A Demonstrative Model of First-Person Thought

Presenter: Daniel Morgan, UCL

Commentator 1: François Recanati, Institute Jean Nicod/Center for the Study of Mind in Nature

Commentator 2: Edison Barrios, Independent Researcher

Advertisements

13 Comments

  1. Hi Daniel

    Thanks for the great paper. I’m afraid that I haven’t had enough time as I would have wished to devote to your paper, but I have a few quick questions.

    1. This is side-issue, really, but I would want to know how you respond to a variant on the causation account of reference determination in the perceptual that appeals to the causal relation and triangulates with this dispositions to act towards a particular object (I’m thinking of Burge’s response to Davidson’s triangulation argument). If asked who he is talking about, Emmanuel will point to Celia, not to Elsie or to any other causal intermediary.

    2. An version of the the Simple Rule Model that is properly explanatory (that has, one might say, psychological reality), would be a descriptive one, according to which the I-thought “I am F” is a thought to the effect that the subject of this thought is F (or, alternatively, the producer of this token). I realise that you have some qualms about the notion of a producer, but I’m not yet too worried. Do you think that thought-insertion cases show this to be problematic?

    3. I’m not sure why we should accept that K-max entialls that the referent of I in the proprioception based “I have crossed legs” is me, rather than my body. You allow that reflection on the content of bodily awareness alone does not show that I am the object of bodily-awareness (I agree), so it may be that I am a distinct (perhaps constituted) object. In which case, don’t we need some further reason to suppose that the referent is me?

    4. A worry about the inferential glue account might arise if people (falsely) believe that different uses of “I” refer to different things. So, for example, a Cartesian might claim that uses of “I” in psychological self-ascriptions refer to a soul, whilst uses in bodily self-ascriptions refer to a body (I think that Galen Strawson holds a view not entirely unlike this – although he’s not a Cartesian). If they are wrong, then inferential glue will give the wrong result won’t it?

  2. Francois:

    Thanks so much for your challenging comments. Here are a couple of initial replies.

    I think that my position and your “mental file” picture may not be quite as similar as you suggest. Take a token thought of mine that I would express with the sentence “I was born in Cork” and that I judge, and know, on the basis of testimony. I think that the right explanation of why this token thought is about me is that (i) it is based on a way of gaining knowledge – testimony – that on the occasion on which I’m having the thought provides knowledge about me and (ii) assignments of reference should maximize knowledge. My view isn’t that that the right explanation of why that token thought is about me has to mention other token thoughts, ones that are based on reflexive ways of gaining knowledge. On your view as well as mine, there is a split between two kind of first-person thought tokens. But, on your view, the split is between first-person thought tokens that are based on reflexive ways of gaining knowledge and those that there are not. Whereas on my view the split is between first-person thought tokens that are knowledge and those that are not. Admittedly, it isn’t obvious how to turn this difference into an argument for one of the two positions and against the other. It’s not as though we have strong intuitions about whether the best theory of first-person thought will draw the line in the way that I do or in the way that you do.

    Another, and closely related, difference between our two positions is in how they can be motivated. I motivate my position by reference to knowledge-maximization, which I take to be a defensible though contentious principle about how reference should be assigned to thoughts. I take that it that your position will need to rely on some principle in the same ball-park, but I don’t think it can be the very same principle. You claim that “an indexical file refers to the object to which a thinker stands in the ER relation which it is the function of the file to exploit”. Suppose someone asks to be given a reason to believe this. You might in replying to them try appealing to the knowledge-maximization principle. But that doesn’t seem to quite fit. On your picture, it is only the ways of gaining knowledge that are associated with the identity relationship – the reflexive ways of gaining knowledge – that get to do any reference-fixing work (in relation to SELF, a particular kind of indexical file). But, if knowledge-maximization is what underlies the reference-fixing work done by ways of gaining knowledge, non reflexive ways of gaining knowledge ought to be just as capable of doing reference-fixing work. So it looks as though you would have to appeal to some principle that is a bit like knowledge-maximization, but also a bit different. I don’t have any very definite ideas about how to formulate that principle. Once it is on the table though, it might be easier to adjudicate between your position and mine.

    You raise the excellent question of why the position I defend does not just count as a version of The Simple Rule Model Plus. What I say in my paper is that the position I defend, unlike The Simple Rule Model Plus, does not require us to accept as an unexplained primitive that all the tokens of the first-person type are tokens of the same semantic type. You ask “But what is the first-person type if it is not a semantic type?” Actually though, I don’t want to deny that all the tokens of the first-person type are tokens of the same type (or that that type is a semantic type, which it surely is). I just want to avoid accepting it as an unexplained primitive. I want this to be something that falls out of a story about reference-determination for first-person thoughts, rather than something that is built into the foundations of that story.

    The defender of The Simple Rule Model Plus assumes that all the different tokens of the first-person type are of a single type and that what needs to be explained is just why that type is governed by one particular rule – the simple rule – rather than by any of the other rules it might have been governed by but isn’t. In explaining why the simple rule is the right rule, The Simple Rule Model Plus appeals to knowledge-maximization. But it has nothing illuminating to say about why the hunt is on for a single rule in the first place, why just these tokens – all and only first-person tokens – are to be given a unified treatment.

