So, in the revised case Objective O entails that Margaret ought to choose envelope one or envelope two. But what of the original case? Margaret is in a position to know Disjunction-3 in the original case; thus, it’s in S. However, Disjunction-3 is not nearly as weighty as Disjunction. In fact, it is intuitively plausible that Margaret’s reason to choose envelope three decisively outweighs Disjunction-3. Thus, it’s intuitively plausible that Objective O still entails that Margaret ought to choose envelope three in the original case.

As it turns out, this cannot be the end of the story; that’s because Objective O gives rise to another puzzle. What that puzzle is and my proposed solution will have to wait for another day.

sympoze

Andy Cullison has set up a philosophy Digg, which he is calling Sympoze. You can find it here. Its success depends on philosophers signing up and using it. I think that it would be very cool if there were a lot of users. So, go sign up philosophers.

the subjective ‘ought’ and the three-envelope problem

Suppose Margaret is put in the following choice-situation. She is placed in front of three envelopes. She is told that she will be allowed to choose one envelope. She is also told that there is $1500 in either envelope one or envelope two, and that there is $1000 in envelope three. Let’s say that the $1500 is in envelope one. Following Jake Ross, call this the three-envelope problem. Intuitively, Margaret ought to choose envelope three. Unfortunately, this case causes all sorts of trouble for most views about ‘ought’ (Coincidentally, there has been some recent discussion about this type of case on the blogosphere. See here for my, Alex’s, and Richard’s discussion at Richard’s blog; see the comments here for some discussion at Pea Soup).

Many ethicists and epistemologists already think they know the answer. They want to say that there is a distinction to be made between the objective ‘ought’ and the subjective ‘ought’. One objectively ought to iff one’s fully informed self ought to . What one subjectively ought to do, on the other hand, is somehow determined by features of one’s subjective situation. Once we make this distinction, continues the canon, we will see that Margaret objectively ought to choose envelope one, but subjectively ought to choose envelope three.

I’m going to argue here that two popular ways of formulating the subjective ‘ought’ cannot give an adequate explanation of why Margaret ought to choose envelope three.

Let’s begin by discussing what (Jake) Ross calls objective reductionism about ought. On this view, subjective ‘ought’ is defined in terms of the objective ‘ought’. Subjective-O1 is a rough characterization of the view:

subjective-o1: Necessarily, A subjectively ought to iff A would be objectively obligated to if A’s beliefs were true.

Parfit holds a view similar to this (the differences don’t matter for my argument). Unfortunately, this view doesn’t conclude that Margaret subjectively ought to choose envelope three. That’s because if her beliefs were true, then she would be objectively obligated to choose either one or two. In other words, suppose her belief that the $1500 is either in envelope one or envelope two is true. Then the $1500 is either in envelope one, in which case she objectively ought to choose envelope one, or it is in envelope two, in which case she objectively ought to choose envelope two. This view, then, cannot explain why Margaret subjectively ought to choose envelope three.

Mark Schroeder defends a view that is in some ways quite similar to Parfit’s view, but is in many ways much better. He defends what we’ll call objective reductionism about reasons. This view builds subjective reasons out of objective reasons. It then defines subjective oughts in terms of subjective reasons. Subjective-R is a rough definition of subjective reason:

subjective-r: A has a subjective reason R to iff R is the content of one of A’s beliefs and R has the following property: If it were true, then it would be an objective reason to .

The subjective ‘ought’ is then defined in terms of subjective reasons. Subjective-O2 is the definition of the subjective ought on this view.

subjective-o2: Necessarily, A subjectively ought to iff there are decisive subjective reasons for A to .

This view can explain why Margaret subjectively ought to choose envelope three. That’s because on this view what one subjectively ought to do is determined by the weight of one’s subjective reasons, not by the weight of one’s objective reasons. The subjective reasons that are relevant are that there is a .5 chance that $1500 is in envelope one and that there is a .5 chance that there is $1500 in envelope two and that there is $1000 in envelope three. Plausibly, the weight of the third subjective reason is decisive. Thus, on this view, Margaret subjectively ought to choose envelope three (when she has the relevant beliefs).

But I don’t think this is captures why we think Margaret ought to choose envelope three. To see this, imagine that Margaret carelessly listens to what she is told. She thus believes that the $1500 is in either envelope one or envelope three, and that the $1000 is in envelope two. On objective reductionism about reasons, Margaret subjectively ought to choose envelope two. But, I still have the intuition that even if this is true, there is still some sense in which Margaret ought to choose envelope three. Carelessness is not a way to dodge that obligation.

