The varieties of skepticism

skeptic-communityby Massimo Pigliucci

In this video Dan and I talk about different types of “skepticism.” We start by recapping the history of philosophical skepticism, moving on to the (by now) classical question: how do we know that we don’t live in the Matrix? We tackle what we call Descartes’ “quaint” rationalism and we talk about a new and rather counterintuitive take on the concept of causality. Toward the end, we ask whether we really need solutions to the challenge posed by (radical) skepticism, and comment on what we think is the skeptic’s fundamental mistake.


Massimo Pigliucci is a biologist and philosopher at the City University of New York. His main interests are in the philosophy of science and pseudoscience. He is the editor-in-chief of Scientia Salon, and his latest book (co-edited with Maarten Boudry) is Philosophy of Pseudoscience: Reconsidering the Demarcation Problem (Chicago Press).

Daniel A. Kaufman is a professor of philosophy at Missouri State University and a graduate of the City University of New York. His interests include epistemology, metaphysics, aesthetics, and social-political philosophy. His new blog is Apophenia.


77 thoughts on “The varieties of skepticism

  1. Coel:
    Interesting response, but I think it still might be problematic. Let me try to re-express your argument in a more precise form, and then point out where I think there is a problem.

    Start with a sequence of N white balls, followed by one black ball. The white balls represent the days in which nature proceeds as usual, and the black ball represents the day at which nature changes. Then let the probability space consist in the collection of n element subsequences of white balls, where each subsequence consists of contiguous balls (in other words, we cannot skip a day). Let us denote this space S(N,n). S(N,n) consists of N-n+1 subsequences in total. All but one of these subsequences will be followed by a white ball.

    Therefore, if N represents the actual number of days at which nature proceeds as usual before changing, and we have been making observations for n days, then the probability that nature will proceed as usual tomorrow is given by the probability that we are not in the exceptional subsequence (represented by n white balls followed by a black one). The probability of this event is P = 1 – (N-n+1)^(-1). In the case that N is much larger than n, it follows that P is close to 1. On the other hand, in the case than N = n, P = 0.

    The problematic aspect is that in order to estimate probabilities we need an idea of the value of N, or at least some probability distribution for the possible values of N. It seems impossible to get any empirical data about this number.

    Before you try to think of a variation of your argument, consider that in order to apply a “frequentist analysis to the ensemble of possibilities” you must be able to say something about the probability space. In the above example, this is entirely contingent on the value of N, an unknown quantity.

    Liked by 2 people

  2. Massimo,

    >>“Many logicians would buy that probabilistic reasoning, of which induction is an instance, can be put into the form of deduction argument”

    >Can you provide me with sources of this? I assume you are not just talking about the literature on the validity of inductive arguments, nor about the literature of probabilistic deductive arguments (where the premises are of the type “most X are Y”).

    This is what I mean (sorry that I can cite only Wikipedia):

    Note the “simple induction” example. I would say more generally that all inductive arguments have at least a tacit deductive structure. I might even say that Hume’s problem of induction consists in there being no premise that satisfactorily completes a certain tactic deductive argument. The uniformity of nature is an attempt at this.

    >>“As to deductive reasoning “depending” on induction, I’m not sure what you mean by an “inductive premise””

    >Take the standard syllogism:

    >p1: All men are mortal
    >p2: Socrates is a man
    >C: Therefore, Socrates is mortal

    >How do we know that (a) all men are mortal, and (b) Socrates is a man? The first one by induction, the second one by direct observation.

    That’s a good point, but it more makes my point that induction exists with an deductive framework than the opposite, which seems what you are suggesting.

    >>“when we say regarding an inductive inference that the premises make the conclusion likely, what is the source of our certainty that the premises make the conclusion likely”

    >We have no such certainty.

    So with an inductive argument, there’s only a probability the the premises make the conclusion probable? Setting aside the potential for an infinite regress, if this is true then when we say:

    The premises make the conclusion likely

    We’re expressing more confidence than we’re warranted. We should say something like:

    It’s likely that the premises make the conclusion likely.

    But are we certain of this? Perhaps we should say:

    It’s likely that it’s likely that the premises make the conclusion likely?

    This is an another of example of why deduction must enter the picture somewhere, in my view.


  3. Quine:

    “As an empiricist I continue to think of the conceptual scheme of science as a tool, ultimately, for predicting future experience in the light of past experience. Physical objects are conceptually imported into the situation as convenient intermediaries not by definition in terms of experience, but simply as irreducible posits comparable, epistemologically, to the gods of Homer . . . For my part I do, qua lay physicist, believe in physical objects and not in Homer’s gods; and I consider it a scientific error to believe otherwise. But in point of epistemological footing, the physical objects and the gods differ only in degree and not in kind. Both sorts of entities enter our conceptions only as cultural posits.”

    The constancy of nature is also one of those posits that differ only in degree and not in kind from Homer’s gods. Not exactly a great foundation for induction.