    By contrast, my account approaches individual token first-person thoughts without assuming anything about there being a single (semantic) type they all fall under. For some of the tokens, the explanation of why they have the reference they have is in accordance with the knowledge maximization principle (thought t is about object o because it is based on an exercise of way of gaining knowledge that provides knowledge of o and assignments of reference maximize knowledge). For the rest of the tokens, the explanation of why they have the reference they have is in accordance with the inferential glue explanation (thought t is about object o because the subject treats it as being about the same object as t* and t* is about object o). It this all works out, the model will end up having explained why any token first-person thought is about the person whose thought it is. That in turn – that this highly interesting generalization holds of them – in my view explains why first-person thought tokens constitute a single semantic type.

  3. Thanks, Daniel. But you do not answer the question I asked in my comment: How do you account for the important distinction between ‘genuine’ first person thoughts and first person thoughts thoughts that concern the subject only accidentally? If I am the winner and through knowledge-transmitting testimony I gain the information that the winner will go to Tahiti, my thought that the winner will go to Tahiti counts as a first-person thought by the characterization you give. Or so it seems.

  4. Edison:

    Thanks so much for such excellent comments. Here are a few initial replies (the numbers for which correspond to the sections of your reply).

    (1)
    You write:
    “I think we should avoid the strategy of supporting views that are already plausible (i.e. that proprioception enables self-reference and other related phenomena in humans) by means of theses whose status is less obvious or more controversial, and whose defense requires elaborate argumentation or the adoption of contentious theoretical commitments (such as K-max)”
    This seems like a good general principle. It didn’t occur to me that it was applicable here, because it didn’t occur to me that the claim the Demonstrative Model makes might be obvious. Claims about obviousness are a bit hard to prove or disprove but let me mention one aspect of the literature on the first person that made me think the Demonstrative Model is non obvious.

    Anscombe’s sensory deprivation chamber case has been massively influential in the literature on the first-person. But the most flat footed way of saying what interesting thing the case shows is that first-person thinking is not dependent on the operation of ways of gaining knowledge, such as proprioception, and can proceed perfectly well in their absence. Treating that as the take home message of Anscombe’s case seems to be in tension with accepting the Demonstrative Model’s claim that the ways of gaining knowledge of which first-person thoughts are based are also what fix their reference. Of course, I don’t think the sensory deprivation chamber case in fact establishes anything that disproves the Demonstrative Model. But, on my reading of the literature, it has often been taken to.

    (2)

    You distinguish two kinds of question, questions of enablement and questions of selection. I don’t feel I have entirely got to grips with the distinction. An example you give is the distinction between the question “What enables the cheetah to run at N mph?” and the question “What determines whether a cheetah, in a given chase, will actually reach 60mph?”. I agree these two are different questions but I think the difference mainly has to do with the difference in the material towards the end of each sentence, rather than the difference between “enables” and “determines”. (For what its worth, the question “What determines that the cheetah is able to run at N mph?” does seem equivalent to the question “What enables the cheetah to run at N mph?” despite the fact that one of these involves “determines” and the other “enables”).

    One difference between the question “What determines the reference of my first person thoughts?” and the question “What enables me to latch onto myself in first person thought” is this. The first question is naturally read as presupposing that my first person thoughts do refer, and as just asking why my first-person thoughts have the reference they do and not some different reference. By contrast, in answering the second question it would be natural to mention material relevant to the question of why my first-person thoughts refer at all, as well as material relevant to the question of why they have the reference they do. I hadn’t spotted this difference when writing the paper, and so I was happy to slip between the two questions. As a matter of fact, I intended the Demonstrative Model, in conjunction with knowledge-maximization, to explain both why first-person thoughts have the reference they do (as opposed to some different reference) and why they have reference at all (as opposed to not having any reference).

    (3)

    You write:
    “Now, remember that DeM was able to mimic first-person rules in their reference-assignments, and that this depended on the fact that one can only receive proprioceptive information about oneself, thus narrowing down the set of possible referents down to one: the thinker. But this is a contingent fact, due to the actual wiring of our nervous system, among other things. The implications of this contingency and what it means for the comparison between semantic accounts and the DeM can be highlighted by the following scenario: we could at least conceive of intelligent beings whose constitution prevented their ways of gaining self-knowledge from guaranteeing a unique referent, but who nonetheless were able to enjoy first person thought, in such a way that (4) and (5), for instance, would also be self-defeating for them. If so, then they would be adequately characterized (at least in part) by SR or some homologous rule, but not necessarily by DeM.”
    We may disagree here about what is contingent. It is certainly possible that one be wired up to someone else in such a way that there is a causal chain that begins with their body and ends with one’s having experiences that one cannot distinguish from experiences of genuine proprioceptive awareness of one’s own body (quasi-proprioceptive experiences). But my claim is that ways of gaining knowledge like proprioception fix the reference of our first-person thoughts. The contingency question I think relevant to that claim is whether it is contingent that proprioception only puts one in a position to have knowledge of oneself. I say that that isn’t contingent.

