The view from the top

In my last post, I tried to describe what is for me a very familiar pattern of intellectual struggle, followed by what feels like understanding, followed by a zealous over application, and then, finally, a more modest re-assessment of the new idea. My best personal example of this phenomenon came with my introduction to the ideas of the philosopher Karl Popper, particularly with regard to the scientific method.

I first met Popper’s work in summary form, in a book called The Great Philosophers, and clearly the professor doing the summarising was a fan. I was hooked both, because the description given was compelling, and also because the central topic, the special nature of scientific enquiry, suited my interests and prejudices. I felt as if I was on the verge of being delivered up a robust intellectual foundation for the things I most wanted to believe.

Later I spent a year at a genetic research centre where Popper’s take on science was central to the ethos of that community. Indeed the two principle academics there had risen to fame many years before by specifically meeting a Popperian challenge to show evolution constituted real science. And that centre was such an exciting and admirable point of intellectual endeavour that my fondness for Popper was always going to blossom.

So I read. I worked my way through The Open Society, The Logic of Scientific Discovery, Conjectures and Refutations, other books whose names I’ve since forgotten. And although the writing style was deceptively accessible, it took work for the ideas to settle, and it was precisely the effort required, plus the desire to believe, that saw me overlook the most obvious weakness. This will make a little more sense if I outline the theory in question.

Popper’s description of the scientific method claimed to be addressing two puzzles, the problem of demarcation (what makes science different from other forms of enquiry?) and the problem of induction (how is it that we are justified in assuming the things we have observed today will continue to hold tomorrow?) In essence, he proposed a single solution to them both. And that solution was falsification. Science is special, he suggested, because it proceeds not by finding support for its theories, but by finding refutations. A theory is scientific if it makes a prediction which, if it doesn’t hold, will show the theory to be wrong. Hence, it is the business of science to generate brave and novel predictions. A new theory makes a prediction not contained in the existing theory, the prediction is tested, and either it holds, so the old theory has been falsified and must be rejected, or it doesn’t hold, and so the new theory has failed to establish itself and can be safely ignored. Popper’s favourite examples included the ability of General Relativity to predict the light-bending properties of mass (tested by measuring the apparent shift of a star’s position when its light passes close to the sun, for which one must wait for an eclipse).

Through falsification, demarcation is made apparently simple. Is this the sort of theory that generates tightly measurable and novel predictions? If yes, science, if not, it’s some other form of knowledge. Popper liked to use examples like Marxism, which by his assessment are not scientific because they give their explanations after the event, twisting the narrative to support the hypothesis.

And induction? Well, Popper suggested science doesn’t use inductive reasoning. Which is to say, we don’t really see a particular relation holding, and then conclude that it will continue to hold into the future, even though it often feels as if this is what we are doing. Rather, all we are saying is that a theory that has not yet been falsified might be true, whereas one that has been falsified is certainly not true (see one black swan, and the hypothesis that all swans are white is no longer viable). And so, by slowly but surely working our way through certainly false propositions, and replacing them with possibly true ones, we are refining our understanding of the world without ever committing to saying anything about whether or not our current propositions will hold into the future.

And this feels like such a satisfying theory. Science is special, because it has a rigour about it that other endeavours do not have. That is not science! one can confidently claim, it is nothing but pseudo-science. Where are its predictions? How can it even be falsified? Although it is never made explicit, the assumption that non-scientific enquiry is in some way inferior is a natural fit with the theory, and the scientifically minded may feel a certain smugness is licensed. And, because the requirements for falsification can be made tremendously narrow, just about any form of knowledge you have reservations about can be dismissed as not real science.

Further, for anybody who has ever worried over the problem of induction (okay, I admit, we’re a small and odd minority) the thought that one can advance knowledge without having to make what is essentially a leap of faith to sustain it, is more than a little appealing. Clearly, in opening up lines of division between those of faith, and those of science, it’s a particularly beguiling offer.

