New APPS

Art, Politics, Philosophy, Science. Mostly Philosophy.

recent posts

about

    Wildcard Skepticism

    Might there be excellent reasons to embrace radical skepticism, of which we are entirely unaware?

    You know brain-in-a-vat skepticism — the view that maybe last night while I was sleeping, alien superscientists removed my brain, envatted it, and are now stimulating it to create the false impression that I’m still living a normal life. I see no reason to regard that scenario as at all likely. Somewhat more likely, I argue — not very likely, but I think reasonably drawing a wee smidgen of doubt — are dream skepticism (might I now be asleep and dreaming?), simulation skepticism (might I be an artificial intelligence living in a small, simulated world?), and cosmological skepticism (might the cosmos in general, or my position in it, be radically different than I think, e.g., might I be a Boltzmann brain?).



    “1% skepticism“, as I define it, is the view that it’s reasonable for me to assign about a 1% credence to the possibility that I am actually now enduring some radically skeptical scenario of this sort (and thus about a 99% credence in non-skeptical realism, the view that the world is more or less how I think it is).

    Now, how do I arrive at this “about 1%” skeptical credence? Although the only skeptical possibilities to which I am inclined to assign non-trivial credence are the three just mentioned (dream, simulation, and cosmological), it also seems reasonable for me to reserve a bit of my credence space, a bit of room for doubt, for the possibility that there is some skeptical scenario that I haven’t yet considered, or that I’ve considered but dismissed and should take more seriously than I do. I’ll call this wildcard skepticism. It’s a kind of meta-level doubt. It’s a recognition of the possibility that I might be underappreciating the skeptical possibilities. This recognition, this wildcard skepticism, should slightly increase my credence that I am currently in a radically skeptical scenario.


    You might object that I could equally well be over-estimating the skeptical possibilities, and that in recognition of that possibility, I should slightly decrease my credence that I am currently in a radically skeptical scenario; and thus the possibilities of over- and underestimation should cancel out. I do grant that I might as easily be overestimating as underestimating the skeptical possibilities. But over- and underestimation do not normally cancel out in the way this objection supposes. Near confidence ceilings (my 99% credence in non-skeptical realism), meta-level doubt should tend overall to shift one’s credence down.

    To see this, consider a cartoon case. Suppose I would ordinarily have a 99% credence that it won’t rain tomorrow afternoon (hey, it’s July in southern California), but I also know one further thing about my situation: There’s a 50% chance that God has set things up so that from now on the weather will always be whatever I think is most likely, and there’s a 50% chance that God has set things up so that whenever I have an opinion about the weather he’ll flip a coin to make it only 50% likely that I’m right. In other words, there’s a meta-level reason to think that my 99% credence might be an underestimation of the conformity of my opinions to reality or equally well might be an overestimation. What should my final credence in sunshine tomorrow be? Well, 50% times 100% (God will make it sunny for me) plus 50% times 50% (God will flip the coin) = 75%. In meta-level doubt, the down weighs more than the up.

    Consider the history of skepticism. In Descartes’s day, a red-blooded skeptic might have reasonably invested a smidgen more doubt in the possibility that she was being deceived by a demon than it would be reasonable to invest in that possibility today, given the advance of a science that leaves little room for demons. On the other hand, a skeptic in that era could not even have conceived of the possibility that she might be an artificial intelligence inside a computer simulation. It would be epistemically unfair to such a skeptic to call her irrational for not considering specific scenarios beyond her society’s conceptual ken, but it would not be epistemically unfair to think she should recognize that given her limited conceptual resources and limited understanding of the universe, she might be underestimating the range of possible skeptical scenarios.

    So now us too. That’s wildcard skepticism.

    [image source]

    [Cross-posted at The Splintered Mind]

    Posted in

    16 responses to “Wildcard Skepticism”

    1. Anon Avatar
      Anon

      I’ve always read Kierkegaard idea of the “absolute paradox” as a version of something like this (though I’m not sure that’s what SK intended!).
      As I interpret it: the paradox is that I can know everything but not know that I know everything–a problem of, as Rumsfeld would say, “unknown uknowns.”
      But the paradox arises only with the specific possibility of an absolute unknown, which I take to mean both beyond my capacity to know and absolutely different from me and from my knowable experience.
      Because the absolute unknown is absolutely different from what I know, cannot be measured in terms of its probability or its impact on my current state of knowledge. As the possibility of something absolutely different from me (in SK’s case, God), it includes the possibility of something that could completely subvert my ability to assess my present state of knowledge.
      So, the Kierkegaardian problem of the wildcard isn’t just that there are other, unknown possible sceptical scenarios, but there are unknown possible truths that would specifically disqualify my present assessment of probabilities. The absolute unknown is the possibility that my measure of the known is radically wrong.
      But the difference from your version is it doesn’t require there to be any conceivable skeptical scenarios at all (which would be relative unknowns), it only requires that anything be unknown, or rather that whether there is anything unknown be unknown.