    One way of defending this claim would be to say that in cases of quasi-proprioception, there is a genuine way of gaining knowledge one is employing, and it can provide you with knowledge of someone else, but this way of gaining knowledge is not proprioception. Proprioception, by definition, only provides knowledge about yourself.

    This strikes me as a bit lame. It leans very heavily on a certain, contentious way of individuating ways of gaining knowledge. It also raises the question of why quasi-proprioception couldn’t play a role in fixing the reference of first-person thoughts, even granted that quasi-proprioception is a distinct way of gaining knowledge from proprioception.

    A better way of defending the claim that proprioception only puts one in a position to gain knowledge of oneself is to say that that, in the case of quasi-proprioception, one is not in a position to gain knowledge on the basis of one’s experience at all, one is simply suffering from an illusion with an aetiology that involves someone else’s body. The “on the basis of one’s experience” clause here is important. If one knows about, or suspects, the unusual causal set up of the case, one can use the premise about one is having a quasi-proprioceptive experience as a basis for inferring something about someone else’s body. But this, I think, isn’t relevant. Given appropriate collateral premises, almost any proposition can be used as a basis for inferring almost any other proposition. For example, given knowledge of the collateral premise that whenever I seem to see a dagger before me, that is because there’s a bread-knife in my vicinity, I can use my knowledge that I am having an experience as of a dagger to infer that there is a break-knife in my vicinity. We do not need to understand knowledge-maximization in such a way that it implies that in this kind of case the experience as of the dagger will enable reference to the break-knife. Equally, we do not need to understand knowledge maximization in such a way that it implies that in the quasi-proprioception case the quasi-proprioceptive experience will enable reference to the person from one whom the experience derives.

  5. Hi Daniel,

    Here are four things I have been thinking about in connection with your paper — sorry if any of them duplicate things you’ve said already.

    1. I do want to hear what you have to say in response to Francois’ question that he reiterates above. Perhaps the answer is that the Demonstrative Model isn’t supposed to provide an account of what’s distinctive about de se thought as opposed to other ways of thinking about oneself (including e.g. Perry’s thought about the person making the mess), and that some kind of supplementary account would be needed to explain this distinction. Is that right? Or do you have something else in mind? (And if it is right, do you have any view on what the supplementary account might look like?)

    2. I have a half-baked worry about the inferential glue account. Imagine a case where someone takes himself to be tracking a single object and forming various thoughts about it, but in fact the object keeps being switched out: perhaps the red thing he sees at t1 (when perhaps he can’t also make out its shape) is a different thing than the round thing he sees at t2 (when he can’t also make out its color), etc. In this case, when at t1 he thinks “That thing is red” and at t2 he thinks “That thing is (also) round”, will the inferential glue account predict that he’s thinking of the same thing at both times, since he’s going to infer “That thing is both red and round”? This seems like a bad outcome. (Perhaps I am misunderstanding the account, though.)

    3. One phenomenon that seems to me to help you in your case against the Simple Rule Model is that of very young children who use the pronoun “you” to refer to themselves. (In my experience this happens especially to first children, since they often hear themselves referred to as “you”, and don’t have any siblings whom they also hear addressed in that way. In my house this led to hilarious attempts to clear things up, which led to even more hilarious attempts to explain the use-mention distinction to a two-year-old … but I digress.) When such a child forms a thought that she’d express with the sentence “You need a snack”, clearly we want to say that she’s thinking about herself rather than someone else. But the Simple Rule Model can’t explain this, right? Whereas your view can: she’s thinking about herself because it’s her own hunger that she has knowledge of, rather than someone else’s. I wonder what you think of this.

    4. Finally, a quick thought on your discussion of Anscombe’s example of the person who forms “I”-thoughts despite having no sensory feedback from his body. I agree with what you say about introspection being a source of knowledge in this sort of case, but would add that something similar could be said by philosophers who think that our main source of knowledge of our propositional attitudes is not our introspection of them, but rather our capacity to make up our minds. (I am one of these philosophers.) That is, supposing that I am in the situation Anscombe describes, I will be in a position to know that (say) I believe that there will be a third world war not by looking inward, but rather (as Evans says) simply by considering *whether* there will be a third world war, and knowing my belief in virtue of my reflection on the relevant facts. The main reason I bring this up is not to push such an account on you (though I do think it’s pretty much the right one), but rather because it points to a pretty significant way in which your account goes beyond Evans’s, at least in terms of rhetoric. This is that insofar as he treats self-reference as a kind of “information-based” thought, he seems to have to center his account of it on cases where people are *in receipt of* information about themselves, and so runs into trouble in situations where such information is hard to come by. By contrast, since your account is *knowledge*-based instead, you can appeal to *any* of the ways we have of gaining knowledge about ourselves, whether passively (by receiving information through inner or outer sense) or actively (by exercising our agential powers, either in thought or in action). Does that make any sense at all?

