Where is the problem, guys? What I am missing?

Hi Augustin,

Thanks for your message. Please do send your new draft to briscoe@ohio.edu. We’re reading Millikan now in the tutorial (“Biosemantics” and some chapters from Varieties of Reference) and will be considering potential responses to Burge.

All the best,
Robert

Hi Robert! Just in case you see this: I greatly appreciate your interest. I haven’t worked much on this paper since Marc posted it, but I would prefer, in case you are going to use it, that you use another draft. Email if you want it. Agustin Vicente

Welcome! Sorry for the late reply, I was away for a week and did not have access to internet. Of course, I think it would be great if you could share your ideas and the comments of your students with us.
Indeed, in a month we are going to have a workshop in Barcelona with Burge on his recent book, so I in the next weeks I’ll be working on it. Thus, any comment, reply, idea, question or interesting paper concerning this topic will be very helpful. I myself will probably put some questions or comments on the blog, and you are very welcome to start any discussion you whish.

Yes, they have been helpful. Thanks, Marc.
Of all the things that you say, I think that there are two that are particularly interesting and important.

The first is: ” If Teleosemantics can account for sensory systems and ‘representational’ systems using the same reductive tools, then it seems sensory systems and ‘representational systems’ are two subkinds of the same natural kind (as you claim). But then, it seems to follow that both systems generate representations (or if you want, ‘intentional icons’, ‘signs’, or whatever have you) with content (or accuracy-conditions or truth-conditions….)”.

I don’t really know why it should be that if we use the same reductive tools for sensory systems and for representational systems bot of them must have veridicality-conditions. Sensory systems and representational systems could both be biological functions of the same generic type, yet be different regarding their having or not veridicality conditions. It seems that being representational -and having veridicality-conditions- has to do with being able to reach the external world by reconstructing distal causes, i.e., by implementing formation laws. So, I don’t know why I should be committed to holding that both sensory systems and representational systems have veridicality conditions. Perhaps I have overlooked something.
Anyway: would you say that teleosemantics necessarily blurs the distinction between representation and sensation?

The second interesting piece is the last, that is, the one about whether what I call “backward-looking functions” are effects or not. I guess you are right here: I have been misled by Dretske’s initial account. Still, there seem to be a difference between representations and the effects of representations. I think Burge’s insistence that NS selects on the basis of the effects of the representations, and not on the basis of their content has a point. I have to think more about it. It would be good to do it together if you go on with Burge’s book.

I think it is a very interesting paper and I have to say I agree with most of the things you say. In particular, I agree with the first two replies you give to Burge’s arguments.

Just two comments on them: sometimes, Burge seems to be arguing against the idea that psychological explanations are reducible to biological explanations. But Teleosemantics does not seek to reduce kinds of explanations, but entities (and it is not obvious the latter implies the former). Teleosemantics tries to reduce certain psychological kinds (representations) into biological kinds; but both psychological explanations and biological explanations can be vindicated. Even if one thinks everything can be reduced to physical facts, one need not deny that all explanations have to be reduced to physical explanations. For sure, Millikan thinks that explanations in biology and psychology are perfectly ok.

Secondly, I agree with you that Teleosemantics may be able to account for sensory systems with the reductive tools used in representational (cognitive) systems. Nevertheless, I think there is a tension here between that claim and an assumption that you seem to grant to Burge, namely that there is an important difference between sensory and representational systems. If Teleosemantics can account for sensory systems and ‘representational’ systems using the same reductive tools, then it seems sensory systems and ‘representational systems’ are two subkinds of the same natural kind (as you claim). But then, it seems to follow that both systems generate representations (or if you want, ‘intentional icons’, ‘signs’, or whatever have you) with content (or accuracy-conditions or truth-conditions….). So that would show that, after all, sensory and ‘representational’systems are not as different as Burge thinks and you seem to grant (personally, I’ll be happy with this conclusion). Of course, a lot depends on the vague expression ‘different in important respects’. But at least, it is important to stress that there is a tension here.

Now, most of my questions concern the discussion of Burge’s second argument. As you present it, the argument seems to be that, since sometimes producing a false representation is adaptive, the representational function cannot be reduced to biological function.

First, I am not sure I fully understand the argument. As I understand it, the claim is that sometimes a false representation may generate a successful behaviour. That is surely true, but I do not see why it should be a problem. In general, for any trait (hearts, kidneys…), failing to fulfil a biological function may sometimes lead to an adaptive behaviour. Perhaps in certain occasions the fact that that, for instance, an animal’s leg is slightly injured causes it to stay home and that fact might (luckily) help him to avoid a predator; that is, dysfunctional traits can be adaptive in certain occasions. Still, the function is the effect that (in a preponderant number of cases) explain why the trait exits. So, the case of ‘false but adaptive representations’ would be a problem only if these cases were very common and somehow systematic, in such a way that they figured in an explanation of why the trait exists. I am not sure such thing can happen.

