In conversation with Anthony O'Hear
*Induction (definition): “the doctrine of the primacy of repetitions” – Karl Popper
If you are going to be a non-hypocritical Popperian, then you are going to have to love your enemies – those people who go out of their way to kick holes in your philosophy. Looking at what he wrote attacking Karl Popper, Anthony O'Hear admits now that perhaps he wasn’t “generous enough” all those years ago, that his “book has a certain nit-picking quality” to it. So let’s be thankful that it does!
In many ways, Popper built his reputation and career against the back of the inductive method, stabbing wilfully into the flesh, weakening the giant to its knees, and then to its death; waiting for its enormous shadow to finally clear from the landscape of philosophy and logic. The fact that this hasn’t happened, is explainable firstly by the craft and intellect and rigor of modern inductivists like O'Hear, and secondly by Popperian scholars failing, almost entirely, to recognise the work that such people have done; instead, continuing to argue against a much cruder – and long since dead – inductivism from philosophical history.
Carrying around the words of Popper, and parroting them in back-slapping agreement, there are vastly too many Popperians out there in the world today who are great at remembering the prose, and yet terrible at living the philosophy. Terrible at admitting to the uncomfortable paradox that: if it is true, then it also must – in many different ways – be false! That these falsehoods need to be chased and hunted and celebrated by Popperians themselves. And that when we run out of ideas in this pursuit, criticism from without should be welcomed as medicine for a sick patient.
When O'Hear says – after his brief nit-picking admission – that, “On the other hand, I don’t think I would go back on the things I have said in the book in general about induction and verification”, he deserves our thanks, our admiration, and our love. He has, after all, written the most punching, and demanding, and difficult to evade assault on Popper’s work that exists today. And so, if we are to be true to ourselves and to our philosophy and to the man himself, it is also the best book on Popper’s work that exists today!
Let’s start at the beginning then, with the simplest argument for induction, the crude one: “Suppose, for example we are trying to discover the cause of cancer. We examine a large number of cases and notice that they are all heavy smokers, but that they seem to have nothing else in common. We would then naturally formulate a hypothesis to the effect that it was the smoking that caused the cancer.”
This type of claim might appear reasonably profound and uncontroversial, until you begin to think about all the possible things which could have been “naturally formulate[d]”. In this case it was smoking, but only because on the short list of things in common, it seemed to be the most likely culprit. There were infinitely many more things which all those cancer patients had in common, which were ignored or not noticed: their clothes, their language, where they live, the things they eat, the air they breathe,… However it came about that the inductivist researcher selected smoking, it couldn’t possibly have been from compiling lists of all the features of the patients’ lives, and then selecting the one feature which repeated the most (or they would still be working on those lists today).
Our impossible troubles with observing all aspects of an event or phenomenon, also stretch-out to the question of what constitutes a repetition: “a farmer may see his ploughing his field on two days a repeated task” writes O'Hear, but the “mouse whose nest is ploughed up on the second day will be more impressed by the distinguishing features of the two days’ activity.” The lesson being, every time we notice a repetition of some kind, it always involves the prior adoption of one or another points of view. This is O'Hear giving Popper his theory-laden dues.
Popper’s view on this is fairly straightforward: nothing that you could learn about experiencing Phenomenon A could help you to understand – or reason about – Phenomenon B, which you have not yet experienced. And when inductivists have tried to moor their theory to firmer ground, they generally haven’t helped themselves, by arguing in regressive circles: what justifies the inductive method? Past successes of the inductive method! What justifies those past successes? The successes before that, and so on.
What happens when the inductive method fails, and an inductivist is facing the disproof of his theory, even within that infinite regress? He runs a little further into the darkness, arguing only that the method or the principle has been misapplied, that relevant differences between connected events weren’t noticed, or that other relevant connections were overlooked. When the inductivist sees the sun rising and falling with regularity, he forms the theory that it rises and sets every 24 hours. He then visits Norway, sees the midnight sun and, instead of admitting that his method and his theory were wrong, he claims they were still correct, just missing a few extra details; details that he can now add after seeing them.