    John

  6. Joel:
    Thanks for the four great questions, which I’ve posted below along with my replies to them.
    (1) “This is side-issue, really, but I would want to know how you respond to a variant on the causation account of reference determination in the perceptual that appeals to the causal relation and triangulates with this dispositions to act towards a particular object (I’m thinking of Burge’s response to Davidson’s triangulation argument). If asked who he is talking about, Emmanuel will point to Celia, not to Elsie or to any other causal intermediary.”
    Suppose we agree that vision can on some occasions be something that helps secure reference (say, in my thought, “That is a finch”). It still seems like a good question to ask how vision manages to do that. Knowledge-maximization is one general principle that you could appeal to to answer that question. Now suppose someone says that dispositions to act can on some occasions help secure reference. It equally seems like there’s a good question to ask how dispositions to act manage to do that. But I’m not sure what general principle there is to appeal to to answer this question. Until I’ve heard what the general principle is, or why there doesn’t need to be one, I’d be agnostic about the potential for dispositions to act to help secure reference. (I should probably go away and read up on Burge’s suggestion, and see what he says though.)
    (2)“An version of the Simple Rule Model that is properly explanatory (that has, one might say, psychological reality), would be a descriptive one, according to which the I-thought “I am F” is a thought to the effect that the subject of this thought is F (or, alternatively, the producer of this token). I realise that you have some qualms about the notion of a producer, but I’m not yet too worried. Do you think that thought-insertion cases show this to be problematic?”
    Would this view be properly explanatory? I’m not sure. It would replace the question of how I refer to myself in first-person thought with the question of how I demonstratively refer to my own thoughts. As Evans pointed out, this seems to be a question of the same general kind.
    I would also want to know what the positive motivation for the view is. There seem to be several arguments against it. There’s the thought insertion point you mention. Phenomenologically, it seems completely implausible. Finally, it seems to imply that first-person thoughts are thoughts of an very cognitively sophisticated kind, despite the fact that the view that first-person thoughts play an extremely basic role in our lives – e.g. in the explanation of action – is a commonplace of the literature on first-person thoughts.
    (3)“I’m not sure why we should accept that K-max entails that the referent of I in the proprioception based “I have crossed legs” is me, rather than my body. You allow that reflection on the content of bodily awareness alone does not show that I am the object of bodily-awareness (I agree), so it may be that I am a distinct (perhaps constituted) object. In which case, don’t we need some further reason to suppose that the referent is me?”
    I was assuming that “I” refers to a kind of thing that can have bodily properties, and about which bodily awareness can therefore provide knowledge. For example, it might refer to a body, or to a human animal. Or it might refer to something constituted by a body or by a human animal. Even I am distinct from my body but constituted by it, it remains plausible that bodily awareness will provide knowledge about me as well as about my body – in that sense, there will be more than one object of bodily awareness. By analogy, vision provides knowledge about statues as well as about lumps of clay, even if the former are distinct from and constituted by the latter.
    You’re right though that if I am distinct from but constituted by my body, the knowledge-conferring potential of proprioception doesn’t on its own explain why my “I have crossed legs” thought is about me, rather than about the body that constitutes me. Rather, it rules out all the candidate referential assignments for “I” other than those two. Inferential glue is then needed to explain why the right candidate is me rather than the body that constitutes me. I will be prepared to move from “My legs are crossed” and “I am hungry” to “Something whose legs are crossed is hungry”. On the assumption that the body that constitutes me cannot itself be hungry, that provides a reason to prefer the assignment on which “I” refers to me, not to the body that constitutes me.
    Admittedly, there are some views in the personal identity debate that I, in defending The Demonstrative Model, presupposed the falsity of. E.g. I presupposed that “I” does not refer to a Cartesian Ego. But I don’t feel bad about that. Claims like “I have crossed legs” are sometimes true. On the face of it, this requires that “I” refers to a kind of thing that has some bodily properties, which a Cartesian Ego does not. Of course, you could try providing a non-standard semantics for the sentence “I have crossed legs” – e.g. one on which it involves deferred reference (the “I” in it refers not to the thing that it normally refers to, but to something that stands in some relationship to). But I don’t think that such a treatment would be semantically plausible.
    “A worry about the inferential glue account might arise if people (falsely) believe that different uses of “I” refer to different things. So, for example, a Cartesian might claim that uses of “I” in psychological self-ascriptions refer to a soul, whilst uses in bodily self-ascriptions refer to a body (I think that Galen Strawson holds a view not entirely unlike this – although he’s not a Cartesian). If they are wrong, then inferential glue will give the wrong result won’t it?”

    I don’t think it has to.

    Suppose Rene judges, and knows, “I am in pain”. Knowledge-maximization predicts that his thought will be about himself. If he makes a different first-person judgment that is not knowledge but that he treats as being about the same thing as his “I am in pain” thought is about (e.g. “I am in pain of threshold level t”), inferential glue explains why that judgment is also about him.

    Suppose he judges, and knows, “I have crossed knees”. Knowledge-maximization again predicts that his thought will be about himself. If he makes a different first-person judgment that is not knowledge but that he treats as being about the same thing as his “I have crossed knees” thought is about (e.g “I have knees crossed at exactly 45 degrees”), inferential glue again explains why that judgment is also about him.

    So, the picture is that all Rene’s mental first-person judgments are to be assigned a single object – because of inferential glue. And all his physical first-person judgments are to be assigned a single object – because of inferential glue. The object in both cases is the same not because he treats his physical first-person judgments and his mental first-person judgments as being about the same thing – that is precisely what he refuses to do. The object in both cases is the same because the object assignment of which maximizes knowledge is in both cases the same – namely Rene himself.

    The inferential glue principle I was relying on is the principle that if a subject treats two thoughts as having the same object, that’s a reason for preferring assignments on which they do have the same objects. Another principle in the same general area that one might consider is the principle that if a subject treats two thoughts as having different objects, then that’s a reason for preferring assignments on which they do have different objects. What I’ve just said about Rene doesn’t require me to reject this principle. It could be that there really is some pressure to treat Rene’s different tokens of “I” as being about two different things – just because he himself treats them as being about two different things – but that this pressure is trumped by knowledge maximization (if there is only one object there to have knowledge of, assigning different objects to Rene’s mental judgments and to his physical judgments will end up making the members of one of the two categories of judgment turn out not to be knowledge). Equally though, I don’t think I’m required to accept the principle.

    I think the moral of the Rene case is that thoughts such as “I am I” can be informative, just like thoughts such as “Paderewski is Paderewski”. If we think of first-person thoughts as associated with special first-person modes of presentation, then the Rene case forces us to acknowledge that there is more than one first-person mode of presentation. That doesn’t mean there isn’t reason to give an account of what these first-person modes of presentation have in common, in virtue of which they are first-person modes of presentation. It doesn’t make the category of first-person thoughts hopelessly heterogenous.

  7. John:

    Thanks for these great questions.

    1. I do want to hear what you have to say in response to Francois’ question that he reiterates above. Perhaps the answer is that the Demonstrative Model isn’t supposed to provide an account of what’s distinctive about de se thought as opposed to other ways of thinking about oneself (including e.g. Perry’s thought about the person making the mess), and that some kind of supplementary account would be needed to explain this distinction. Is that right? Or do you have something else in mind? (And if it is right, do you have any view on what the supplementary account might look like?)

    Reply coming tomorrow…..

    2. I have a half-baked worry about the inferential glue account. Imagine a case where someone takes himself to be tracking a single object and forming various thoughts about it, but in fact the object keeps being switched out: perhaps the red thing he sees at t1 (when perhaps he can’t also make out its shape) is a different thing than the round thing he sees at t2 (when he can’t also make out its color), etc. In this case, when at t1 he thinks “That thing is red” and at t2 he thinks “That thing is (also) round”, will the inferential glue account predict that he’s thinking of the same thing at both times, since he’s going to infer “That thing is both red and round”? This seems like a bad outcome. (Perhaps I am misunderstanding the account, though.)

    This is definitely a tricky issue for me. Consider the following two principles.

    “If two thoughts have the same reference, then assign them the same reference.”

    “If a subject treats two thoughts as having the same reference, then assign them the same reference.”

    It’s clear that there are cases on which these two principles issue different verdicts, and it’s clear that, wherever they do, the first principle’s verdict is right, and the second principle’s verdict is wrong.

    So I can’t endorse the second principle. I meant to endorse a weaker principle that says that a subject’s treating two thoughts as having a single object provides a reason – a defeasible one – to prefer an assignment on which there really is a single object. The hope would be that in all cases where two thoughts have two different objects despite how things seem to the subject, the reason for preferring an assignment on which there really is a single object will be defeased. E.g. if the thing the thinker looks at at t1 really is red, and the thing the thinker looks at t2 really is round, then an assignment on which the first thought is about the red thing and the second thought is about the round thing may maximize knowledge – by making each of the two premises of the inference knowledge – even though it doesn’t make the conclusion of the inference (“Something is both red and round”) knowledge. This kind of “trading off” manoeuvre seems legitimate given that the underlying reason for taking a subject’s treating two thoughts as having the same object as a reason for assigning them the same object is knowledge-maximization.

    3. One phenomenon that seems to me to help you in your case against the Simple Rule Model is that of very young children who use the pronoun “you” to refer to themselves. (In my experience this happens especially to first children, since they often hear themselves referred to as “you”, and don’t have any siblings whom they also hear addressed in that way. In my house this led to hilarious attempts to clear things up, which led to even more hilarious attempts to explain the use-mention distinction to a two-year-old … but I digress.) When such a child forms a thought that she’d express with the sentence “You need a snack”, clearly we want to say that she’s thinking about herself rather than someone else. But the Simple Rule Model can’t explain this, right? Whereas your view can: she’s thinking about herself because it’s her own hunger that she has knowledge of, rather than someone else’s. I wonder what you think of this.

    Thanks, that does seem potentially helpful. I wonder though whether a defender of the Simple Rule Model might claim that the thought that the child would express, or attempt to express, with the sentence “You need a snack” is a first-person thought. Whatever definition of first-person thoughts we give will have to allow for the possibility of non-linguistic creatures being first-person thinkers. So, we need something like “A first-person thought is a thought that the thinker could express in language using the first-person pronoun if they had that pronoun in their language” rather than something like “A first-person thought is a thought that the thinker does express in language using the first-person pronoun”. But, if the first of these is the definition we have, it looks as though the child’s thought, which they express, or attempt to express, using “you”, will count as a first-person thought.