But, more importantly, I am not sure this is a good description of what the teleosemanticist says. For instance, Millikan claims that the content of a representation is the state that must be cited in the most proximal Normal explanation of the consumer perform its function. So it is not true that Millikan is merely reducing content to function; Millikan uses the notion of function, but other concepts are also important(e. g. Normal explanations and Normal circumstances are also crucial). So I am not sure, Burge is not arguing against a strawman. Content is not merely reduced to biological function, but to biological function plus Normal circumstances.

(p.7) Just after presenting the argument you consider a possible reply (that utterly fails); I do not fully understand this reply. The reply seems to be that if one thinks the content were something like sound in my vicinity (instead of danger around), then “it may possible to hold that the good or bad functioning of a representation entails increases or decreases in adaptability”. I do not see why the claim that it represents one thing or the other makes any difference; either if it represents sound in the vicinity or danger around the representation can surely be false and still the behaviour be adaptive, can’t it?

Finally, you claim that backward-looking functions are biological functions but not selected effects. Of course, that clashes with the standard etiological analysis of functions. Now, I think there is a way backward-looking functions can be conceived as selected effects, and so the etiological account can be vindicated. Let me explain.

The claim that backward-looking functions are not selected effects derives from the way you describe them. I take it that you are trying to establish a distinction between forward-looking functions (C being supposed to M) from backward-looking function (C being supposed to indicate F), in Dretske’s spirit. Now, you claim the (backward-looking) function ‘C being supposed to indicate F’ derives from the disposition of ‘C being caused by F’. So, I take it that you are interpreting ‘indication’ as carrying ‘natural information’ in the sense of Dretske (1981) (S carries information about F iff S has been caused by F). But one could assume a different analysis of indication: for instance, in terms of covariation or correlation (S covaries with F iff P(F|S) greater than P(S)). On this (weaker) interpretation of ‘indication’, C being supposed to indicate F’ derives from ‘C correlating with F’. And ‘C correlating with S’ can perfectly be said to be an affect, and hence a selected effect. (In contrast to ‘being caused by F’, correlations can be selected effects: think about chameleons. The biological function (selected effect) of the pigment-rearranging device is producing a color that correlates with the color of the surface he is sitting on.) If that is true, backward-looking functions can be selected effects. So, I think your argument against Burge is right but does not imply a revision of the etiological account of functions.

Well, I hope these comments are helpful.

Best,

Marc

About the fallacies like affirming the consequent, the problem is the following; on a broad representationalist theory of mind, you explain behaviour by appealing to certain mental representations. Now, you say that my conditional “If A, then B” does not map onto anything unless I also believe A (or not B). Thus, if in fact I believe “If A, then B” and B, on your view the conditional is not a representation at all. So if I behave following this fallacy, you cannot explain why I behave as I do by attributing certain representations (that I believe B by itself does not explain why I behave as if A where present). Similarly with conditionals in which I will never know the truth of the antecedent, cannot explain my behavior. This view seems to clash with a broad understanding of the Representational Theory of Mind. The point is quite general: denying that conditionals are representations has some explanatory costs. In fact, can you even say that someone believes that if A, then B, if the latter is not a representation and doesn’t map onto anything? I guess some of the previous questions turned on that issue.

There is a point in your proposal I don’t quite get. You wrote “I wouldn’t say that the content of “if A, then B” is B, assuming that we also believe that A. Rather, I would say that a conditional “If A then B” jointly with an affirmation of its antecedent maps B”. Now, I think if a representation A (let’s call it “REP-A”) maps onto A, then REP-A means A. (REP-A’s content is A). So, “If A then B” jointly with an affirmation of its antecedent maps B, then I guess “If A, then B” jointly with an affirmation of its antecedent means B (has a its content B). That is what looks a bit odd. Is a consequence of your view that if I believe A, “If A, then B”, and B, the last two believes have exactly the same content? (this is a better formulation of one of the worries I had above)

As for the last point; I thought you were intending to offer a teleosemantic account of conditionals, which is supposed to explain how we use conditionals in thought. Now, I think that, intuitively, if we deny the antecedent of a conditional, the conditional is true, while if we don’t affirm it nor deny it, then we still don’t know its truth value. What you seem to be saying is that in the second case it does not have truth value (in fact, it does not map anything yet) and in the first it is not clear (you do n’t want to commit yourself). I just say that you need to say something about these cases, because this is the main argument for rejecting Jackson’s porposal. You must show that at least you are in a better position to solve it.

Well, good luck tomorrow! 🙂

Marc

Hey Marc, cool, thanks for your comments! I think you’re right about PPR: I spend a disproportionate amount of space on the issue. That needs to change. As for your other comments:

First, you’re right that on my story a conditional that never has its antecedent affirmed or its consequent denied or (perhaps other things too–I leave this open) will not be an intentional icon. Compare a sentence that is begun but not finished immediately: if it doesn’t get finished somehow, it doesn’t manage to be an intentional icon. But that doesn’t mean it doesn’t have functions. It just doesn’t map anything in particular. It’s not problem for you to believe such things. You know what the conditional would map, if you were given any of the various additional items such that, jointly with the conditional, there is a mapping.