And yes, I was smitten: along for the ride, ready to hurl my insults at those left standing on the roadside . The wheels first wobbled when I tried to describe Popper’s contribution to scientific method in my book Falling for Science. I found that I was forced to use the same few examples again and again, because whenever I tried to apply it to other areas of science I was familiar with, the facts never seemed quite straightforward enough. There was always room for uncertainty and interpretation. As there is in non-scientific endeavour. The line of demarcation was worse than murky and whatever this thing called scientific knowledge was, it was going to be awfully small.

More importantly, even those examples (like the relativity one) appeared to carry their emotional weight not because of the falsification of a rival, but because of the verification of the new theory. Yes, they’re two sides of the same coin, but it’s the power of verification that really impresses us. As, I propose, it should. Because it is verification that gives us the confidence to apply our theories. When we did A and observed B, again and again. Hence we are able to trust this relation, and build technology and future theories upon it. The belief that this sort of relation holds Falsification is important too, one should always be open to negative evidence, but a huge part of the scientific enterprise is interested in verifying what is suspected, rather than attempting to falsify it. Why on earth deny that?

The answer is, because if we are able to frame scientific enquiry purely in terms of falsification, we don’t have to worry about this business of the rules all changing on us tomorrow (this is the problem of induction: just because rocks don’t float today doesn’t imply they won’t float tomorrow, unless you assume that all of our physical laws will extend into the future, and there is no evidence available to support this view, so we do just have to assume).

But even if the rules might change, so the Popperian reasoning goes, we are still confronted with one theory that has already been falsified, and one that hasn’t, and hence the non-falsified is in some sense a better bet. Unfortunately, as soon as we attempt to apply this bet to the future, the wheels come off completely, and this is the gaping hole in the Popperian description of science. We can hypothesis A is a better description than hypothesis B of the world we have experienced, but we can’t predict that it will out perform it tomorrow. If Popper’s description of science is correct, then science is entirely unable to tell us what to expect, ever. Or, as one critic rather wearily put it, ‘when are the Popperians going to admit, that every time they turn the light switch and expect the light to shine, they are contradicting their own theory?’ Or, as someone else said, nothing that is correct about Popper’s description of science is novel, and everything that is novel is incorrect.

And, while Popper’s contribution was undoubtedly important, and gives us new and helpful ways of framing the problems, I now think the critics have a very good point. Yet, how easy was it for me to ignore this? Even when I came across a lecture on youtube, where the academic pointed out precisely this problem, using the decision to get on an airplane as his example, I somehow managed to miss his point. My ignorance had become all but wilful. The trouble, I think, is that learning why a theory doesn’t work is every bit as hard as learning how it does work, and nowhere near as satisfying. It is the difference between climbing the mountain, and returning to home base. All that extra effort, just to find you’re back where you started, only more tired. No wonder we get trapped in error, the views from the top are gorgeous.

23 thoughts on “The view from the top

  1. JP says:

    Hi Bernard,

    I have not read Popper and can’t comment directly on his ideas. However, on induction, it does not seem absurd at all to say that it’s not used in scientific theories.

    Take the theory of gravitation. It can be seen as an abstract system stating mathematical relationships between certain entities. As such, it has no place at all for induction (or perhaps even for causality). Seen this way, the theory is not about reality at all – only an abstract system.

    When comes the time to apply the theory to the real world, what we need to do, it seems to me, is to establish a coupling or mapping between the entities of the theory and real world objects, through well defined observations/experiments/measurements. Then, if the theory states that some mathematical relationship holds between entities X and Y, we can translate this into an experiment (using well defined rules) and see whether or not the equivalent relationship holds in the real-world version. We can perhaps distinguish between valid theories like relativity (with very strong coupling) and pseudo-science like astrology (no coupling at all).

    The question of whether or not the apple will still fall tomorrow seems to be of a different kind. Not sure what it is.

    Now, perhaps I am just stating the obvious or not making sense at all – just trying to think my way into these issues.