Of course, there are countless other ways in which Margaret could turn out not to be subjectively obligated to choose envelope three on this proposal. She could simply fail to believe anything at all about the envelopes. In all of these cases I still think she ought to choose envelope three. If this is right, then there is something fundamentally misguided about trying to explain why Margaret ought to choose envelope three by appealing to this type of subjective ought. I want to suggest that what is misguided is forging such a tight link between what the agent ought to do and what the agent believes. What is really driving our intuitions is that the information is available. In other words, what determines what Margaret ought to do in the sense we are after is the information state she is in. Any view that attempts to explain why Margaret ought to choose envelope three that doesn’t take this into account is doomed to be unintuitive.

I have a positive account to offer, but that will have to wait for another day.

wordle

I’m late in the game noting how awesome wordle is. If you are reading this on the main page (as opposed to on some type of reader), you’ll notice that the header is a wordle image. This is the wordle image for this blog. I also created one for a few papers. Below is from my long (8,000 word+) ‘Correctly Responding to Reasons and Internalism about Rationality.’

Update: this is quite the distraction. I’ve created some images from some books I have in pdf. First, Parfit’s On What Matters:

Next, Mark Schroeder’s Slaves of the Passions:

Finally, Schroeder’s Being For:

ought, ambiguity, and reasons

It is often said that ‘ought’ is ambiguous in the following kind of way. When one utters ‘A ought to ,’ one might mean that A morally ought to , or that A prudentially ought to , or A ‘chessly’ ought to etc. Judy Thomson has a quick and persuasive argument that this cannot be right (she gives it in ‘Normativity’ in this book; see also this taped talk of that paper). Imagine that Philip is playing chess with Michel. Michel utters ‘You ought to move your rook.’ Their sociopathic friend Pascal is watching their game, and he tells Philip that if he moves his rook he (Pascal) will torture and kill a small child. Michel, upon hearing this, utters ‘You ought not to move your rook.’ Philip is a little slow, and so he asks Michel what he really ought to do. If ‘ought’ was ambiguous in the way that it has often been claimed, Philip’s question wouldn’t make any sense. The most Michel could say is ‘Well, you chessly ought to move your rook, but morally you ought to not move your rook.’

The intuition that Thomson has is that this cannot be right. I share this intuition, and nearly everyone feels the need to try to account for the case. One popular way of trying to account for the case is by arguing for an all-things-considered ‘ought’. This ‘ought’, as the name suggests, takes into account all of the considerations bearing on whether someone should do some thing. What’s not clear is whether those who propose the all-things-considered ‘ought’ think that this is just another type of ‘ought,’ and thus when some says ‘A ought to ‘ one disambiguation is ‘A ought to all-things-considered ‘.

For reasons that won’t be gone into here, Thomson doesn’t buy this, and thus argues that the moral ‘ought’ is the only ‘ought’. Thus, in normal cases one ought to move his rook iff one is morally required to move her rook. I don’t think that this can be right either.

In fact, I think that there is only one kind of deontic ‘ought’. Moreover, it is analyzed in terms of reasons. Thus, A ought to iff there are decisive reasons for A to . Contextual factors can limit what reasons are relevant. So, when we are in a normal context of playing chess, the reasons that matter are the reasons that one has when one plays chess. When moral factors are contextually important, moral reasons are the reasons that matter. But, in many contexts (if not most), what we want to know is what we ought to do considering all of the reasons. And in these contexts, moral reasons are generally much weightier than chess reasons. But this is not to say that there is an ought and an ought and an ought and an ought. It’s the same concept across all the contexts, it’s just that context determines which reasons are important.

logical form and non-indexical relativism

Although at points the terminology used in the contextualism/relativism literature has been very confusing, recent work by MacFarlane and Weatherson seems to be focusing the terminology a bit. MacFarlane’s paper ‘Non-Indexical Contextualism’ does a really nice job mapping logical space (here, pdf). I’m going to follow Weatherson because it’s going to be easier for my purposes (the details of this are here (pdf), but I’m quoting from here). Weatherson distinguishes between two different ways a view can neglect assessor-sensitivity. That is, they can either accept (T) or (U):

(U) The content of what is said by an utterance is the same relative to any assessor

(T) The truth value of something that is the content of an utterance is the same relative to any assessors

Obviously, there is a view that holds (U) but rejects (T) about some domain . On this view, the content of -propositions is the same relative to any assessor, but the extension of those propositions is relative to assessors. Weatherson labels this view non-indexical relativism.