      Like

      Reply
    2. Eric Winsberg Avatar
      Eric Winsberg

      I guess I have radically different intuitions about this sort of thing. The first clue was when I read this: “In Descartes’s day, a red-blooded skeptic might have reasonably invested a smidgen more doubt in the possibility that she was being deceived by a demon than it would be reasonable to invest in that possibility today, given the advance of a science that leaves little room for demons. ” I think contemporary science leaves just as much room as it ever did for Cartesian demons. That’s because all of our present evidence is entirely consistent with the existence of demons.* All evidence is! So, I don’t even know how to begin to think about my credences in skeptical hypotheses. I guess I think all of the credences I have are conditional on the belief that some set of my perceptions are in some way veridical. Take away that belief and my very sense of my own credences starts to melt away, the idea of a credence in anything starts to seem incoherent.
      P(D/E)= (P(E/D)P(D))/P(E), but P(E/D) and P(E) are the same no matter what E is, so all the work is done by my prior on D, (where D is the demon hypothesis). In other words, no novel evidence can change my credence in D. The same is true of any good radically skeptical scenario.

      Like

      Reply
    3. GFA Avatar
      GFA

      Nice! You probably already know this, but the move here is pretty close in important ways to Kyle Stanford’s “Problem of Unconceived Alternatives,” which grounds/ explains his “New Induction” over the history of science. The ‘unconceived alternatives’ are scientific theories that would explain all the data available to the scientific community at time t (at least as well as the currently-accepted theory), but won’t be thought of until some later time. (E.g. Newton’s theory can explain the data available to Ptolemy, but nobody thought of it until centuries later.)
      There’s an interesting potential disanalogy too. In his book, Stanford is at pains to distinguish his skepticism/ instrumentalism about (highly) theoretical science from the sort of general, external-world skepticism covered in this post. Stanford says the difference consists in the fact that although we have given up the (successful, long-lived) geocentric cosmos for the heliocentric one, and given up the (successful, long-lived) Newtonian theory of gravitation for the general relativistic and quantum ones, there is no parallel case of us giving up the external-world hypothesis.

      Like

      Reply
    4. Sylvia Avatar
      Sylvia

      Hi Eric, your idea reminds me of Shimony’s catch-all hypothesis (which in turn seems very similar to the Unconceived Alternatives to which GFA refers). I have submitted a paper on the old evidence / new theory problem (joint work with Jan-Willem Romeijn) in which we try to figure out how to assign and revise probabilities to the catch-all hypothesis and to new theories that emerge from it. Our framework would allow you to include nondescript sceptical scenarios in the catch-all. However, we deliberately refrained from assigning a specific probability to the catch-all hypothesis. Instead, either we assign it an undefined symbolic value, or we only assign conditional probabilities (conditionalized on the union of explicit hypotheses). I can send you the draft if you’re interested in the details.

      Like

      Reply
    5. Eric Schwitzgebel Avatar
      Eric Schwitzgebel

      Thanks for the interesting and thoughtful comments, folks!
      Anon 23 July: Thanks for the pointer to Kierkegaard. Is there a particular part of his corpus where you think he is especially clear about this? The way you put it sounds like a radical rejection of probabilistic thinking. But I’m inclined to think we’re stuck with probabilistic thinking, like it or not, explicit or only implicit, as soon as we make choices. Do I drink this apparent coffee, or not? Do I try to fly as I’m walking across campus, or not? These choices reveal my subjective credences. They might be a problematic lot, but I’ll need to make the best of them or become a vegetable — although even the latter is a kind of choice.