  2. Edwin says:

    If the argument that the fact something has always occurred a certain way cannot be considered evidence that it will continue to occur that way is untrue, then all and any predictions about the future are invalid. Since this is clearly incorrect, it seems obvious that one can make this logical conclusion. Therefore, I don’t understand Popperian reasoning. How could anyone subscribe to such an obviously false assertion?

  3. Hi Edwin

    I think the thing to notice is that Popper, like most philosophers, took Hume’s challenge to induction seriously. Which is to say, while intuitively it does indeed seem just obvious that making predictions about the future is valid, it is devilishly difficult (some say impossible) to justify this intuition.

    Popper then, was attempting to find a way around the logical impasse.


  4. Edwin says:

    So the problem was solving Hume’s challenge to induction, namely, that the prediction things will continue to act as they have in the past is circular logic because there can be no other justification for this belief other than that’s how it has always happened in the past? What then is your solution to this, now that you’ve rejected Popper?

  5. Hi Edwin

    Well many have tried, and failed, to find a solution, so I’m not about to propose a new one. Of the various takes on this, I like Bertrand Russell’s observation that maybe the best we can do is accept induction as an independent principle of logic and get on with it. After all, the only way we can establish deductive logic is via deduction (circular) or by appealing to it’s success (which is inductive reasoning). So perhaps we do just have to quit seeking independent justification for the way we think and accept our intellectual limitations.

    The evolutionary perspective to some extent backs up this view. Given selection is in essence a retrospective process (it has no foresight) then all out behavioural and intellectual instincts have been based upon past success, hence there’s something about inductive thinking that is inescapably hard wired.

    Of course, both these perspectives do highlight the fact that our notions of reality are, at least in some sense, make do constructions, that can never be fully justified. Personally, I’m okay with that, but others seem to find it deeply worrying.


  6. Edwin says:

    Personally, I think applying the mathematical principles of probability to this scenario works. What we are doing when predicting the future is basing it upon a sample size of what we know to have occurred in the past, and thus we can never say for certain what will occur, but we can say it is more likely the sun will rise tomorrow than not. Obviously we may not be correct, but this will in most scenarios, when we have a sufficiently representative sample size, prove to hold true.
    Justifying this, however, is admittedly impossible due to the reliance on circular logic. However there’s plenty of other things that we can’t prove either, such as that our senses speak the truth and we aren’t just a “brain in a vat”. Making certain assumptions is necessary for any and all science to be possible, I believe.

  7. JP says:

    Hi Edwin,

    I’m not sure induction is necessary to do science, as such. We need it when we want to claim that a scientific result applies universally, say, or will continue to obtain in the future (the apple will fall tomorrow). But this type of claim may be more of a “meta-claim” than a scientific one.

    I think also we can consider a scientific theory on its own (like the theory of relativity), in this case as a set of mathematic relationships between abstract entities – no induction needed at all. We can also study how a given theory relates to reality by using various kinds of criteria (experiments, etc.) Here, too, we don’t need to invoke induction, unless we want to claim that a particular method of verification is “better” than another, but this, too, is a meta-claim.

    What do you think?

  8. Edwin says:

    I think the induction problem is philosophy and not science. There is no reason why a scientist should concern himself with such matters, because it’s not a scientific problem, merely a philosophical one. So yes, induction is not necessary to do science, but the assumption that things which have held true in the past will continue to hold true in the future is necessary for science, because otherwise, for example, how can you expect the lights of your lab to turn on when you flick the light switch? Let alone get the same result each time under the same circumstances, which would make experimenting pointless if you expect to do the same thing and get different results.

  9. Factpinion says:

    Hi Bernard,

    I’m not quite sure how you read “Logic of Scientific Discovery” and “Conjectures and Refutations”, and got from that that Popper is essentially proposing radical skepticism (as suggested by the critic you quote, according to whom “Popperians” contradict themselves when they expect the light to come on when flicking a switch).