    4. Finally, a quick thought on your discussion of Anscombe’s example of the person who forms “I”-thoughts despite having no sensory feedback from his body. I agree with what you say about introspection being a source of knowledge in this sort of case, but would add that something similar could be said by philosophers who think that our main source of knowledge of our propositional attitudes is not our introspection of them, but rather our capacity to make up our minds. (I am one of these philosophers.) That is, supposing that I am in the situation Anscombe describes, I will be in a position to know that (say) I believe that there will be a third world war not by looking inward, but rather (as Evans says) simply by considering *whether* there will be a third world war, and knowing my belief in virtue of my reflection on the relevant facts. The main reason I bring this up is not to push such an account on you (though I do think it’s pretty much the right one), but rather because it points to a pretty significant way in which your account goes beyond Evans’s, at least in terms of rhetoric. This is that insofar as he treats self-reference as a kind of “information-based” thought, he seems to have to center his account of it on cases where people are *in receipt of* information about themselves, and so runs into trouble in situations where such information is hard to come by. By contrast, since your account is *knowledge*-based instead, you can appeal to *any* of the ways we have of gaining knowledge about ourselves, whether passively (by receiving information through inner or outer sense) or actively (by exercising our agential powers, either in thought or in action). Does that make any sense at all?

    I didn’t say what I meant by “introspection”. What I had in mind though was a light weight conception of it as the ordinary source of our knowledge of our mental state, whatever that turns out to be (the source could turn out to be our capacity to make up our own minds). So, in that sense, even Evans believes in introspection. But he doesn’t seem to give it any role in fixing the reference of first-person thoughts. This is because, I take it, he thinks an account of how the reference of first-person thoughts is fixed is identical with an account of how a thinker of a first-person thought will know which object they are thinking about, which is in turn cashed out in terms of the thinker’s capacity to locate themselves. Introspection doesn’t seem to be of much help in locating oneself, so it gets excluded (though it’s not obvious that some other sources of knowledge that Evans does want his account to mention – such as one means of knowing that one skin is clammy – really differ much from introspection in this regard.). But, as you say, if the motivation for the Demonstrative Model is knowledge-maximization, there is no reason to exclude introspection.

    Another related difference is that Evans tries to deal with the sensory deprivation case by appealing to dispositions for one’s thoughts to be sensitive to ways of gaining information. That appeal to dispositions might or might not be legitimate given the motivation Evans has for his account (Lucy O Brien argues that it isn’t in her book “Self-Knowing Agents”), but it definitely isn’t if the motivation for one’s account is knowledge-maximization. Dispositions for one’s thoughts to be sensitive to ways of gaining knowledge don’t make those thoughts into knowledge.

  8. Hi Daniel,

    Thanks. That all sounds compelling. I’m a little puzzled, though, about how your responses to my Q3 coheres with what you say about Rene. On the latter, you write:

    “Suppose he judges, and knows, “I have crossed knees”. Knowledge-maximization again predicts that his thought will be about himself. If he makes a different first-person judgment that is not knowledge but that he treats as being about the same thing as his “I have crossed knees” thought is about (e.g “I have knees crossed at exactly 45 degrees”), inferential glue again explains why that judgment is also about him. So, the picture is that all Rene’s mental first-person judgments are to be assigned a single object – because of inferential glue. And all his physical first-person judgments are to be assigned a single object – because of inferential glue. The object in both cases is the same not because he treats his physical first-person judgments and his mental first-person judgments as being about the same thing – that is precisely what he refuses to do. The object in both cases is the same because the object assignment of which maximizes knowledge is in both cases the same – namely Rene himself.”

    Well, you effectively accepted in answer to Q3 that K-max alone does not entail that the “I” in “I have crossed knees” refers to Rene himself. Rather, you said that inferential glue was required to secure that verdict. But your reason for thinking that latter was that you would be prepared to move from “I have crossed knees and I am hungry” to “there is something that has crossed knees and is hungry”. Fine. But that’s a transition that Rene refuses to make. So what justifies the claim that Rene’s use of “I” in “I have crossed knees” refers to himself?

    Or, to put it another way, given that Rene refuses to make such transitions, and supposing that he is happy to make transitions such as “I have crossed knees and I will outlive my body” to “I will outlive something that has crossed knees”, we have at least some reason to think that at least some of Rene’s uses of “I” do not refer to him (at least, if Rene is not identical to his body).

    Perhaps the answer is the same as you gave to John re the demonstrative case (i.e. a reason, perhaps, but not a conclusive one)? Basically, my worry is something to the effect that inferential glue leaves the referent of “I” hostage to what crazy people happen to think. And that doesn’t look like a virtue.

    Sorry if this is a muddle (it’s Friday night…)

  9. Hi Daniel,
    I really enjoyed the paper. Your appeal to the knowledge maximization principle is interesting and quite convincing.
    I have a sort of a “zoom out” clarificatory question, which might stem from ignorance of the literature (specifically, Williamson stuff). The knowledge maximization principle appears to be a principle about the existence of a metaphysical connection between reference and knowledge, in that the former is (at least) partially determined by the latter. Now mainstream cognitive scientist (as far as I know) utilize the notions of representation and computation, but not the notion of knowledge. In other words, in order to explain how cognition works – how it accomplishes various tasks – cognitive scientists do not (and need not) appeal to the notion of knowledge. Furthermore, traditionally, if I am not mistaken, the issues of representation and reference were taken as very closely related, if not identical.
    My question is: do you think that reference is a useful notion in cognitive science? If it is, then it seems it (the notion of reference) should be divorced from the notion of knowledge (contrary to K-max), or perhaps cognitive science should be reformed so as to utilize the notion of knowledge?
    Perhaps you think that reference isn’t a useful notion in cognitive science. But this seems odd, because our ability to refer to ourselves in first-person thoughts appears to help us perform various cognitive tasks.

    Best,
    Assaf

  10. Joel:

    You’re absolutely, right. There is an incoherence between what I said in answer to Q3 and what I said about Rene. The only option I can see at the minute is to write another cheque on the personal identity debate and say that the correctness of the account of reference-determination I have been defending depends on our being identical with, and not merely constituted by, the objects of bodily awareness. A cost, no doubt.

    More generally, your point about making the referent of “I” hostage to what crazy people happen to think is certainly a worry. It’s interesting that in “The Philosophy of Philosophy”, Williamson uses just that point in arguing against inferential role accounts of concept possession. So it seems plausible that Williamson wouldn’t say that, for example, the reason all of Emmanuel’s visual demonstrative thoughts about Celia are to be assigned a single object is that Emmanuel treats them as being about a single object. I don’t have any ideas about what else he might appeal to.

    One way an inferentialist might try replying to Williamson’s objections is to appeal not to beliefs subjects actually have and to inferences they actually make but to beliefs they are disposed to have and to inferences they are disposed to make. The odd theoretical commitments some subjects, e.g. Rene, have would be conceived of as noise that causes performance errors and thereby disguises what their underlying competence consists in. As far as I can see though, appealing to knowledge-maximization means that this reply is not open to me. Even if Rene is in some sense disposed to treat his different first-person judgments as being about a single object, his in fact not doing so deprives him of the knowledge he is in a position to have. Someone who wanted to appeal to Rene’s underlying disposition would need to appeal to something like a potential-for-knowledge maximization principle.

    Francois, John:

    Sorry, still thinking about whether and how my account can get the right results about which thoughts are first-person thoughts.

    Assaf:

    Great question, thanks.

    A possible analogy. Williamson elsewhere considers the suggestion that our folk epistemology – which gives a central place to the concept of knowledge – is a relic of stone-age thinking and that it ought to be replaced by Bayesian epistemology. His riposte is that Bayesian epistemology presupposes a notion of evidence and, on his view, our evidence is just our knowledge. So it turns out that Bayesian epistemology – or, anyway, the correct version of Bayesian epistemology – does give a central place to the concept of knowledge. The analogous view for the case you focus on would be that cognitive science, or anyway the correct version of cognitive science, will give a central place to the concept of knowledge.

    This view seems a bit heroic to me (and isn’t something Williamson ever suggests). I think the right think to say is that knowledge-maximization is relevant to the story about how thoughts – things at the personal level – get their reference but probably not relevant to the story about how the subpersonal states posited by cognitive science get their reference. Divorcing the notion of knowledge from the notion of reference (or the notion of representation – I wasn’t intending any interesting distinction between reference and representation ) to this extent does not go against knowledge-maximization if knowledge-maximization is construed as a principle about how thoughts get their reference. It does go against knowledge-maximization if knowledge-maximization is a principle about how intentional (/representational/referential) states get their reference. I argue in my paper that the examples Williamson gives support knowledge-maximization only if it is construed as a principle about how thoughts get their reference.

    One might argue that we should expect a unified theory of how intentional states get their reference. In that case, the fact that the knowledge-maximization explanation of how thoughts get their reference doesn’t generalize beyond thoughts would be a point against it.

    I’d feel more persuaded by this argument if we had a candidate for being a unified theory of how intentional states get their reference that was anyway plausible. It looks to me as though the most plausible views about how certain intentional states get their reference, aren’t at all plausible when construed as views about how all intentional states get their reference. Knowledge-maximization is one example. An opposite example is Ruth Millikan’s teleology-based theory, which looks like a much better theory for lower level states than for higher level ones. Nick Shea defends what he calls pluralism about the metaphysics of content in his paper “Naturalising Representational Content”. I am sympathetic to this pluralism.

  11. Hi Daniel

    “The only option I can see at the minute is to write another cheque on the personal identity debate and say that the correctness of the account of reference-determination I have been defending depends on our being identical with, and not merely constituted by, the objects of bodily awareness. A cost, no doubt.”

    Especially if we allow, as some do, that the object of bodily awareness can incorporate tools such as the white cane for the visually impaired.

    Actually, I think that the point probably dovetails with François and John’s question regarding the de se. In fact, there’s a more general question in the background concerning the notion of sense. Consider the following inference: (1) Hesperus is a planet, (2) Phosphorus appears in the morning, so (3) there is a planet that appears in the morning. It’s quite possible that both (1) and (2) are knowledge, yet this transition does not look as though it preserves knowledge.

    Now consider the “I” case. I form a judgement formed on the basis of bodily-awareness will be “I have crossed legs”, and K-max plus the identity of self and body (object of bodily awareness) entails that the “I” here refers to myself. Now both Rene and I each judge “I am hungry”. Rene refuses to infer “there is something that has crossed legs and is happy”, whilst I do so. We want to say that Rene is wrong, whilst I am right. But unless we have an account of what determines the sense of “I” in “I have crossed legs”, it’s possible that the inference is just like the bad Hesperus/Phosphorus one above.

    That’s all a bit of a muddle, but I guess my thought is just that even if we wheel in a metaphysical claim to fix the referent, we’re not yet going to be in a position to say whose inferential dispositions are the “right” ones, until we’ve got an account of the sense of “I” in the relevant cases.

  12. Joel:

    You noted in comment 8 that there was an incoherence in the combination I attempted of (i) an appeal to inferential behavior to explain why thoughts like “My knees are crossed” are about me and not about my body and (ii) the concession that some first-person thinkers , e.g. Rene, might not go in for the relevant inferential behavior.

    My first response was to drop (i) and make the Demonstrative Model conditional on my being my body (in which case there’s no task of explaining why a thought is about me rather than about my body). You’ve convinced me that the better response is to claim that Rene, even if he does not in fact infer in a way that treats his different first-person thoughts as being about a single thing, will at least find it primitively compelling to do so. He is dissuaded from inferring in the way that is primitively compelling to him by his philosophical commitments. In keeping with this, I would appeal not to a knowledge-maximization principle but to a potential-for-knowledge maximization principle. (I mentioned the possibility of giving something like this reply in my last comment, but suggested it wouldn’t interest me because it would involve abandoning knowledge-maximization. In fact though, I don’t see on reflection why I shouldn’t abandon knowledge-maximization, given that I’d be abandoning it in favour of something so similar, and something that can be motivated by reference to similar examples.)

    You point out in comment 11 that inferences in which a subject trades on co-reference, and which do in fact involve co-reference, divide into two kinds – the ones where the subject is entitled to trade on co-reference and the ones where the subject isn’t entitled to trade on co-reference. Just to have a label we might call the ones in which the subject is entitled to trade on co-reference same-sense-involving inferences. Intuitively, “I have crossed knees, I am hungry, So someone who is hungry has crossed knees” is a same-sense-involving inference. What explains this? If first-person thoughts were governed by a token-reflexive rule, one could say: treating the two thoughts as being about the same object is something you’re entitled to do because they are guaranteed to be about the same object. They are guaranteed to be about the same object because they have the same producer, and each is governed by the token-reflexive rule which says that it will refer to its producer. The defender of the Demonstrative Model can’t say this. They can say that treating the two thoughts as being about the same object is something you’re entitled to do because treating them as being about the same object is productive of knowledge. But they cannot offer a further explanation of why treating them as being about the same object is productive of knowledge. That is where the explanation bottoms out.

    Francois, John:

    Francois wrote

    “But you do not answer the question I asked in my comment: How do you account for the important distinction between ‘genuine’ first person thoughts and first person thoughts thoughts that concern the subject only accidentally? If I am the winner and through knowledge-transmitting testimony I gain the information that the winner will go to Tahiti, my thought that the winner will go to Tahiti counts as a first-person thought by the characterization you give. Or so it seems.”

    It would clearly be bad if my account made my thought that the winner will go to Tahiti count as a first-person thought. I don’t see how it does that though. Since I acknowledge that there is such a thing as the first-person thought type, I can say that my thought that the winner will to Tahiti isn’t a first-person thought because it isn’t a thought of the relevant type. Saying that isn’t especially illuminating – in effect it amounts to taking the distinction between first-person thoughts and non first-person thoughts as primitive – but it seems enough to show that the account doesn’t succumb to the “The winner will go to Tahiti” counterexample.

    For a more illuminating explanation, we need, as Francois says, an independent characterization of first-person thoughts. On the other hand, not just any independent characterization will do. Suppose that first-person thoughts are Bill’s favorite type of thought. Then I can say that the thought that the winner will go to Tahiti is not a first-person thought because it isn’t a token of Bill’s favourite type of thought. That doesn’t sound as impressive as saying that the thought that the winner will go to Tahiti is not a first-person thought because it isn’t reflexive. Presumably this is because reflexivity is a more intrinsic or explanatory property of the first-person thought type than being Bill’s favourite type of thought.

    But there are properties of the first-person thought type other than reflexivity that do seem fairly intrinsic and explanatory and that nothing prevents me from appealing to – e.g. its functional role properties. I can say that the thought that the winner will go to Tahiti is not a first-person thought because it isn’t a token of a type some tokens of which are included among the immediate deliverances of perception, and are immediately motivating of action. It’s controversial, or anyway I think it should be, whether it’s possible to distinguish the first-person thought type from all other thought types just by reference to such functional role properties. But uniqueness isn’t required here. All that’s required is that tokens of the first-person thought type have some functional role properties that the thought “The winner will go to Tahiti” lacks.

  13. Dear All,

    Thanks everyone for such great comments. I feel like I’ve learned a lot!

    Thanks so much to John Schwenkler for giving me the opportunity to present the paper on such a useful and fun platform.

    Best wishes,

    Daniel

Comments are closed.