      Like

      Reply
    6. Eric Schwitzgebel Avatar
      Eric Schwitzgebel

      Eric Winsberg: We seem to be starting pretty far apart, but here goes.
      On demons: I was thinking primarily of real, observable demons who cast evil spells and the like. It seems reasonable to me to suppose that credence of the average Anglophone philosopher today in the existence of such demons should reasonably be less than the credence of the average Western philosopher in Descartes’ day. And I think that if such demons are real it’s more reasonable to give a little credence space to an amped-up Cartesian demon than if such demons are not real.
      On the assumption that your perceptions are “in some way veridical”: I think here of Wittgensteinian “framework assumptions”. Are you thinking of something similar? I suppose I’m more of a Quinean here: Anything can be challenged, even core things, with enough evidence. Add a bit of Bayesianism to that and you’ll want to avoid p’s of 1 or 0. So then the question arises, what p’s should I give to these framework things? I agree with Wittgenstein that it’s silly to doubt (at any more than a negligibly tiny p) without grounds for doubt. But I do think I have grounds, for at least some slender doubt, regarding dream skepticism, simulation skepticism, and cosmological skepticism. This post adds the thought that there are meta-level grounds for wildcard skepticism.

      Like

      Reply
    7. Eric Schwitzgebel Avatar
      Eric Schwitzgebel

      GFA and Silvia: Thanks for the pointers on this. I am familiar with a bit of the philosophy of science literature on these issues, but it’s been long enough since I looked at it that I didn’t think to draw the (in retrospect obvious) connection to the content of the post. (I actually did one of my qualifying exam questions on the problem of old evidence.) So I appreciate the pointers! I agree that there’s some analogy, but also disanalogy, as GFA indicates. Silvia, yes, I’d be interested to see that paper. I should get my head back into this literature a bit.

      Like

      Reply
    8. Eric Winsberg Avatar
      Eric Winsberg

      “And I think that if such demons are real it’s more reasonable to give a little credence space to an amped-up Cartesian demon than if such demons are not real.”
      Yeah. I really don’t see this. Let’s go whole hog. Suppose you are walking around and you see some real observable demons. Red guys with pointy ears and horns. And you see them turn people into goats. And you see them put people into trances. In fact, you even see them put people into the sort of trance that, from all appearances on the outside, makes it appear as if they themselves are certain they are on a tropical island in the south pacific hunting unicorns with lollipops. What credence do you now give to the hypothesis that you yourself are being punked by an evil cartesian demon. The same one you do today, I would say. Because if an evil demon was trying to deceive you, Descartes-style, why would he include, in the hallucinations he gives you, images of demons like himself?
      Put more carefully, if D is the hypothesis of a Cartesian demon deceiving me, and E is some set of appearances I am having, the P(E/D) is the same no matter what E is. Because on D, I could equally well be having any set of appearance. On the assumption that E is always certain (its a reasonable approximation that I cant be wrong about what appearance I am having), P(D/E) is independent of E. I’m curious which premise you doubt in that argument.
      On the second point, I’m not saying that I am certain that my appearances are in some way veridical. I am saying that assuming that they are is a precondition for thinking about assigning credences to any of my beliefs. As soon as I entertain a radically skeptical hypothesis, the whole machinery of credences blows up in a way that I don’t think you are considering.
      Take Boltzmann Brains. Suppose I think, (as I indeed do!), that its appropriate to put some uniform probability measure of the set of microscopic initial conditions that are compatible with whatever I take to be the earliest macroscopic state of the universe–the one that I would ordinary assume it to have based on my observations. If I do that, then it seems as if physics tells me that it is overwhelmingly likely that I am a BB. BBs are much much more common in such a universe than are real people who are veridically experiencing the effects of a big bang. (that went fast, but it can be spelled out much more carefully). But if I am a BB, then the very physics which led me to believe I am a BB is bogus, and I no longer have reason believe I am a BB. The hypothesis that I am a BB is epistemologically unstable. So, I simply have to exclude it from my probability calculations by some kind of fiat. I see no other way out. (I discuss this in two papers. I will post links if people are interested.)
      A similar thing can be said about any of your examples. If the hypothesis D (cartesian demon) is true, then nothing whatever about what I take myself to be observing is at all informative about the probability of D.
      I was similarly puzzled about this line: ” It would be epistemically unfair to such a skeptic to call her irrational for not considering specific scenarios beyond her society’s conceptual ken, but it would not be epistemically unfair to think she should recognize that given her limited conceptual resources and limited understanding of the universe, she might be underestimating the range of possible skeptical scenarios.”
      Why should my understanding of the universe have any bearing AT ALL on the credences I should assign to various skeptical scenarios. It seems to me just as likely that I am dreaming everything I experience as it is that blzgebit is happening the real world that is causing me to experience the fake world that I experience, where “blzgebit” is some effect that has zero correlate whatsoever in my observable world. I might be dreaming in the real world, I might be matrixing, I might be evil demonized, or I might be blzgebting. Blazgebting might just be the kind of effect that makes me hallucinate a world where there is no such thing as Blzgebting. I shouldn’t take anything whatsoever that i take myself to be experiencing as remotely informative about the credence I ought to have in any of those hypotheses happeening in the real world.
      So: you are welcome to assign any a priori credence you like to such hypotheses. But I think you are underestimating the deeply epistemically undermining nature of these hypotheses if you think one’s credence in them ought to reflect, in any way whatsoever, what we take ourselves to be observing.