    Popper repeatedly writes that we always have theories that have held up well to attempts at falsification – and that there’s nothing suspect or “un-scientific” about that. The importance of not being too *dogmatic* in our expectations, and/or trying honestly and seriously to refute our theories, is when we’re in the arena in which we’re trying *expand our knowledge*. He certainly doesn’t mean every time we put an old shoe on, we ought to be surprised if it fits. (Already in the introduction to “Logic…” Popper makes clear his focus is the growth of knowledge.)

    Honest attempts at falsification and peer criticism are the ways our knowledge grows (i.e. when they’re applied towards new theories), because a theory that holds up to falsification seems justified in being held for the time being – provided we allow that one day it might not. But the falsification attempts *themselves* aren’t the main point – obviously the content of the theory is; only that we’re interested in seeing if it might be true. So, when you say that in B following A again and again comes our confidence in a theory, I don’t think Popper has any argument with you – so long as you’ve 1) tried reasonably to see if in some instances B won’t follow A (which would refute a theory “B will ALWAYS follow A”), and 2) don’t get so dogmatic about B following A that you assign it to a “fluke” or whatever. Popper’s main point is not to think by B following A you’ve *proven*, for all time, that this is so.

    In the case of a the flick of a light switch, after all, sometimes the light blows out and it doesn’t come on. And actually – we’re surprised! 🙂 We’re reminded that our unconscious assumption, “this light will come on – it always does” is, technically, false.

    But to say (and act as if) it will *usually* come on is fine – Popper explicitly says we do this all the time (as with the example he gives in “Logic…” of the bridge builder).

  10. Hi Factpinion

    Thanks for your thoughtful comments.

    I agree, it’s entirely possible to read in Popper the down to earth view of scientific progress that you outline, and he often goes to pains to endorse this approach. Nevertheless, he is also explicit in making the claim that his approach answers the problem of induction, and the method by which he does this directly implies the scepticism you note.

    Hence, when reading Popper, we appear to be faced with the choice of taking his pronouncements on induction seriously, in which case the scientific method he describes is rendered impotent (the light bulb example) or we ignore this aspect and focus more on the many broad brush descriptions of discovery that he offers. In the second case, however, we find that much, if not all, that is novel in Popper’s description of scientific discovery is lost, and we have something akin to the common sense description offered by Huxley a century earlier (itself an echo of a tradition thousands of years old).


    • Factpinion says:

      Hi Bernard,

      Thank you for your reply. If I understand you correctly, you’re presenting a dichotomy – either a solution of the “problem of induction” means abandoning everyday inductive reasoning wholesale, a-la radical skepticism, or in fact Popper doesn’t solve the problem and in essence is just offering a wordy version of some simple, common-sense advice – “be critical and self-critical in your work” – which isn’t new, or particularly revelatory.

      I think whether or not Popper’s is a solution to the “problem of induction” – let’s call it POI for short – depends on how you understand that problem, and therefore what solving it means. What you seem to suggest is that any “solution” to the POI would necessarily entail *never using* an induction-like process from then on, and so would by that very definition mean radical skepticism. Solving the POI necessarily means never using anything like it ever again.

      But I don’t think that’s what Popper is suggesting. For Popper, the POI is not so much the practical problem but first and foremost the *logical* problem that no amount of repetition can *prove* something to be true. This logical problem, of course, has a bearing on practical problems, but isn’t the same as them. An obvious example of the logical problematics of inductions would be testing an occurrence that happens only one in ten-thousand tries, and concluding, say on the basis of 10, or 20, or 30 tries, that we’ve *logically proven* in can’t happen. *This* is the POI that Popper’s looking at: that we don’t know – and can’t logically derive – how much something needs to be tested/repeated to know it’s the case.

      It’s to this problem that Popper offers what I think is a satisfactory *logical* solution – not a prescription for hands-on science that means there’s no point in conducting tests anymore. In doing so, I find his contribution to have been substantial, if today seeming alternately either self-evident or extremely counter-intuitive.