MacFarlane and Mark Richard have defended non-indexical relativism (see here and here (both pdfs) for MacFarlane and here for Richard). MacFarlane has defended non-indexical relativism about epistemic modals (see the first paper linked in the first sentence above). Non-indexical relativism, he thinks, allows him to solve the problem of disagreement that plagues the indexical contextualist (which is a popular view on the matter). The indexical contextualist accepts both (T) and (U). The problem of disagreement can be seen by considering Ned and Fred (an example similar to this is given by Kratzer somewhere; I cannot remember where). Ned, let’s say, is 15 feet away from a street corner. Fred is 50 feet from the same street corner. A person is approaching the corner. Fred has poor eyesight, and thus cannot see the corner all that well. He utters (1)

(1) That might be Zed.

Since Ned is closer and has better eyesight, he knows that it isn’t Zed. So, he utters (2),

(2) That cannot be Zed.

The indexicalist holds that the content of (1) and (2) is a function of what Ned knows and what Fred knows (obviously I’m assuming some stuff about epistemic modals; hopefully mostly everyone follows what I’m skipping in the interests of time). But since Ned and Fred know differing amounts of information, the indexicalist has it that Ned and Fred aren’t disagreeing. That is, Ned isn’t denying a proposition that Fred is affirming. But, intuitively, he is, and thus they are disagreeing. Nevertheless, it seems as if both Ned and Fred are saying something true. The indexical contextualist gets this result only if she saddles herself with the problem with disagreement.

The non-indexical relativist, however, can solve this problem. Since she’s a non-indexicalist, she can say that Ned and Fred are disagreeing about the same proposition. However, since she is a relativist (i.e. since she denies (T)), she can hold that (1) is true from Fred’s point of assessment and (2) is true from Ned’s point of assessment.

So far, I’ve just sketched some background for what I’m worried about. Kai von Fintel and Thony Gillies have attacked relativism (see here, pdf). They focus on epistemic modals, and they explicitly say that MacFarlane’s view is one they are attacking. They offer up a proposal about the logical form of sentences like (1). So, they hold that the logical form of (1) is (3),

(3) might (B) ()

where is the prejacent (or the proposition being quantified over) and B restricts the quantification. They then give an interpretation of B for the contextualist (i.e. someone who accepts (U)). That interpretation is (4):

(4) [[B]]c ,i = {v : v is compatible with the c-relevant information at i}

where c is the context and i is an world-time index. From (4), we can formalize what it takes for a proposition like (1) to be true,

(5) [[might(B)()]]c ,i =1 if and only if ∃w ∈ [[B]]c ,i: [[]]c ,w,ti =1

So far, so good. They then try to give similar formalizations for the relativist. (6) is how they assign the denotation for B,

(6) [[B]]c ,i,a ={v : v is compatible with what ja knows at ta in wa}

where ja is some judge (or assessor) and ta is the time of assessment and wa is the world where the assessing takes place. They they give (7) as an analogue of (5)

(7) [[might(B)()]]c ,i,a =1 if and only if ∃w ∈ [[B]]c ,i,a : [[]]c ,w,ti ,a = 1

However, thisdoesn’t seem like the correct formalization for non-indexical relativism. Instead, this seems like a formalization of indexical relativism. That’s because (3) and (6) result in points of assessment getting into how content is fixed. Thus, it’s not at all clear that indexical relativism can solve the problem concerning disagreement. Thus, when von Fintel and Gillies attack views that accept (6), they are not attacking MacFarlane. This is a problem considering they think they are attacking MacFarlane.

An interesting question arises, however. How should we formalize non-indexical relativism? As far as I know, Andy Egan’s centered world approach is the only explicit attempt at fixing what the content is supposed to be (I think Tamina Stephenson’s view is like Egan’s, but I’m not entirely sure). Any help with how to do this is greatly appreciated.