      Like

      Reply
    9. Eric Schwitzgebel Avatar
      Eric Schwitzgebel

      Eric, I really appreciate your pushing me more on this, and please send the papers. I’m embarrassed to confess that I hadn’t realized you’ve written about Boltzmann brains! I have a blog post about the extent to which the BB hypothesis undermines itself on the Splintered Mind here:
      http://schwitzsplinters.blogspot.com/2014/02/might-i-be-cosmic-freak.html
      I think that credences over 50% in the BB hypothesis are epistemically untenable, but I don’t think that the same instability objection undercuts small credences. And it would be an odd result, in my view, to think that one is compelled to assign a zero credence to being a BB, since it doesn’t seem like something I can know for sure (maybe .01% is far too high, but is 1/10^100 clearly too high?). But I’ll be interested to see your reasoning on this it your papers.
      On your point about demons: (1.) Why are you confident that demons wouldn’t include demons in their deception? I might think they probably wouldn’t — but if I see demons, the probability of deceiving demons shoots up so much that whatever reduction I give it on the assumption that demons probably wouldn’t show me demons might not put me back to my pre-demon-seeing Cartesian demon credence. And (2.) I would deny that p(E/D) = p(E/-D) for all E. It depends, in part, on demon psychology! Demon psychology is of course a very speculative topic, but we’ve already expressed some tentative opinions about it in our discussion of whether demons who wanted to punk me would want to let me experience a world containing demons.
      Now I don’t think demon doubt gets much traction now — compared, maybe, to a culture in which serious belief in demons might give me a decent credence in their existence and some sensible credence distribution on their psychology if they exist. So it’s pretty much groundless doubt, to which I would assign a negligible credence compared to dream skepticism and simulation skepticism (far below 1/10,000). But for dream and simulation skepticism I can start to give structure to conditional credences in ways that violate independence. For example if F is “succeed in having an experience of flying if I do something that feels like flapping my arms”, then p(F/dream) > p(F/-dream) — the topic of my very first NewAPPS post!

      Like

      Reply
    10. Eric Winsberg Avatar
      Eric Winsberg

      I’ll send you the BB papers, but the example might be a bit of a distraction here, since in the BB case, we are starting by talking about a probability measure which I take it has to represent physical chance. I agree that we can know for sure we aren’t BBs, but its hard to know how to set up the physical chances in such a way that ch(BB) doesnt come out to zero or 1. BBs are very different than other skeptical scenarios because if we are not careful, physics gives us the chances of them being true.
      Our more relevant disagreement come out here: ” For example if F is “succeed in having an experience of flying if I do something that feels like flapping my arms”, then p(F/dream) > p(F/-dream) — the topic of my very first NewAPPS post!”
      I guess I only agree with that reasoning condition on ~dream. If I am now and have always been dreaming (or if I am dreaming now but cant sort my real memories from my fake in-dream-only memories) then I have no idea how to set my credences, since all my a posteriori beliefs, including all my second order beliefs, are out the window. Similarly,
      ” (1.) Why are you confident that demons wouldn’t include demons in their deception? I might think they probably wouldn’t — but if I see demons, the probability of deceiving demons shoots up so much that whatever reduction I give it on the assumption that demons probably wouldn’t show me demons might not put me back to my pre-demon-seeing Cartesian demon credence.”
      Sure, I agree with the first part. I can’t be confident in this at all. But I completely disagree with the second part. Me seeing demons doesnt do anything to the probability. This comes out more clearly in my response to (2).
      As to your (2), I’m not sure which of my following premises or inferences you want to deny,
      1. P(D/E)= P(E/D) *P(D)/P(E), Bayes
      2. P(E)=1, for any possible E. Frist person authority on my own appearances.
      3 P(E/D)=1, for any possible E. D is, by definition, the hypothesis that a demon is giving me the E that I am having.
      therefore 4, P(D/E) is never updated in light of any new evidence.
      (I never said that P(D/E)=P(E/~D), I said it was equal to 1). You say that P(E/D) depends on demon psychology. I think this is the core of our disagreement. I guess I agree that it depends on demon psychology, but conditional on D, the only thing I can possibly learn about demon psychology is that they tend to give me experience like the on I am having now. The fact, that, for example, in my experience, they tend to hate their mothers tells me nothing whatsoever about whether they in fact hate their mothers (conditional on D).