      Regarding the second option I outlined in the first paragraph, I think there’s a difference between merely saying “be critical and self-critical in your work” – because it’s “good”, or “important”, or “keeps you unbiased” or some handy common-sense reasoning – to giving a far deeper, wide-ranging, logical, epistemological and even metaphysical reasoning for doing so, the consequences of not thinking that way, etc. His is not a prescription for every level of practical scientific work, but a (dare I say “healthy”?) logical framework for interpreting our findings, considering problems, approaching seeming confirmations, etc.

  11. Hi Factpinion

    Thanks for engaging with this. I wouldn’t want to suggest one can’t get a lot that’s interesting and provocative from Popper. Rather I was interested to observe how easily I overlooked a central contradiction.

    The thing I missed, and I think Popper at times makes it rather too easy to miss it, was the full implication of the problem of induction. If we do not assume regularity, then it is not just that no amount of confirmation can establish the truth of a relationship, it is also that no amount of confirmation can even establish the likelihood of a relationship.

    Hence, the folk intuition that those regularities established in the past are the best bet for the future (our bridge builder’s intuition) are entirely dependent upon inductive reasoning. Popper’s attempt to show that we can prefer one hypothesis over another whilst sidestepping the need for induction thus sets up a radical scepticism that contradicts much of what he has to say about the way we reason. it is this contradiction that I initially overlooked.


  12. Burk says:

    Hi, Bernard-

    Sorry to take so long finding your blog.. Hope you are well. I was just reading Popper a few weeks ago myself, from some kind of collected writings. In the end, I was far more impressed by his social commentary, like on Marxism, than by his philosophy.

    The “problem of induction” strikes me, after reading all these comments, as something like the problem of trascendence that religious people pose for themselves (and often on our behalf) all the time. Why, oh why, aren’t we clairvoyant? Well, we just aren’t, so we can never answer questions of the future with 100% certainty.

    I tend to view science as dealing in models. We are always working with models of reality, however ill defined and implicit. While falsification isn’t the solution to the scientific method or to induction, it does get at an important point, which is that scientists spend their time trying to define, and then test their models in the most demanding ways they know how. If the model makes one prediction that differentiates it from competitor Y’s model, then that is the experiment that needs to be done. Whether the counts as “falsification” or as “proof” is perhaps somewhat subjective.

    I’d go with a consistency theory, where people synthesize all the evidence, pro and con, to come with the best model given the weight of the evidence (including probabilities that the future resembles the past). Only in rare and dramatic instances does this involve radical paradigm shifts, the need for the old believers to die off, etc. Most of the time, normal science is perfectly fluid enough to find, test, and propagate changes to its models as it goes along, building asymptotically more accurate models.

    Best wishes…

  13. Hi Burk

    Good to hear from you. Yes, I come and go on the importance of the problem of induction. A part of me is very sympathetic to the ‘shut up and test’ approach, much as you outline above. Sure, we can not use past evidence to draw conclusions about the future without assuming some sort of regularity, but so what? The fact is it’s worked pretty well in the past, and short of somebody providing a better method of refining our guesses, then it’s apparently the best hope we have of knowing anything. Maybe, at some point in the future, all the rules will indeed break down, but it’s hard to see how planning for such an event makes any practical sort of sense (or why we should still expect to be around to have to cope with it).

    Against that, there are days when our inability not only to be certain that the rules will hold into the future, but also to assign any meaningful probability to the chance that they will, strikes me as profoundly important.

    As you know well, a certain type of theist likes to make use of the inductive conundrum by claiming that if we are justified in trusting our instincts on this one, despite having no applicable evidence available, then why not extend this approach to other spheres of knowledge? Not being a believer myself, I feel compelled to have some sort of an answer to this question, but it’s something of a work on at the moment.


    • Burk says:

      I think it is overly dramatic to say that no meaningful probability can be assigned. Perhaps these expectations were most markedly raised by the Newtonian revolution. Now, we could predict the future in amazing detail and accuracy (no instincts involved, incidentally). And Hume said… but there are no guarantees. The probabilities are exceedingly high, (which seems meaningful, if qualitative), even if the future has unknown aspects to it.