As for your second point here, I discuss related issues in footnote 3, though I want to remain noncommittal about this. It might turn out that (deductive) fallacies like affirming the consequent are quite practical ways to reason, under certain circumstances; it even might turn out that reasoning in accord with this pattern when certain further conditions are met (amongst them, when one knows that the cost of being wrong is sufficiently low) accounts for the proliferation of mental conditionals. If that’s so, then a conditional and an affirmation of its consequent and an affirmation that those further conditions are met maps the affirmation of the antecedent. I think this could be a virtue of my account: it may explains why we do so often reason in accord with these deductive fallacies. But again, I don’t know about this case; it doesn’t seem to me something we can easily settle with philosophical methods. That’s why I focus attention on affirmations of the antecedent and denials of the consequent.

I wouldn’t say that the content of “if A, then B” is B, assuming that we also believe that A. Rather, I would say that a conditional “If A then B” jointly with an affirmation of its antecedent maps B. The conditional has no content (in the sense of mapping). It only has relational and triggered functions. I don’t understand the problem surrounding ‘If not-A, then not-B’ and affirmations of A: I’m not committed to that mapping anything (though I did, in the last paragraph, leave that possibility open). Can you try to restate the problem?

Fourth, I very much want to avoid the claim that the potential content of “If A, then B” is equivalent to [(A and B) or (not A and not B)]. Rather, what I say is that the Normal explanation for the success of consumers of mental conditionals on occasions when those consumers also had categorical representations of the antecedent will appeal to the fact that the consequent held, indeed, that the state of affairs mapped by a categorical affirmation of the consequent was mapped jointly by the conditional and the affirmation of the antecedent (And similarly for denials of the consequent). This way we avoid all of the problems that attend truth-functional accounts of conditionals. I consider two truth-functional accounts in the paper and reject them both (one of them rather quickly, since no one ever proposed it anyway!)

As to your last point, I do make some comments relevant to this in footnote 3 and above. But suppose I didn’t make those comments (I might not want to make them anyway!). Why would that be a problem? Of course there is an important difference between having no opinion about whether A and believing that A is false. But it might be that neither of those epistemic states combine with the conditional “If A, C” to yield a mapping. Why is that a problem? Combine “If A, C” with “D” and you don’t get a mapping (beyond the mapping of D); why should denials of the antecedent be any different? Sorry, I guess I just don’t understand the question!

Thanks for taking the time to go through this with me! And sorry I haven’t responded to your last comments on your own paper yet. I’ll get there soon. But we have a big day coming up tomorrow: I’ll find out whether my job here lasts another three years or not!

Take care-
-Brian.

There are many interesting ideas in the paper. Here are some comments:

I liked very much the introduction, but I would not spend so much time presenting PPR. You do not really use it in your proposal and it does not seem to be important for the discussion. So I would suggest either to eliminate this part or reduce it.

Your proposal is intriguing. According yo you, “if A, then B” is not an intentional icon untill we know either that A or that not B. What I haven’t understood well, is what happens if it is impossible to know whether A or B (as in “If there is live after death, I will never know it”. Does that mean that this condition will never count as an intentional icon, and hence, it will never represent anything? If it is not an intentional icon, can I believe it?

On the other hand, the idea that “If A, then B” does not represent anything without A or no-B seems to have striking results. For instance, suppose someone (wrongly) reasons as follows: “If A, then B; B, therefore A”. We know this kind of reasoning is extremely common. But if the first premise is not an intentional icon at all, how can we explain what is going on here?

I also have some problems understanding the idea that the content of “if A, then B” is B, assuming that we also belive that A. That seems clearly counter-intuitive. One thing is to say the conditional implies B (assuming A), but saying that it literally means B is surprising. Does that mean, for instance, that (given that I believe A) the content of “if A, then B” and the content of “If no-A, then not B” is the same, namely B?

Here is related worry. Suppose we assume (as I think you are assuming) that the potential content of “If A, then B” is equivalent to (A and B) or (not A and not B). We could even assume that the content is (A and B) or (not A and not B) or (not A and B). Secondly, I think it is also plausible that a teleosemantic explanation of how conjunction and disjunction originate in cognition is quite straightforward. Then, why couldn’t we just explain conditionals in terms of mental conjunctions and dijsunctions? In that case, we will not have to deny that mental conditionals are representations, even if the subject does not believe A or not B. (Maybe that does not look very realistic, but that is an empirical question).

The last comment is that you haven’t explained what happens when I believe “If A, then B” and “not A”, according to your proposal. You claim that if I believe that no-A, the conditional doesn’t fulfill its function, but that seems to me insufficient. If I do not believe A, nor not-A, it does not fulfill its function either, but there seems to be an important difference between knowing that no A and not knowing whether A or not A.

I hope the comments are helpufl!
Best,

Marc Artiga