      Like

      Reply
    11. Diana Tsang Avatar
      Diana Tsang

      The 1% skepticism argument sounds like an argument that Kierkegaard once advanced in the Postscript. Robert Adams skillfully summarized that argument in his paper, “Kierkegaard’s Arguments Against Objective Reasoning in Religion”
      Here’s a link to the paper if you’re interested:
      http://philosophyfaculty.ucsd.edu/faculty/rarneson/Courses/Adams2phil1reading.pdf

      Like

      Reply
    12. Anon Avatar
      Anon

      Of course, no part of his corpus is “especially clear” about anything! 🙂
      The passage I have in mind is the chapter “The Absolute Paradox” in “Philosophical Fragments.” It doesn’t explicitly talk about or deny probability, but I’ve read that into the text in order to make it to make sense to me.
      I agree that skepticism about probability is practically impossible, but it’s not impossible as an intellectual position. I can judge that my judgments of probability may be wrong while still acting as though they aren’t–even dispositionally and emotionally believing they aren’t.
      Kierkegaard would probably limit his skepticism to what he calls “existential truths” about value. So he might accept probabilistic judgments about mundane matters, he would insist on rejection of probability and the “leap of faith” in matters of religion and morality. I don’t find that satisfying myself, since I don’t think see why such “leaps” are better than probabilistic or thoughtless judgments, and I don’t think they’re psychologically possible anyway.
      Another variation of this is Quentin Meillasoux’s idea of absolute contingency, the possibility that the laws of nature might change at any time for no reason at all. His argument seems to be that if we cannot know any reason for natural laws being as they are, then it’s possible there isn’t any reason, from which the possibility of absolute contingency follows. That would then suggest any assessment of probability could be false.
      Meillassoux goes further in Beyond Finitude to claim absolute contingency is knowable not just a possibility, but I haven’t been able to make sense of his argument for that.

      Like

      Reply
    13. Eric Schwitzgebel Avatar
      Eric Schwitzgebel

      Thanks, Diana — I appreciate the reference, and I’ll definitely check it out!
      Eric: I suspect our disagreement is so complicated and multifaceted that it will take a long conversation to sort it out. I’m looking forward to reading your papers, and then I’ll send an email with further thoughts. In the meantime, though, I’m fine with (1) of course and I’ll spot you (2) for the sake of argument since I don’t think that’s the relevant issue (though I’ve argued elsewhere against strong first-person authority about experience). But I don’t understand why I would accept (3). Will I understand your thinking about this better once I’ve read your papers?

      Like

      Reply
    14. Eric Winsberg Avatar
      Eric Winsberg

      well, maybe only in the sense that thinking about boltzman brains specifically has influenced my thinking on this, and I think (3) is definitely true of the BB hypothesis. P(E/BB)=1 is true, I think, by definition of what it is to be a BB.

      Like

      Reply
    15. Anon Avatar
      Anon

      Thought I had replied to this, but must not have worked. Hope this isn’t a double post.
      I don’t ever find Kierkegaard to be “especially clear,” but this idea comes from the chapter called “The Absolute Paradox” in his book Philosophical Fragments.
      I agree that practically we cannot ignore probability, but we can still theoretically accept the possibility that probabilistic judgments are false.
      In K’s case, he limits his skepticism only to “existential truths” about morality and religion where a leap of faith is required to make authentic choices, so he probably accepts such judgments about mundane matters.
      A similar argument can be found in Quentin Meillassoux’s suggestion that the inability to demonstrate the necessity of the laws of nature implies the possibility of a radical contingency–that laws of nature could change at any time for no reason–which would, in turn, give us reason to be skeptical about probability judgments.

      Like

      Reply
    16. Eric Schwitzgebel Avatar
      Eric Schwitzgebel

      Eric, I’m not sure I agree with that, but maybe I would grant it for the BB hypothesis specifically. Part of the issue is whether in evaluating the BB hypothesis, you can make a prediction and update — maybe not, depending on how the thought-experiment is set up. More by email.

      Like

      Reply

    Leave a comment