      I think the religiously and metaphysically inclined are assuming that by logic alone, we can give ourselves guarantees about the future. And they pine for that certainty, either by logic or by brute faith and authority. For the “view of god”, in all its omniscience. Science has provided alot, and engendered such expectations, but can never completely fulfill them. Another way to pine for all this is to fantasize about time machines.. another way to *transcend* the arrow of time.

  14. JP says:

    Hi Bernard,

    I understand what you’re saying when you compare belief in induction with other beliefs (religious) – we need some way to distinguish between these cases. It doesn’t seem we can do much better than say we deliberately decide to accept induction even in the absence of formal justification. But there are other ways to distinguish various kinds of beliefs.

    Consider the claim: “the sun will rise tomorrow”. Our confidence that the claim is true is based on induction but, independently of this, the claim stands on its own by being susceptible of verification: we can wait until tomorrow and see what happens. I like to say the claim is “strongly coupled” with experience. There is nothing similar with religious claims, especially now that they seem designed to preempt any attempt at verification.

    So, I think the key difference is procedural: we can apply some procedures (verification) to some claims (scientific) but not on others (religious) and we can decide to trust the ones but not the others.

    Now, to say we can wait and see or that this procedure has been followed many times in the past and the test succeeded does not mean that the claim is “true”. But, perhaps, this search for Truth is a red herring and we should look at this differently.

  15. Hi guys

    If I may play the devil’s advocate, perhaps.

    Burk, you’re right to say we can make extremely powerful predictions with high degrees of likelihood. The issue, I think, is that the very process of assigning probability relies upon an inductive leap, which is to say to believe an outcome is likely is to implicitly assume regularity continues. Hume doesn’t just say we can’t be certain the future will look like the past, he says we have no reasonable foundation for thinking any one prediction to be more likely than the other, no matter how well our models have performed in the past. This is why Bertrand Russell ended up saying inductive reasoning must simply be seen as an independent principle of logic, without which science would be impossible. At this level, perhaps it’s exactly like deductive reasoning, unable to be established without circular appeals (or an appeal to induction). So perhaps we ought to just accept it as one of our philosophical foundations and get on with it?

    JP, I’m not sure verification provides a useful demarcation. I suspect a theist would argue that many of their central beliefs (say life after death) are ultimately verifiable. Further, one might propose rather loopy, but entirely verifiable hypotheses (Mayan end of the world prophecy) but to my mind these represent different kinds of beliefs than ‘the sun will rise tomorrow’, despite having verifiability in common, in that the second leans much more heavily on the reliability of the model in the past. What do you think?


    • Burk says:

      Perhaps the issue is whether one aspires to make universal claims in a philosophical sense, which are equivalent to guarantees, as one gets from logic. That leads to the “problem” of induction. But if one instead has the expectation of making the best possible Bayesian inferences about the future, perhaps the problem dissolves. One does not need a new logical principle, if one is not pining for logically airtight (transcendent) claims.

      I do part with Hume if he claims that no one prediction is more likely than any other. That is empirically false. What might be true is that induction is not a (valid) logical operation, but a practical one. As Hume claims, induction is logically circular anyhow, so it is logically invalid in a fundamental way.

      So as you say, it is a matter of practical, intuitive, and natively bayesian intelligence. This hardly warrants all other uses of our intuitions to believe anything that falls into our heads. Induction by the method of bayes still has rules and structure. When based on rigorously observed experience, it is highly practical and accurate. When based on myth and tradition, not so much.

  16. JP says:

    Hi Bernard,

    I think we need to distinguish many related but different things. First, there is the question of what statements/claims are really about. My take is that we should take these as propositions within a model or some abstraction of a part of reality. As such, a model stands on its own – it has no necessary connection to reality at all.

    Now, and this is where verification comes in, different models have different degrees of coupling with the real world (or none at all). Again, the way a model relates or maps to the real world is more or less an objective matter. We can pretty well describe how the theory of gravitation maps to real objects by describing what experiments to perform, and so on. Likewise, the Mayan end of the world prophecy has a well defined verification procedure (obviously different from the one used with gravitation) – I would say the coupling is weaker. Likewise, possibly, with life after death.