    On a lighter note, I like the way Quine’s web subtlety, almost imperceptably implies the Homeric god-like figure of Spider-Man. Such a superhero of course needs no “epistemological footing”.

    Liked by 1 person

  4. Massimo: “Nobody has figured a decent way around Hume’s problem…”

    I would rather say that no one has found a justification for induction (in the relevant sense). Nor will they, since there cannot be one. However, it’s quite possible to find a “way around”, in the sense of explaining why the absence of any justification is nothing to worry about, and why (in a sense) it was a mistake to look for a justification in the first place. Indeed, Hume himself claimed to give a “skeptical solution” to the problem, whatever that means, and I think he was on the right track.

    In my view, what leads to excessive worrying about the problem is a misguided epistemology. We tend to see justification in an overly metaphysical way, and that leads to the feeling that we need justifications. I say we should see justifications as a useful tool, and not as a requirement.

    Liked by 1 person

  5. Coel,

    your explanation/answer to Goldman’s grue problem again shows that you don’t appreciate what the problem actually is, for the reasons that Aravis has already explained.

    “I’ve argued that on frequentist-probabilistic grounds”

    No probabilistic ground can ever get around the problem of induction, as several people have already explained to you, because all such arguments start with some sort of underlying distribution / prior, which needs to be justified… you guessed it, by induction!

    “What we’re suggesting has been argued by academic philosophers such as David Stove.”

    Ah, Stove. Let’s set aside for now his misogynist views as irrelevant, and his labelling of Popper and Kuhn as “irrational” (which is a bit more relevant, given that Popper explicitly took on the problem of induction).

    As far as I can tell, Stove’s alleged solution of the problem of induction is not taken seriously in epistemology. Other, more clever people (e.g., Samir Okasha, not to mention Popper himself) have also tried and have claimed success, but much of the philosophical community doesn’t seem to register the validity of such claims (though at the least Okasha contributed significantly to our understanding of the problem, and the nature of induction).

    Stove’s negative argument against Hume boils down to accusing the latter of “deductivism,” which is simply not the case. Indeed, Hume explicitly explains why any probabilistic argument is not going to work, and Stove’s attack is yet another version of the probabilistic argument (which I suppose is why you like it).

    As for Stove’s positive argument, it is again probabilistic, and therefore again misses the point entirely. Nobody has argued that it isn’t reasonable, on pragmatic grounds, to use frequency distributions and sampling to make predictions about the future. To insist that this sort of attack solves the problem is simply to misconstrue what the problem is.


    “This is what I mean (sorry that I can cite only Wikipedia):”

    Right, but the term “statistical syllogism” is a misnomer, since in a deductive argument the conclusion has to be entailed by the premises. If the conclusion is only likely given the premises then we are doing induction.

    Indeed, this is made clear a few sections above the part you link to on the same Wiki page:

    “Unlike deductive arguments, inductive reasoning allows for the possibility that the conclusion is false, even if all of the premises are true. Instead of being valid or invalid, inductive arguments are either strong or weak, which describes how probable it is that the conclusion is true”

    “That’s a good point, but it more makes my point that induction exists with an deductive framework than the opposite, which seems what you are suggesting.”

    I fail to see how, since the example I gave you is of a deductive syllogism based on inductive premises, not the other way around.

    “So with an inductive argument, there’s only a probability the the premises make the conclusion probable? Setting aside the potential for an infinite regress”

    That infinite regress is Hume’s problem.


    “I would rather say that no one has found a justification for induction (in the relevant sense). Nor will they, since there cannot be one”

    That’s correct.

    “Indeed, Hume himself claimed to give a “skeptical solution” to the problem, whatever that means, and I think he was on the right track.”



  6. Joe:
    “you indicate Quine thinks the core beliefs should be those with the greatest evidence. That is something I disagree with. I think the core beliefs should be those that align best with answering the most important questions”

    “But some of the latter may actually have the least epistemic warrant, so why would one put them at the core of the web?”

    It depends how you view warrant. If you think only in terms of theoretical rationality then yes my approach would have little warrant. But if you view this from the view of practical rationality arranging your noetic structure in this way may be warranted.
    This link explains how I understand the distinction:

    Looking at our situation it seems to me that we can assume we have some control over how we arrange our noetic structure. To the extent I view the arranging of our noetic structure as volitional, I use both theoretical and pragmatic reasoning. Thus how we do this will depend on what goals we are trying to achieve.

    We need to start somewhere in our beliefs. I start with the belief that it is important to act morally. I don’t think this belief has less warrant than say just jumping to the conclusion our senses accurately depict the external world. I do that as well, but that decision is even in some way after my first belief that I should act morally. Even if there were no external world, or it was drastically different than we thought, we might still be able to act morally in our thought processes.

    If your goal is to fill your noetic structure with beliefs that are more likely than not true, then I think Quine’s approach is good. And I do concede there is something odd in how the pragmatic approach works. When I say I “believe” something to some extent people would say that *means* I view it as more likely than not true.