By the way, aside from being interested in epistemic modals, I’m also interested in how this all relates to the semantics of ‘ought’. I want to defend non-indexical relativism about ‘ought.’ I’m writing a paper on this right now, and I initially used the von Fintel and Gillies formalizations. But after trying to convert their formalizations about ‘might’ and ‘must’ to ‘ought,’ I realized that they didn’t give correct formalizations for non-indexical relativism.

Update July 7th: Tamina’s view is equivalent to Egan’s. Also, as Nate has pointed out in the comments, Kai and Thony’s relativist is equivalent to Tamina and Egan’s. Finally, and also because of Nate, I should point out that Kai and Thony’s relativist proposal does preserve content across points of assessment. Unfortunately, I’m still puzzled about the dialectic. See my comment below.

the reasons program, the normativity of rationality, and rationality as coherence

According to a popular position we will call The Reasons Program, what it is to be normative is to be analyzable in terms of reasons. This is a popular position and ‘trendy’ position in ethics. Two theorists who are often cited as card-carrying members of The Reasons Program are Tim Scanlon and Derek Parfit. And although they do explicitly say that they believe The Reasons Program’s main tenet, I don’t see how they can do this without contradicting themselves.

The problem has to do with (1) what rationality consists in and (2) the normative status of rationality. Both Parfit and Scanlon think that rationality just has to do with coherence in a broad sense of coherence. In chapter four of Climbing the Mountain, Parfit holds that one is rational (in most cases–the exceptions won’t help him with my problem) just in case one correctly responds to one’s beliefs. More specifically, one is rational just in case she does things that she would have decisive reasons to do if her beliefs were true. The contents of one’s beliefs are what Parfit calls apparent reasons. Under the right conditions, one’s apparent reasons are real normative reasons. However, Parfit contends, in some cases one’s apparent reasons aren’t real reasons at all. Thus, he writes,

(1) Our desires and acts are rational when they causally depend
on beliefs whose truth would give us sufficient reasons to have
these desires, and to act in these ways.
(2) With a few exceptions, it is irrelevant whether these beliefs are
true, or rational.

Thus, rationality doesn’t have any tight connection with reasons. However, Parfit maintains that rationality is normative (I just did a quick scan of ch. 4 to find him explicitly mention this. He writes on p. 88 that ‘the rationality of our desires and acts normatively depends…’ (my emphasis)).

Scanlon holds a similar position (to simplify, I’m only going to talk about his view in What We Owe to Each Other; note, though, that his view is more sophisticated than this now. See ‘Structural Irrationality’ in this book). In What We Owe to Each Other, Scanlon defines irrationality in the ‘narrow sense.’ Thus, according to the view there, one is irrational just in case one believes that she has decisive reason to but doesn’t . Obviously, though, one can believe that one has decisive reason to even if one doesn’t have any reason to . Scanlon explicitly acknowledges this. He also explicitly says that the requirements of structural rationality are normative.

I think that there is an obvious contradiction between the following three claims, all of which Scanlon and Parfit seem committed to:

(1) The requirements of rationality are normative.

(2) The requirements of rationality are not analyzed in terms of actual reasons.

(3) In order for some concept C to be normative, C must be analyzed in terms of reasons.

So, what’s the deal? Parfit, I think, is in a better position to amend his view so he can deny (2). He can say that apparent reasons aren’t always objective normative reasons, but that they are always subjective normative reasons. This is, in effect, Mark Schroeder’s view (see here, pdf). However, I highly doubt he wants to say that what he calls apparent reasons are always normative. I am partial to this view; so, I think he should. But I doubt that he is sympathetic to it.

Scanlon, on the other hand, seems to flat out reject (3) in this paper. He there distinguishes between normative claims that are structural–i.e. about structural rationality– and substantive–i.e. claims made about reasons. I think that people should stop citing him as an adherent to The Reasons Program.

Of course, one could deny (1). Niko Kolodny, at least the Niko Kolodny of ‘Why Be Rational?,’ seems to want to deny (1). Although in more recent papers (see here) he has cited Schroeder’s view as a view compatible with the type of criticisms he is making against coherence views (I will be posting about that issue eventually).

I would like to vindicate (1) and (3). And so I deny (2). In fact, I deny (2) in two interesting senses. First, I deny that all requirements of rationality are coherence requirements. Thus, some requirements are requirements to respond to objective normative reasons. Second, I deny that coherence requirements aren’t analyzed in terms of reasons. They are analyzed in terms of subjective reasons.