So it is a mistake to build a theory on the assumption that the future will resemble the past. But is it wrong to say that the accumulation of past evidence makes certain aspects of the future probable? For Popper it is, because a theory is either correct or incorrect – probably correct is a meaningless statement. And when you turn the wheels of probability theory on this, Popper appears vindicated. To run the experiment in a “universe such as ours”, the ratio would have to factor-in all possible (conceivable) tests and counter-theories. This being infinitely large, the probability of any one theory being correct would be zero (“or something very close to it”).
Regardless, in small and failed shifts like this, you can begin to see the steady refinement of the inductive method in response to Popperian criticism. And it continues. In fact the hard centre of O'Hear’s inductivism is a chase of sorts – Popper running slightly ahead, O'Hear snapping at his heels, getting closer by each step, until finally they turn down a bricked-up, blackened alley; nowhere else to run: “Popper’s attempt to dispense with induction is unsuccessful. We have found that inductive reasoning, removed from one part of the picture, crops up in another.”
All that talk of probability theory failing to help the inductivists out, is turned around on Popper, and turned around on the principle of falsifiability. If it is meaningless to say that a theory is probably correct based on some criterion, then it must also be meaningless to say that it is probably wrong! When a theory fails a test of some kind, at the bones of things this means that an empirical statement has clashed with an empirical observation. But – again back to that endlessly complex world of ours – there are infinitely many ways that this might happen, and not happen. So to talk about falsification without the help of probabilities, Popper and Popperians need a classification system, something that sizes different types of clashes (empirical statement vs. empirical observation), and designates what the consequences should be of each.
And O'Hear is happy to volunteer some of the preliminary work: “If the class of potential falsifiers of x include the classes of potential falsifiers of y and some more as well, x is more falsifiable than y.” For example, take these two statements and try to falsify them: 1. The planetary orbits are circular, 2. The planetary orbits are elliptical. The latter theory sits in a different class to the first, as it requires “six singular statements describing a curve” in order to falsify it, whereas the first theory requires only four.
O'Hear’s help and kindness quickly escapes him as we run deeper into things. Back in that infinite human world of possible experiences and observations and statements, theory selection is a fraught and difficult place for Popperian philosophy. How is it – from that limitless space within, and beyond, us – that we ever get around to choosing a single theory and calling it true? Millions upon millions of different theories are capable of explaining any single observation, the vast majority of which are not falsified and never will be. O'Hear and his inductivists have a simple answer: you choose the theory that has been most successful to date, the one whose predictions have most consistently come true.
The Popperian answer is the least impressive part of his philosophy: you choose the theory with the highest degree of corroboration! Meaning you choose the theory that is best corroborated by surviving the most severe tests. (At this stage you should be sensing the blood in the water! But let’s continue). Successful predictions, lots of them, are important things for a theory to have before we consider it to be true. And clearly we prefer theories that survive tests to theories which fail them. Now even if Popper says, something like all the successful corroborations in the world won’t give us reason to think that the corroborated theory will continued to be true into the future, will continue to be corroborated, he does still have a problem on his hands here.
Imagine you are a commander on a battlefield, and in the next few minutes you have one of two decisions to make. The enemy is attacking from an unknown side of the mountain range in front of you, so you need to either send your reserve troops to the left to reinforce the line, or to the right. With easier terrain, more cover, and better firing angles, all the attacks to date have come from the left flank. So the theory that the enemy prefers to attack from the left is a well corroborated theory. The inductivist agrees – completely! For them, the theory that the enemy prefers to attack from the left is true because it has been successfully predicted time and time again. The difference? The inductivist thinks that this should inform your decision about where to send your troops, whereas Popper thinks it shouldn’t.