    The point is that for each kind of model, we can identify objectively what kind of coupling is available between the model and reality. Possibly, no coupling at all is defined (as for example with some theological claims designed to preempt any empirical verification).

    Up to this point, it seems to me there can be agreement on what the models and corresponding couplings are. Either this particular verification procedure is available or not, and so on.

    It is only at this juncture that the notion of belief comes in: should we believe the propositions of this model in preference to another? What I think we do, in fact, is choose models according to which types of coupling we value most – and this is a deliberate choice. For example, I certainly prefer models that have a verification procedure that has been effective in the past – but this, I think, must be seen as a personal choice.

    No magic bullet here, of course. But I think that by distinguishing between models, coupling and conscious preferences, we may perhaps see more clearly what is actually going on.

  17. Hi JP

    Yes, I agree the difference between beliefs and preferences is very important. For my part, I like to think of it this way. When it comes to my preferences, I feel I have no right to impose these on others. I take a certain stance because it suits me to do so, and that is where it ends. When, however, I have a belief, I am prepared to extend that to others. So, for example, I believe walking out in front of a moving bus is dangerous, and not only do I refrain from doing so myself, but I would actively encourage others to do the same. In this way, my belief in the regularity of physical laws, for which verification has been successful in the past, is something more than a preference or personal choice (as I see it).

    Hence the acknowledgement that this belief can not itself be grounded in past evidence, nor indeed deduced from other accepted principles, feels important to me.


  18. Hi Burk

    Yes, I think the Bayesian procedure describes pretty well the way we use evidence to reason about likelihoods. And I also agree that it’s a sensible way of proceeding; in the face of any better alternatives it seems to me to be the way we should reason.

    Nevertheless, and this is a subtle point that perhaps am leaning on too heavily, the probabilistic reasoning at the heart of Bayesian inferences is reliant upon induction. We can use evidence from the past to generate expectations about the future only because we believe there will be no discontinuity, beyond which the observed relations no longer hold.

    It is this belief (I think some philosophers use the example of ‘Grue world’) that Hume challenges. He doesn’t say we shouldn’t have the belief, he just points out that it appears to be impossible to ground it, either rationally or empirically. We must take it as a given before we can proceed, even to the tentative point of claiming some outcomes are more likely than others.

    As such, we can’t ever say that we have empirical evidence that one prediction is more likely than another, without first making the metaphysical commitment that we live in a world of continuing regularities. And that commitment we do apparently just have to pull out of thin air, so to speak (or rather, perhaps, the retrospective nature of evolution has forced it upon us).

    Does this make any sense?


    • Burk says:

      Hi, Bernard-

      Absolutely.. evolution and every other aspect about living in the real world forces it upon us. So it seems more than a metaphysical principle, it seems downright physical, as a fundamental and necessary principle of existence.

      It does raise an interesting question for me about the boundaries between logic and metaphysics. I have not given much credit to metaphysics as much use on its own. But perhaps I should. In any case, this hardly licenses any sort of metaphysics. This case is one that is richly supported by empirical experience.

  19. JP says:

    Hi Bernard,

    Yes, when it comes to practical matters it seems we have no choice but assume some regularity in the world. In fact, this assumption is no doubt embedded in the deepest recesses of our minds and bodies – every form of memory, for example, is dependent on it.

    The problem, as you point out, is to justify this belief. I don’t think saying “out intuition tells us so” quite does it – something stronger seems at play and, furthermore, we know that not all intuitions are reliable.

    This is why I suggested to simply say we decide to use induction – without any justification whatsoever. Making it axiomatic, so to speak. (Note that this is not the same as saying we assume it is true.)

    Would it do to argue simply that this is the way our mind works? That this mechanism of anticipating the future from our past experience is so embedded within us we have no other choice? Seems a little weak but perhaps this is the way to look at this. This is of course related to the evolutionary argument: our brain (and brains in general) ended up working this way for a reason – because it works.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: