Thursday, January 05, 2006

Transcendental Arguments

I want to discuss a type of argument which might be similar to what Kant called "transcendental arguments". (I'm not sure because I've never actually studied Kant. Must get on to that some day.) What I have in mind are those assumptions that we must make as a precondition to any sort of intellectual progress. Or, more generally, those things that we ought to believe because we've got nothing to lose by doing so. If they're false then we're screwed anyway, so we might as well just believe them and hope for the best. "Practical" might be a better label for this class of argument. After all, it's not as if the arguments do anything to establish the truth of the belief in question; they merely show that we might as well believe it.

The example I've used in the past is free will: If we can choose at all, then we must have free will. Therefore, anyone faced with the choice ought to choose to believe in free will. You can't possibly go wrong that way. Either you've got no choice at all, or else free will is the correct choice. Makes it pretty easy, don't you think?

This shows that we might as well believe in the preconditions for choice (namely, free will). For similar reasons, we might as well believe in the preconditions for rationality. Consider the laws of logic. I can't think of any non-question-begging justification for something so foundational as, say, modus ponens, or the law of non-contradiction. But we can't get anywhere without them. So we might as well just accept them, and hope for the best. (Thankfully, this seems to work.) The same goes for induction, and our assumption that the past is a reliable guide to the future.

Perhaps we should add in the notion of normativity (which I have difficulty getting a firm grip on). If there are normative truths, or facts about what we ought to do, then we ought to believe this. So either we ought to believe in normativity or else there aren't any facts about what we ought to do. So we can't possibly go wrong (i.e. against what we ought) by believing in normativity. Indeed, it's only by having this belief that we could possibly be doing what we ought to do. So, again, with nothing to lose, we might as well believe in it.

Is this good reasoning? Can you think of any more examples?


  1. Russell offers (in an offhand manner) and argument against free will: either our actions are caused and therefore not free, or they are random and therefore not free. ;-)

    On a serious note, Popper thinks that the thing you must believe is mathematics. i suppose you could add that we have to believe that we are not dreaming rocks -- or brains in vats. that our senses tell us something about the real universe which actually exists out there.

  2. Gawain,

    I think Hume offered that argument long before Russell. Although, it seems to me, its rubbish: That actions must either be caused or random is only true if you've assumed that there is no third - free - type of action. That is, the argument is question-begging.

    And on brains-in-vats, there's actually a very serious semantic transcendental argument that we aren't. Simplifying hugely (I can't be bothered to drag out my notes), according to some externalist theories of meaning, we cannot adequately refer to sceptical scenarios, such as brains-in-vats, and therefore they cannot be true.


  3. Alex,

    There is a famous article by Putnam where he makes the semantic argument to which you refer. (Is it a transcendental argument?)
    On Putnam's argument we can indeed refer to sceptical scenarios. I have the power to refer to brains and to vats - both refer to the objects of my experience. I can then imagine that some brain has been placed into some vat and fed artificial experiences.
    But I cannot suspect that I myself am a brain in a vat, for 'brain' and 'vat' refer to objects in my experience. If there is some other 'real' world beyond what is available to my experience, I have no means of referring to such a world or the objects in it, so I can't imagine it to be radically different from my experience.
    But this doesn't really solve the skeptical problem, for I can still ask whether I am in a position analogous to those of the brains in vats that I can imagine. (I believe this latter argument can be found in an article by Graeme Forbes; I can look for the citation if anyone is interested).

  4. Gawain - "Popper thinks that the thing you must believe is mathematics."

    Where's that from? I've only got LoSD and TOSaiE handy right now, but I've read most of his stuff and I honestly don't remember him making that argument.

    Richard - I don't see why "induction" is "foundational" at all, though I suppose that depends what you mean. If you mean something like "formulation of universal theories based on limited observations of recurring phenomenal patterns," then that's just something we do and has the same status as the fact that dogs bark.

  5. > Is this good reasoning?

    it is a bit like a weak form of pascal's wager guess. It runs into the same problems (ie alternative "religions" and so forth). Of course there is the "it works" argument which is presumably measured against some set of things that make you feel good about it.

    Also I am concerned that there might be an issue with overshooting the mark - for example I see there being many ways to look at "free will" I am unconvinced that because we dont want to accept the most anit free will view point that that means the most pro free will definition is the logical conclusion.

  6. Derek,

    Thats the one. I always found it unconvincing for the reason that you mention, although my lecturer did at one point have me convinced it wasn't so simple, but I've forgotten her reasoning.


  7. I think a more accurate way of outlining the general form of a transcendental argument is to put it as follows: 1. Identify the necessary preconditions for phenomenon P. (This is the major step, of course, and you just argue that, assuming P, some other things have to be true - it's, especially in Kant, mostly conceptual analysis.) Then, 2. Argue that it would be useless to deny P (for one of two reasons, that I think you blur together slightly: either 1. It is just stupid (pragmatically) to deny P, or 2. Any denial of P is automatically a (performative, usually) contradiction.) This then establishes the preconditions argued for in step one.

    So, for example, you could argue as follows:

    1. When I say "The cat is on the mat", it means that the cat (there) is on the mat (over there). This could only be true if my words are meaningful, and that could only be the case if I have some sort of intrinsic intentionality all to myself. (We need not bother with whether this is a good argument - that step there is dubious - just whether or not it's a transcendental one, broadly construed.)
    2. And it would be fruitless to deny that my words have meaning, since any denial would, a fortiori, be meaningless and hence not a denial of anything at all. The sentence "My words have no meanings" is only ever false when uttered, because when true it cannot be true.
    3. So I have a sort of intrinsic intentionality.

    When read this way it strikes me as being a little more powerful than just a 'nothing to lose' sort of appeal to pragmatic considerations.

  8. The problem with many transcendental arguments is that sneak in their premises and then make "magical" things appear. I think Kant does that. He starts with basically a Cartesian mind (minus the dualism - that is consciousness is present and inner). Then he asks what the requirements for knowledge are. In the original work this was largely a metaphysical question, although certainly by the mid 19th century the neo-Kantians had turned it primarily into an epistemological one.

    The problem, of course, is that many of Kant's premises might be wrong, with lots of post-Kantian thinkers (Hegel, Peirce, Neitzsche, and others) adopting different ones.

    The same is true of your example of free will, where people making these arguments sneak in their conclusions (IMO).

    The one that I dislike the most of this sort is radical emergence. All we have is brain stuff and consciousness has properties we can't explain by neurons therefore there must be radical emergence which produces new properties (consciousness). The problem is that there are two assumptions: that neurons described mechanically/chemically is all there is and that we can't explain consciousness by them.

  9. two questions on free will:

    1) what is the best argument for "free will" out there? and isn't there empirical evidence that refutes the idea of free will?

    2) why is it useful or necessary? that is, what happens differently if you don't "believe" it?

  10. But Dr. Pretorius, just because something is unassertable (on pain of "performative contradiction") doesn't mean it's false. So I don't see how we can move beyond pragmatics. Your example argument, for instance, is simply invalid. Premise 2 is far too weak to establish the conclusion; we require it to be true that our words have meaning, and not merely that "it would be fruitless to deny" this.

  11. Indeed - I am inclined towards denying free will (not that i think about it much). I assume I must be thinking of free will (and choose) in a slightly different context to richard.

    Otherwise a train and I would probably refute richard's hypothesis.

    anyway here we will do a litle substitution to see if we can make sense of it.

    "If a computer can choose at all, then it must have free will. Therefore, any computer should (choose to) declare (believe - same diference) it has free will." It can't possibly go wrong that way....

  12. I take it you're referring not specifically to what you earlier (in the linked to article above) called theoretically self-defeating principles, but only practical ones, right? It strikes me that any theoretically self-defeating principles are perfectly fine in the structure of a transcendental argument. (And I'm not convinced that my example isn't, at the least, more like your theoretically self-defeating cases than practically self-defeating cases.)