Well what is the point of talking about corroboration, or truth in general, if it does not help to guide our decisions, if it does not help us with theory selection? Because, after all, you do need to make a selection in moments like the one just mentioned, despite the best efforts of some Popperians to pretend like you don’t – to say obfuscating things like, I would first have to more fully understand the ‘problem situation’. When induction does its best to meet Popper’s criticism head-on, the shame only belongs with the people playacting as if nothing were actually being said.
The question remains: as the commander on the battlefield, do you take into account the past behaviour of your enemy or not? Do you make your decisions with those patterns in mind, or do they hold zero value? In other words, back to that previous question: what is the point in corroboration if it’s only worth is to retrodict an explanation for past events, and not to predict future ones? And if it does have some predictive value for the future – if those previously observed repetitions should be factored into your decision making process in some way – then corroboration begins to sound a lot like induction.
It is important to head things off here, and remind any angry Popperians out there just who it is they are talking with. It’s been said once, but it’s worth echoing: O'Hear is not some unsophisticated lout from an earlier time, shouting about the inductive method through drunken breath at passers-by. Because the enemy has always attacked from the left does not mean that they will always attack from the left in the future (they might choose to attack from the right to catch you off guard, or the weather might change making the right flank easier to traverse), but only that this past behaviour should be a very important part of your calculations.
So not a hard inductive theory, nor the primacy of repetitions; but a theory of inductive inferences, and the importance of repetitions.
Either way, it is verisimilitude to the rescue – to save Popper’s corroborations from the “whiffs of induction’ and from that question: what does it mean to say that something is true? And what does it mean to select one unfalsified theory over another unfalsified theory? To give all of this the sense of purpose and progress that our “intuitive desire” needs, verisimilitude is a way of speaking about truth in terms of our distance from it. Theory A explains a phenomenon, and so does Theory B. Neither are falsified, both produce accurate predictions, but if Theory B explains the phenomenon as well as others, and/or has more precision, we can reject Theory A not because it is false, but because it is less true (has less truth content). In this way we can see the verisimilitude of any theory or statement being “its truth content minus its falsity content.”
So we can now make Popperian sense of the act of comparing theories, and of theory selection. But not quite! If you want to appraise the verisimilitude of a theory, then that appraisal will have to rely upon how we view the tests it has passed – it will have to rely upon “inductive background assumptions”. Call it “background knowledge” if you like, but the problem remains – the outcomes of all those tests matter – and if they matter then you have a spoon in the inductive soup. Verisimilitude 2.0 drills harder into the two categories that a theory can hold (truth or falsity) and talks instead of “excess truth-content” and “excess falsity-content”. 3.0 involves deriving numerical values from the accurate predictions of theories. 4.0 is a “language-dependent” version, taking apart the propositional language and primitive sentences.
Collapsing back into the uncertainties of probability theory, into inaccuracy, and into induction, whichever way verisimilitude has been constructed over the years, it has always failed to “fulfil all Popper’s original desiderata.” Adding another layer of problems without solving any, Popper stuck with the theory as it chased its own tail, “continuing to stress the importance of the idea” and trying desperately to save it; behaving in a very un-Popperian way!
So much of this comes back to testing, severe testing. This involves not just trying to shoot down a given theory, but doing so in places where that theory appears weakest and most likely to break: the riskiest predictions, the most unlikely consequences, and the most probable types of counterexamples; tilting the scale – as much as possible – towards falsification. The trouble is, once you have completed a severe test, it becomes less severe the second time, less so the third, and so on. You are repeating the event, and so the risks of falsification diminish. As the tests come and go, the theory in question moves further and further into established fact (background knowledge).
So in straightforward language. How does a theory move from risky theory to background knowledge: the repetition of severe tests… or induction! If a good Popperian doesn’t want to use corroboration as a guide to future success, but does want to claim that well-corroborated theories should become a part of our background knowledge (things that we take for granted in order to test other things), then the inductivist will nod away, saying: at least you have admitted that background knowledge is “covertly inductive”, and that inductive reasoning has its place. Still doubtful? How else, if not inductively, does the reproduction of a test and its outcome make it increasingly less severe?