    But I think there is another aspect to “belief.” And Quine happens to be the author that I think expressed it well. He said a belief is “a disposition to respond in certain ways when the appropriate issue arises.” When we view belief in this way that changes things a bit. Consider the situation where we have several different non-compatible views – say of meta-ethics or even ethics itself. We might think all of these views have less than a 50% chance of being true. Yet we might be “disposed to act” according to one view and I think that we could properly say we “believe” that view.

    I go into beliefs a bit more here:

    Liked by 1 person

  7. 2/2

    “It is rather the skeptic who begs the question if she wants to go so far as to say that we have no reason or evidence for believing anything at all. As Epictetus famously wrote, we don’t see any skeptic walking around as if she actually believed her extreme version of epistemic nihilism.”

    I think the skeptic properly shows that we have no evidence that the external world is not the result of a dream-state or some other scenario. To that question we have no evidence one way or another. I think Quine might just be skipping a step when talking about our senses as if they are evidence of the external world. I do agree he has “reasons” to act that way but I am not sure he has *evidence* to act that way.

    BTW I think Occam’s razor favors the skeptic. The skeptic just assumes we have impressions that there is an external world. The non-skeptic believes that *and* she believes there are all these different things that make up an external world. If all we need is a mind to get the impressions then adding all the other stuff seems to go against parsimony.

    I am not sure how I would expect someone who believed in epistemic nihilism to act. But I think many (not all) people who deny real objective morality act as if it were real and objective. Again if we go back to the notion that beliefs are dispositions I think this is to some extent a contradiction in their noetic structure.

    “But you are talking about values, not facts. And values are always underdetermined by facts, meaning that given the same fact there will always be different courses of action available to you, and the decision on which to take cannot be just a matter of empirical evidence (pace Sam Harris and Michael Shermer, of course).”

    In the law we often think a “fact” has a more specific meaning than how philosophers use the word.
    I think the way philosophers use the term “fact”, moral claims could be facts. E.g., “It was in fact morally wrong for Stalin to starve Ukrainians.” As an objective moral realist I do not have an issue with that claim. If someone denied it then I think they misunderstand reality. I disagree with Harris because I do not think “wrongness” gives off any sort of empirical indicia. So the notion that doing science can directly solve moral issues is ill-informed.

    The fact that moral wrongness gives no empirical indicia might lead some people to think it does not exist, or is not real. But I see no reason to believe that everything that is real in the universe is detectable by our five senses.

    “it just seems odd to suppose that what we really should do is try to fill our head with beliefs that have good evidence even if they seem trivial”

    “Indeed, though I’m not sure anyone here has suggested that.”

    I am not sure either. But if someone puts all the emphasis on theoretical rationality they may be heading there. We need to account for values and we won’t have empirical evidence on how to do that. I just admit that these sorts of beliefs are central to how I arrange my noetic structure.


  8. Patrick G wrote:

    What I sometimes miss in discussions like this is a *constructive* approach. Not an explanation why things don’t work, but why things work.


    A discussion of skepticism is probably not the best place to look for constructive approaches. The whole point of skepticism is to identify limits.

    As I said to Massimo in the dialogue, in my view, one of the chief values of skeptical inquiry is that it disabuses us of a mistaken picture of ourselves and our practices; one that is dear not just to philosophers but to many “ordinary” people as well.

    What it disabuses us is of an overly rationalistic view of ourselves and our practices. It does so by showing that in a fundamental sense, belief is prior to reason. And what that means is that every inquiry and more generally, every practice, begins with beliefs for which rational grounds — and thus, warrant — are not possible. I must already believe a number of things in order to prove, demonstrate, justify, observe…indeed, to engage in any epistemic practices at all.

    You ask “why does induction work” as if, somehow, “working” shows that induction connects us with some indepedent reality, outside our practices and frames of reference. But of course, what “working” means — what it is for something to “work” — is itself a matter of haing a relatively sophisticated conceptual scheme already in place. When we play the science language-game, it includes conceptions of success and failure, working and not working, relative to which we measure methods and tools like inductive inference and thus, unsurprisingly, things “work.” But all of these conceptions and practices are only comprehensible *inside* the framework in question.

    So, the philosopher, if he is good, can point all of this out — as Hume did; as Reid did; as the later Wittgenstein did — but once within the frame of reference, engaging in the practice itself, he has little to contribute. It’s up to scientists to discover and explain why certain practices and methods work and others don’t, within the scientific frame of reference.

    Liked by 5 people

  9. Massimo,

    Yes, there is that initial general definition of inductive arguments, but the examples I linked to are deductive. The examples show how (at least some) arguments of the kind that fit the initial definition can be turned into deductive arguments.

    >That infinite regress is Hume’s problem.

    I agree that that’s one way to describe it, but one might also say that the regress exists because of the absence of a completing premise that makes a sound deduction.


  10. I feel a bit sorry for Aravis (and somewhat Massimo, too) in responding to comments on this issue, especially vis-à-vis Coel and DM but also somewhat others.

    It appears some commenters here do not “get” Hume’s problem of induction, and, via various hand-wavings, whether Bayesian statistics or other probabilistic issues, don’t want to “get” it, either in Hume’s original, or as extended by Nelson Goodman. And, Goodman answers the “grue” issue exactly as how Aravis notes. Maybe Coel is indeed an essentialist; wouldn’t surprise me. This is also why I’ve not “engaged” with either Coel or DM on this issue; I have other things to do with my time; they’re either wanting to hand-wave because they don’t want to “get” it, or they simply don’t “get” it.

    And, Goodman also, to me, refutes the likes of Stove.

    As for reading some of this in general, per Marko or others? I mean, Hume is an easy-to-read philosopher, not just in terms of philosophy, but overall English writing style. A most basic issue of the problem of induction is his being questioned how he could go to sleep at night not even knowing the sun would come up tomorrow. And, with his sense of detachment about that and other things, he said it wasn’t a problem. Thus, he showed that while induction may be a problem philosophically, we normally live as though it’s not, psychologically — part of him arguably being the first psychologist.

    This relates to PatrickG. Cars drive over that bridge, and it doesn’t collapse — until it does. Yes, we can use engineering standards to make probabilistic guesses about a bridge’s lifespan, but, until it actually collapses, it doesn’t.

    Maybe thinking of this as a macro-world, classical statistical probability equivalent of Schrödinger’s cat will help? No, I’m not joking, while I note that’s not the best analogy. That bridge is “suspended” between remaining standing up and an increasing likelihood, defined probabilistically, of collapsing. But, until it actually collapses, it doesn’t “collapse,” so to speak.

    This gets back to Coel/DM. Just as one can’t use probabilities on an individual quantum event, one can’t use them on an individual next occurrence of a macro-world series of events, either.

    That said, yes, per Imad we shouldn’t, IMO, be putting all thinking into just two boxes. After all, we’ve also evolved beyond classical logic into modal and other multivalued logic systems. Indeed, someone like Haack may offer some insights:

    As for Hume, otherwise? He was indeed a skeptic. Whether he was a Pyrrhonic one or not is a matter for debate, in which I’ve been involved before, on an early essay here.

    Liked by 1 person

  11. My fifth post. But I just want to say that I hope we have more such discussions about the foundations of our beliefs. I think that such discussions are not only enlightening but I also think on an interpersonal level we can better understand where others are coming from.

    Hi Coel you say:
    “The justification of Occam’s razor is then again probabilistic. The number of conceivable statements is infinitely larger than the fraction of those statements that are real-world true.”

    It seems both sets are infinite. “There are fewer than 2 people named Thomas Hobbes that live in Urbana Illinois.” If that is true then the following statement would be true as well. “There are fewer than 3 people named Thomas Hobbes that live in Urbana Illinois.” etc……

    “Thus we can discount any statement for which there is no evidence, the chances of it being true are effectively zero. This excises Russell’s Teapot and Descartes’ Evil Genius.”

    I believe this is a poor way to calculate the probabilities of things we have no affirmative evidence for.
    There is more at work in making Russell’s Teapot less probable than just the lack of evidence. He is playing on the conjunction fallacy. Russell is not just saying some unknown general object is in space. Is the chance that there are objects in space that we do not yet have evidence for near 0? I don’t think people would believe that. Some of these stars might have planets and those planets may have moons etc. There is likely some debris or object existing somewhere in space that we currently have no evidence for.

    But whereever you put the probability of there being an object in space that we don’t know of it is less probable that the object will be metal. And no matter what you put the probability that there is some as yet unknown metal object in space it is less likely that it is kitchenware. And no matter how low you put the likelihood of there being metal kitchenware in space the likelihood of it being a metal teapot in space is even lower. And however low you put the probability of there being a metal teapot in space it is even a lower probability that the teapot is in between earth and mars.

    As the person adds details we naturally and properly decrease the likelihood.

    This sort of confusion is often used to try to justify someone having a philosophical burden of proof. For a gross example of this confusion you can see this youtube video:

    There we are not just supposed to believe in something being below the surface of Pluto but a living thing. And not just a living thing but a mammal. And not just a mammal but a walrus. And not just a walrus but a psychic were-walrus. Once you understand how the conjunction fallacy works you will see why these examples are silly and don’t really support the notion of a philosophical burden of proof.


  12. Sorry for coming in so late, but I wanted to listen to the dialogue attentively, which was difficult this week-end. Most commenters have used up their five posts, so I will try not to write contentiously.

    However, I do have to remark the occasional confusions between deduction, induction, and probability theory in some comments. Some, unfortunately, do smack of freshman precocity. Not just Philosophy 101, but Composition 101 (which I taught for twelve years). If one cannot distinguish deduction from induction, one’s arguments become immediately suspect.

    Untying probability theory from the problem of induction is a little more difficult, because induction is necessarily a probabilistic reasoning, as Aravis notes. But probability can never arrive at certainty, and without this, it can never undue the problems of induction, and there is no way around that. Where probability supposedly attains certainty, then we are discussing certainty, and not probability at all. Probability to be probability cannot guarantee certainty.

    I read Rathmanner and Hutter; and while I didn’t grasp all of the probability math, I certainly understood this: “M strongly confirms the black raven hypothesis under any reasonable sampling process, of course provided no non-black ravens are observed.” But isn’t that the whole point? – provided no counter example suddenly appears, in which case the whole theory – as a claim on universality – crumbles to dust.

    The problem of induction cannot be resolved by resort to probability theory. Can it have any resolution at all?

    Technically – no. The problem of induction, in its various permutations, really has to do with the problematic relationship between logic and human experience. Hopes to crunch this relationship into pure deduction or mathematics have proven futile – and hopes to crunch it into either computationalism or some other scientistic paradigm will probably prove similarly futile. Frankly, the effort to crunch it into probability theory seems a mere road-side attraction.

    Our objects of study always seems to come back to who we are – and we don’t know who we are.

    Knowledge is about things that interest us; I don’t think we’re really as interested in ourselves as we think we are. So we reach the limit of knowledge when we are forced to examine ourselves; and then find some way to redirect the study into other interests. This seems to be psychologically normative. Hume’s greatest contribution here may be his recognition of the psychologism grounding reasoning that so many philosophers have abhorred.

    Which goes to the question, about why, despite the limits of various forms of reasoning (and they all are limited, and they all do stand on prior beliefs, as Aravis notes), we still seem to function fairly well in the world and continue to accumulate knowledge both useful and esoteric.

    I wish this discussion had followed this tangent, rather than spurious efforts to deny an obvious logical problem. Epictetus was right – no full-blown skeptic acts as though complete skepticism were valid. That tells us more about this issue than any formulaic theory in logic or probability.

    Liked by 4 people

  13. Aravis,

    > You ask “why does induction work” as if, somehow, “working” shows that induction connects us with some independent reality, outside our practices and frames of reference. But of course, what “working” means — what it is for something to “work” — is itself a matter of having a relatively sophisticated conceptual scheme already in place. When we play the science language-game, it includes conceptions of success and failure, working and not working, relative to which we measure methods and tools like inductive inference and thus, unsurprisingly, things “work.” But all of these conceptions and practices are only comprehensible *inside* the framework in question.

    Hmmm. So that stone age cave dweller who deduced that he would burn his fingers if he kept them in a flame had a frame of reference, practices, a conceptual scheme, was playing language games etc.
    I don’t think so. You can’t make his burned fingers disappear with other schemes, games etc. They burn when you stick them in a flame. They burned thousands of years before humans knew what combustion, nerve cells etc. were. It was pure undiluted induction and it worked. Personally, I think the cave dweller acted rationally when he didn’t stick his hand in the flame.

    So, philosophically speaking, we have this situation (more or less): we have a philosophical tool (rational enquiry), we study something, and we come invariably to the conclusion that the something studied is not entirely rational. It is not entirely rational to deduce that you’re going to burn your fingers when you keep them in a flame. To be entirely rational, you should have justified induction – but you can’t, rationally speaking.

    When reading this blog I’m frequently thinking that rational enquiry is a blunt tool. A bit like using a hand axe to study a swiss watch. Obviously, if you study a swiss watch with a hand axe, you’ll be mystified. How come the damn thing works? Certainly not by virtue of mechanisms that can be studied with a hand axe. Perhaps there’s some circularity in using rational enquiry too, a circularity that’s just as circular as using induction? Pot, kettle, black etc.

    Liked by 3 people

  14. From the various types of inference/reasoning discussed above, for anyone still around with an interest in type theory — is a recently revised introduction — I’ve put a list of references to various types of judgements/judgments here (that now includes abductive and modal — if anyone has additional ones, let me know):
    “Types of judgments in type theory”


  15. Patrick G wrote:

    Hmmm. So that stone age cave dweller who deduced that he would burn his fingers if he kept them in a flame had a frame of reference, practices, a conceptual scheme, was playing language games etc.
    I don’t think so.


    Er…no. This is like saying that because the cave dweller is not aware of the rules of inference, his deductions are not governed by said rules. Or that there is no social contract, because none of us signed a piece of paper. It fundamentally misunderstands what is meant by a rule-governed activity.

    Of coruse science and all other rule-governed activities are language games, in the sense that Wittgenstein means. Reductio-ad-caveman arguments, while cute, don’t get us anywhere.

    Liked by 1 person

  16. Aravis,

    I know that one can use rules of inference even if one is not aware of them etc. I also know that one can frame many problems in terms of language games etc. I even think that such framing is often enlightening.

    I just don’t feel it’s very fruitful when we’re talking about a stone age cave dweller who uses induction to decide that he’ll burn his fingers if he sticks his hand into a flame. Language games etc. are a tool and using this particular tool in this particuar case is not very enlightening (although perhaps it learns you something about the tool).

    The same can more or less be said about Hume’s analysis of induction. In my opinion it tells more about the limits of the philosophical tools he uses than about induction.

    Thanks for the exchange, I have nothing more to add.


  17. Patrick G.

    Hume doesn’t employ any “philosophical tools.” He simply inquires whether something is knowable via sensory experience or reasoning. The conclusions he draws re: induction are based on his answers to questions framed in terms of these very common, hardly philosophically sophisticated epistemic means.

    As for the cave dweller, if you are ascribing language and intentionality to him, then you’ve already described him as operating within frames of reference — the two are not separable. Indeed, your very description of his behavior, insofar as it is rule-governed — and all linguistic and intentional behavior is rule-governed — is the description of a language game, so I just don’t see how you get any mileage against Hume (or Wittgenstein), by appealing to cave dwellers.

    Liked by 2 people

  18. Aravis,

    English is not my first nor my second language (and if you count the dialect I grew up in not even my third language **), so I suppose I didn’t express myself very clearly. I don’t think there’s anything wrong with the very common epistemic means of Hume or the less common approach with language games à la Wittgenstein. They’re tools and like all tools they have a domain in which their application is fruitful, and other domains in which their results aren’t. I personally feel that neither Hume nor language games etc. are telling me something interesting about induction. I’m not looking for mileage against Hume or Wittgenstein – what I know about Wittgenstein II makes him a very interesting thinker for me.

    I promise this is my last contribution to this discussion!

    (**) We have a Jewish lodger now, a phd-student at our university. To my great surprise some expressions in my dialect have exactly the same meaning and almost the same pronunciation as certain Yiddishe sayings.


  19. Like Ej, my apologies for dropping by so late. I did need to do a good bit of reading to hopefully get up to speed.

    While I certainly stand on Hume’s side, I also wonder if it would be more useful for us to not consider “the induction problem,” as a true problem at all? I’d have us simply consider this to be a basic human limitation for us to acknowledge, not something to fight. When compared against the very mechanisms of reality, we humans are surely just idiots. Apparently the only thing that we can truly “know” is that we “think,” as René Descartes observed in the previous century. I would go further however, by attempting to institute a generally useful definition for the term “thought.”

    I consider thought as “the conscious processor,” and it comes in two varieties. The first of them is “the interpretation of inputs” (with conscious inputs existing as “sensations,” “senses,” and “memory”). This variety of thought is roughly analogous with Ned Block’s “phenomenal consciousness.” Then the second variety of thought is entitled “the construction of scenarios,” and here we essentially figure out what to do given our motivation for happiness. This is roughly analogous with Ned’s “access consciousness.”

    I’m obviously quite happy that he’s coming as well!


  20. PatrickG The only way people can unknowingly use philosophical frames of reference, whether in construction of linguistic games or other things, is not if they’re cavemen but if they’re p-zombies, as I see it. Massimo has repeatedly indicated his take on them; I’m sure Aravis feels the same. I know I do. Other than that, inductive reasoning isn’t a “frame of reference.”

    Hope this helps. And, whatever your native language is, since Hume is a simple enough read in English, he should be in your language too.

    The best way to put it? Induction simply is. It’s not a frame of reference, nor is it a “tool.” Then, along comes Skepticism, with the first musings about what eventually became the “problem of induction.” The use of induction as a natural mode of thinking existed long before that, and surely before Homo sapiens evolved from previous hominids.

    Philosopher Eric gets at this himself. Intuition works for macro-world (neither quantum nor cosmo-world) events well enough. Hume knew that and, despite identifying the “problem of induction,” as I already noted, went to sleep happy at night.

    Back to Patrick. The fire example, that said, also likely isn’t inductive reasoning but rather instinctual behavior. But, per the fire example, induction may have first been an extrapolation from more instinctual behavior classes, specifically, reactions to danger. Autos don’t exist in the “natural” world, therefore, we don’t have an instinctual response to them. But, inductively, we learn that a car making a certain level of travel sound or whatever is normally traveling at Speed X, and we therefore decide to cross a street, or not, in front of it based on that.

    In fact, this fits with some of my amateur ideas on the evolution of consciousness.

    EJ is right about skeptics and their actions.

    That said, it should be noted that while empiricism, induction and skepticism have overlaps of some sort like a Venn diagram, they aren’t the same thing. One can be a skeptic without grounding one’s skepticism on the problem of induction, for example. One can bring an empirical mindset to life without worrying about the problem of induction, or without being a skeptic. One can seriously address the problem of induction without necessarily committing to skeptical philosophy, at least a more rigorous version of skepticism.

    As to Hume, I noted there was debate over whether he was a Pyrrhonist skeptic or not. Of course, there’s also debate whether the revival of skepticism in general at his time understood the ancient divisions the same way they were understood at that time. Also, Hume’s skepticism arguably did not derive from philosophy first, nor from the natural sciences; rather, it came from him being a professional historian.


  21. Fifth and last …

    Hi francisrlb,

    In the case that N is much larger than n, it follows that P is close to 1. On the other hand, in the case than N = n, P = 0. The problematic aspect is that in order to estimate probabilities we need an idea of the value of N, or at least some probability distribution for the possible values of N.

    Yes, agreed so far. And N can be n, n+1, n+2 … to infinity, and only N = n would violate induction tomorrow. It would be a remarkable fact if N were that special number n, and not any of the mundane values. Of all the conceivable probability distributions for N, only a tiny fraction make induction unlikely to hold tomorrow. It would be remarkable if it were one of those.

    One can still talks sensibly about sampling from a distribution even if one doesn’t know what the parent distribution is (non-parametric statistics deals with this). And, as above, I’m not claiming you can get certainty, nor that one can develop the argument from first principles, only that one can link the idea sufficiently with the rest of the Quine-style web to give sufficient justification for that likelihood.

    Hi Aravis,

    To ascribe “green” is no less “applying a label” than to ascribe “grue.”

    Agreed, but the nature of the label “grue” is very different from that of “green” for reasons that I explained, and thus the degree of justification for applying an inductive extrapolation — justification deriving from other parts of the Quine-style web — is very different.

    It sounds like you buy into some kind of essentialism about certain kinds of properties …

    No, not at all, I said nothing remotely like that.

    Wrong. As I explained in the dialogue, Quine’s web has two tiers. Observation statements are confirmed directly from experience, not via their inferential relations to other beliefs.

    Not so, there is no “tier” where observational statements are “confirmed directly from experience”. *All* observations are heavily theory laden. You cannot get to any English-language statement about an observation (or any any-language statement) without employing large swathes of the web. You are entirely right that the web is anchored to empirical observations, but it is the whole of the web that is compared to observation, not just an immediate “tier” where observations are “confirmed directly”. Thus it is indeed the case that: “justification never “terminates”, it gets distributed over the web”.

    This is one aspect where the “spider’s web” analogy is misleading. It’s better to think of it as a neural network with sensory feeds, since that is — quite literally — what a Quine-style web actually is.

    [Aside, one shouldn’t be wedded to Quine’s exact wording about the web, especially since the Vienna Circle originated it, not Quine.]

    Hi Joe,

    It seems both sets are infinite.

    True, but one set can still be much bigger than the other (different-sized infinities being a weird but sound concept).


  22. Coel:

    You gave no reasons why ‘grue’ is different from ‘green’ other than that ‘grue’ is merely a label and that ‘green’ describes something *in* the object. Not only is this fallacious, in general, as a distinction, but it clearly is a form of essentialism, your assertions to the contrary notwithstanding.

    It amuses me that you continue to misrepresent Quine’s views, apparently completely oblivious that you are doing so, and in spite of the fact that experts tell you you’ve got him wrong. The fact is, if you had *any* real familiarity with his work or with the scholarship on it, you would know that Quine assigns a distinct status to “observation sentences.” From his essay, “Epistemology Naturalized”:

    “A sentence is an observation sentence, if all verdicts on it depend on present sensory stimulation and on no stored information beyond what goes into understanding the sentence.”

    And from the SEP:

    4.2 Observation Sentences

    Central to Quine’s naturalistic account of knowledge is the idea that all our knowledge is in some way based upon stimulations of our sensory nerves. For much of our knowledge, the relation is quite indirect. (This is one way of expressing holism; see 3.1, above.) Most sentences are not accepted because of a direct relation between the given sentence and stimulations of nerve endings; the connection goes via other sentences, and may be quite indirect and remote. But then there must presumably be some sentences which are directly related to stimulations. This is the role that observation sentences play in Quine’s thought. Acts of uttering such sentences, or of assenting to them when they are uttered by others, are shared responses to stimulation.

    Indeed, this is why Jonathan Dancy, in his “Introduction to Contemporary Epistemology,” says that Quine is not really an epistemic holist.

    Of course, you give yourself an escape route, in your parenthetical at the end. So, even though you originally spoke of Quine’s web of belief, apparently, now, you aren’t talking about Quine, but rather Shmine, who is just like Quine, except that he doesn’t treat observation sentences differently from the rest of the statements that make up the web of belief.

    Of course, this won’t do any good either, because there are good reasons for not subjecting all justification to the web-metaphor. As I already mentioned — more than once — without *some* beliefs that are justified observationally, one could not distinguish an internally coherent fiction from a genuine belief system. In other words, coherentist epistemologies have trouble with truth.


    I will make the same suggestion to you that I made to everyone else: reading the material you are purporting to talk and opine about might be a good idea. It certainly would mean a lot less repetition over and over again of basic points that, with the internet, can easily be learned.

    Liked by 3 people

  23. Hi, I wish I’d had time to take part in this discussion, particularly DM and Coel‘s part regarding probability theory and Quine.

    In short, I don’t believe they overcame Hume’s arguments, and have mistaken a large sample number for adequate data to claim justification for the process itself. So I found Francislb‘s replies particularly interesting and useful.

    I happen to like Quine-style web of knowledge systems, and think Coel is incorrect that Hume’s concerns dissolve within them. Such epistemologies really do involve assumptions which when acknowledged and accepted allow for knowledge claims to be “justified” (internal to that system), though the assumptions cannot be justified. In other words, knowledge comes with a caveat. If one is ok with saying “I know” and not mean “I know that I know that I know that I know… (ad nauseum)” than that should be good enough.

    Looking for foundation-based knowledge sinks one into the I know ad nauseum trap, and I’m more than happy to leave that behind for something that grants practical, if caveated knowledge claims.

    That said, I think Aravis is being a bit unfair to state that Coel is misrepresenting Quine’s views when he explicitly stated you shouldn’t be wedded to them. Quine was not the only one with a web of knowledge scheme and there are other flavors. I believe he (and I) have previously described our epistemologies as “Quine-like” or “Quine-style”. Quine makes a good reference point to start but not end the discussion.

    While I think Aravis is right that eventually an unjustified assumption is required by any such system, an admittedly unjustified assumption does not constitute evidence it is a false one, and coherence and successful predictions aren’t exactly the best evidence against the assumption either.


  24. Shoot, I meant to end my last post with the zinger…

    Indeed, greater coherence and a long string of successes may never grant justification, they do grant a level of confidence.


  25. dbholmes:

    If you notice, in my last post, I accepted Coel’s escape-clause, re: Quine, and indicated why a no-tier, purely coherentist theory of justification is problematic. In other words, Quine is *better* than Shmine, because Shmine has no way of distinguishing an internally coherent fiction from an actual system of knowledge. Without some contact with “reality,” which is what observation sentences are, there is no way to tie the system of statements to the world.


  26. Hi ejwinner. You cpmmented
    “…in which case the whole theory – as a claim on universality – crumbles to dust.”, but the point is rather that if you never see a black raven, your credence for the “all ravens are black” hypothesis gets closer and closer to certainty (but doesn’t reach it except at the limit). The “universality” is that it behaves as well as it can given all the uncertainties – you might still be wrong about the true state of nature. Maybe we do live in a wacky world (a la “crazyism”, mysterianism etc).

    But as I see it, deduction can only ever exclude certain collections of axioms/postulates, you need induction or related processes to actually add to knowledge. You may even need induction to choose the “logical axioms” (in Table 1 of the ref), that is decide if a particular rule of logic (eg excluded middle) gives a system that matches the real world uses you want to put it to – I am generalizing wildly here from the few papers I have read on the nonclassical logics now popular for automated theorem proving.


  27. Great listen.

    I agree with (most) everything that both Dan and Massimo had to say. I think the most important thing to stress is that yes, the skeptical argument is indeed foolproof, but I think Dan was spot on by saying that it really can be “dissolved” in a sense.

    What DM, Coel, and several others might want to add to their philosophical arsenal is pragmatism, which will allow them to continue using induction, deduction, abduction, and probabilistic reasoning without worrying too much about whether or not they’re dreaming/hallucinating everything.

    Pragmatism seems to get short shrift around these parts, but there is a section I’ll cite directly from the Wiki entry on the philosophy that can help show how powerfully relevant it is to this topic:

    “Hilary Putnam has suggested that the reconciliation of anti-skepticism and fallibilism is the central goal of American pragmatism. Although all human knowledge is partial, with no ability to take a ‘God’s-eye-view,’ this does not necessitate a globalized skeptical attitude, a radical philosophical skepticism (as distinguished from that which is called scientific skepticism). Peirce insisted that (1) in reasoning, there is the presupposition, and at least the hope, that truth and the real are discoverable and would be discovered, sooner or later but still inevitably, by investigation taken far enough, and (2) contrary to Descartes’ famous and influential methodology in the Meditations on First Philosophy, doubt cannot be feigned or created by verbal fiat to motivate fruitful inquiry, and much less can philosophy begin in universal doubt.”

    This very idea allows us to establish things like deductive/inductive methods as completely valid when it comes to trying to understand the world. They’ve proven themselves time and time again, and until that ceases to be the case (extremely unlikely, but again, we can never know for sure), we can continue using them based on pragmatic grounds. Now of course I believe, as many others in the comment section (dare I say even Massimo or Dan?) probably do, that these methods are very likely actually valid. But like others have pointed out, we just cant be sure by using those very same methods to assess this, as it would be circular.

    One other point I wanted to make was with respect to foundationalism. Massimo seems to think that it’s failed as a valid epistemic theory, but I think that’s highly debatable. I myself am a fan of modest foundationalism, which plenty of individuals subscribe to as well. In this framework, there’s no need to have foundations that are absolutely infallible. All of our claims are fallible, though over time and exploration we continually refine our foundations. Take the mathematics, logic, and the natural sciences (you might also throw in law/political theory). Each one of these subjects has foundations, and each one provides us with knowledge. They all have foundational “axioms,” “laws,” or “generalities.” With this in mind, I really don’t see the other two options (infinitism, which strikes virtually everyone as absurd, and coherentism, which falls victim to an infinite number of other equally valid webs) as more attractive by any stretch of the imagination.

    Liked by 2 people

Comments are closed.