    As I've generally read transcendental arguments, though, it's mostly a non-starter (free will ones aside, because I find the whole area entirely confusing and possible confused). Arguments where denying the second premise (which amounts to "P exists/happens/etc") is impossible have a very solid second premise, of course. But the fact that some have weaker premises is not a strike against the transcendental argument itself - merely the commonplace observation that a valid argument can have false premises. (And as I've outlined it there will be two easy challenges to any transcendental argument: (1)P doesn't happen/exist/etc; (2)P does happen/exist, but not the way you say it does. These amount, appropriately, denying the two premises.)

    It looks like something more special, I think, because transcendental arguments tend to start with premises like "I have true beliefs about the world" or what have you, whereas in philosophy people are often more used to starting with more abstract principles and eventually getting to the more concrete ones. (Transcendental arguments are, in some sense, backwards.) I don't know that this is a problem with them, though, or at least if it is I'm not sure why. Personally I incline towards that sort of approach to philosophy, at least.

  13. Yup, I was referring to practical self-defeation only. (I'm not sure why theoretical self-defeation would require any further argument?)

    I don't fully follow your response, because as I pointed out, your example argument did not have a false premise in either of the two ways you suggest. Rather, it was simply logically invalid. But perhaps you had other examples in mind.

    Perhaps proper "transcendental" arguments are different from what I meant to be talking about? By your description, it sounds like they rest on substantive premises, which are perhaps common-sensical but nevertheless open to contention.

    The sort of argument I meant to point to, in contrast, makes no such assumptions as "I have true beliefs about the world", or "I make choices", or anything like that. Their premises are of a more conditional form, which I think should make them entirely uncontentious, even trivial. But the conclusions are significant or interesting nonetheless (though perhaps misleadingly so?), and I think this makes them a rather special form of argument.

  14. I'm not sure what you're after, then - aren't all arguments in some significant sense in a conditional form?

    The significance of a transcendental argument, I think, is that it generally appeals to the nature of whatever is being discussed to answer questions, though more and more I'm uncertain if I'm getting this right at all. For example, we might start with experience and consider its nature. Then ask what would have to be the case (metaphysically and otherwise) for it to have that nature, and conclude that things are so - because experience has the nature that it does. In terms of free will, I suppose, the argument would begin by pointing out the nature of free will, and what must be the case if there is free will, and then, finally, establish at the least that believing that there is such a thing as free will is more plausible than that there is not (and this is where the pragmatic considerations come in, I think).

    Since a large part of the argument is establishing a certain way the world could be given free will, then, the positing of free will gives you license to infer that the world is in fact that way. At least, generally I think that's how transcendental arguments tend to work - pragmatic considerations, of course, are not necessarily the best for nailing down the existence of the phenomenon. But sometimes I think that, in conjunction with the demonstration of their possibility, they're at least pretty good.

  15. Okay, well that just sounds different from what I'm wanting to talk about, then. Apologies if the title of my post is misleading.

    The arguments I'm pointing to don't depend on demonstrating "the existence of the phenomenon". In this sense, the conclusion is unusually unconditional. (I said the premises are conditional, which makes them weak and easy to satisfy. And, sure, there's a sense in which conclusions of arguments are always "conditional" on the premises; but since the premises in my arguments are so trivial, that doesn't really weaken the argument/conclusion at all.)

    What I'm pointing to is the fact that my arguments cannot be countered. Just look at my free will argument, or any others discussed in the post. I don't assume anything even remotely contentious (e.g. 'that we make choices', or anything like that). My premises are more or less analytic. Whereas someone could easily deny the premises of what you call "transcendental arguments" (as you yourself note), there simply aren't any assumptions being made in mine that you could possibly deny. Yet the logic seems entirely valid too. So the conclusion is unavoidable.

    That strikes me as a pretty special sort of result, for what appear to be non-trivial conclusions.

  16. err....

    "I don't assume anything even remotely contentious"
    "my arguments cannot be countered."

    when others have indicated your defence reduces it to a misleading / trivial argument.

    reminds me of the last few posts in the debate here

  17. Where did others "indicate" that?

    You asked whether it might apply to computers. If a futuristic computer ever found itself in a position of trying to work out whether it had free will or not, then I think my argument would give it reason to conclude that it does. It's not an argument for the truth of such a belief, remember. The computer might well be mistaken; but if so, it didn't really have any choice in the matter, so there's nothing can be done about that. (It's not as if you can admonish the machine for making the wrong choice!)

    Again, the argument is simple: "Either you've got no choice at all, or else free will is the correct choice." This should make the decision clear for anyone (or anything) wondering whether to believe in free will. Whether human or computer, once they've come this far, they've got "nothing to lose" by believing themselves capable of choice (which is all I mean by the term 'free will').

  18. Richard - as far as the arguments you've made go, my instinct is to say that while occasionally very good they're also deeply susceptible to mistake. For example - the phrase "if we can choose at all then we must have free will" sounds either dubious or perplexing (I'm still unsure what free will, or its absence, is supposed to be). So there's a deep danger of a false dualism at the very least.

    The other problem of course is the more pragmatic one - these arguments all leave a nicely sized space for someone with suitable contrarian tastes to simply bite the bullet and refuse to believe in these things anyway. And I'm not sure what could be said in that case. In actual life that sort of attitude always strikes me as a sort of perverse machismo, but in this case it seems to have a serious impact. Arguments that can simply (and, I suppose, rationally) be ignored by people who disagree with them don't have much force.

  19. matt macintosh:

    i dont think Popper offers this as an argument (that we must believe mathematics), but this assumption underlies his article in "CaR" on Austin's Why Mathematical Calculi Apply to The Real World (or something like this). sorry it is a sketchy reference, but i dont carry my popper around to the internet cafe...

  20. hello alex:

    you wrote: "That actions must either be caused or random is only true if you've assumed that there is no third - free - type of action."

    and such free action would be caused -- or random? (sorry, i dont seem the understand what "free" means in the sense you propose).

  21. upon reflection, i think i see what the problem is. "free will" is a judicial concept -- the courts see us as free to act in one way or another (and treats with lenience those who appear especially incompetent, such as the mentally retarded or the temporarily insane). as a legal theory it is fine (though it has its problems). but metaphysically speaking, there probably cannot be such a thing as "free action" (as opposed to either caused or random). (I dont know any scientist, in any case, who, when confronted with events in a testtube, will assume them to be "free" rather than "caused").

  22. Compatibilists don't think "free" and "caused" are mutually exclusive. See my post 'choosing determinism' linked in the sidebar.

  23. Richard, I am also a fan of this kind of argument. I'm not sure if/where they are discussed in the literature. I call them 'weak sure thing principle arguments' or 'might as well arguments'. They show that regardless of which conceptual possibility is the case, one option has at least as much of some positive property as the other and thus we might as well choose it. It is like the sure thing principle, but with a weak inequality rather than a strict one. This means that you cannot use them to construct an 'ought' or 'rational' or 'reason', but can get a 'might as well' or a 'not ought not' (sometimes defined as permissibility) or a 'not rational not' etc.

    My examples are belief in some kind of free will and anti-nihilism (i.e. against the view that nothing matters). We might as well believe that things can matter. In both cases, getting the arguments to be precise is a bit tricky. For example, is the pro belief in free will argument in favour of making choices, or anti-determinist. Is the view that something matters in favour of goodness or value or reasons? What if we have one without the others? Also, what if the universe is constructed such that belief in free will or in anti-nihilism is bad for us? Then the arguments don't seem to work. They seem to require belief in these claims as being positive or neutral if the claims are true. I'd encourage you to look further into this as it is quite interesting.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)