If not a whiff of induction, then how about a whiff of verificationism. Popper’s philosophy is always running away from direct observations and towards theoretical statements. It goes like this: the statement “this is red” might appear as a clear observation of some object in reality, but by making such an assertion – no matter how clearly red the object in question is – you are impractically committing yourself to that truth holding into the future; for an infinite number of statements and objects. All of which we are not in a position to verify. Will it always look red, under all conditions, from all angles, and with all the coming advances in technology? This scepticism bleeds from Popper’s “general feeling that we can never rule out every possible mistake”.
It is interesting to think in this way, and it is certainly true, but – as O'Hear points out – “this position has no practical import. We do not [and cannot] act as if we might at any time have to revise well-tested empirical judgements about everyday realities”. Popper is making two errors here: 1. Suggesting that there is always evidence for a given observation, 2. Missing the common-sense way in which we talk about things, and why their enduring nature (even if false) is necessary for the “rest of our conceptual scheme” to hold. With Popper speaking in this way, it is hard to imagine how any empirical statement could hold any meaning at all.
If you are sensing some Wittgensteinian tones to this debate, then so is O'Hear, and so is Paul Feyerabend, and so is W.W. Bartley: “Bartley’s innocent comparison of Popperian methodology with a Wittgensteinian language game possibly so enraged Popper because of its closeness to the truth.” And he is running fast through the open door of another house that he doesn’t want to be in.
The scientific realism that Popper defends, is the claim that our theories actually give us knowledge of the world, as opposed to scientific instrumentalism which claims that our theories are just instruments from which we can get accurate predictions (just as a screwdriver doesn’t need to mirror the screw, our theories don’t mirror the world, but simply help us to bridge the gaps to what we want – they are tools, nothing more). Popper instead values greater universality and greater depth of explanations. He believes that – by encouraging the endless probing of even more fundamental truths – he can steer his philosophy into anti-instrumentalist seas. At this point in things, and surprised by the weakness of Popper’s argument, O'Hear has lost some of his patience: “an instrumentalist could agree that they are desirable properties [universality and depth] of a tool, because the more applications a tool has, the more useful it is.”
O'Hear also senses a whiff of – unavoidable – relativism in Popper, but this is best left for another day, and for readers of his book. It can all be brought back to a simple claim about what is reasonable and unreasonable to believe. Might it not be the case – and do we not have appropriately good reasons to believe – that it is just “our biological good fortune” to be able to notice regularities in the world? And to then be able to use those regularities to make successful predictions? It would, after all, be impossible to think and to function and to survive if those regularities were to suddenly disappear tomorrow (you became unable to notice them). And the best arguments for the method of induction come back to just that, practical, common-sense, decision making.
Piecing through Popper’s work with a maliciously sharp blade, O'Hear eventually finds his way to close agreement; or rather he believes that Popper agrees with him, and not the other way around. Especially when it comes to his claim about physical regularities and their importance for knowledge creation (“at least insomuch as this is acquired by ‘physical methods’”). Saying that our assumptions about the world could always fail us, that we could always be wrong, doesn’t hurt the O'Hearian inductivist in any way.
It is impossible to walk away from this book and this man, without thinking that he has a point; a point that scholars of Karl Popper are almost wilfully missing. Other than with his argument about Wittgensteinian language games (which now seems uncontroversially true), there are some good answers to many of O’Hear’s challenges (which he acknowledges himself); but there is no conceivable way of not being a better critical rationalist after reading the work of Anthony O'Hear – the worst enemy, and greatest friend, of Popperian thought.
**** The Popperian Podcast #14 – Anthony O'Hear – ‘Whiffs of Induction’ The Popperian Podcast: The Popperian Podcast #14 – Anthony O'Hear – ‘Whiffs of Induction’ (libsyn.com)