Episode 115: Is Falsification Falsifiable?

• Links to this episode: Spotify / Apple Podcasts

Unofficial, reviewed transcript. It may still contain mistakes; check the podcast when wording matters.

Speakers are labeled by first name where identified. Unidentified-speaker labels may refer to different people. Bracketed question marks indicate uncertain wording.

Transcript

[00:00:00]  Peter: Hello out there. This week on The Theory of Anything podcast, we consider, is falsification falsifiable? Was Popper a naive falsificationist? Is falsification itself a philosophical theory that thus makes it immune from falsification? Does the Duhem-Quine problem, or the assertion that theories exist in an interwoven web of other theories, create a problem for falsification? Bruce considers these questions and more as our infinite journey into epistemology continues.

[00:00:46]  Bruce: Welcome to The Theory of Anything podcast. Hey, Peter.

[00:00:49]  Peter: Hello, Bruce. How are you today?

[00:00:51]  Unidentified speaker: Good.

[00:00:52]  Bruce: Well, this is gonna be another epic length one.

[00:00:56]  Peter: Okay, that's okay.

[00:00:58]  Bruce: We're gonna talk about, is falsification falsifiable?

[00:01:02]  Unidentified speaker: All right.

[00:01:03]  Peter: Something I've wondered myself, and not sure I have a clear, concise answer to that. So I hope I'll come away from this with one.

[00:01:12]  Bruce: I'm gonna muddy the waters as much as possible.

[00:01:15]  Unidentified speaker: Okay.

[00:01:15]  Bruce: So I have an even less clear answer by the time you…

[00:01:17]  Unidentified speaker: Okay.

[00:01:18]  Unidentified speaker: Okay.

[00:01:18]  Peter: That's philosophy, isn't it?

[00:01:22]  Bruce: Let me try to explain Popper's epistemology in my own words, but with a very different emphasis that places the no ad hoc rule at the center of Popper's epistemology. So I did this in episodes 81 and 83, especially 83 called Popper's Second Axis, or aka Bruce's epistemology, question mark. Here's the elevator pitch version of it. If you want to seek the truth, intentionally make your theories very specific, such that they can clash with reality in some way, thereby making your theories as vulnerable, or rather non ad hoc, i.e. testable or checkable, as possible. Then expose those vulnerable theories to as much objective criticism as you can. Here, objective criticism means the type of criticism anyone, in principle, can go check for themselves, and thus has an intersubjective nature. If when you discover a problem with your theory, insist on correcting that problem only by increasing the level of vulnerability of your collective theories overall, though you aren't required to make any one specific theory more vulnerable and thus less ad hoc.

[00:02:34]  Peter: And this is the same as what you also call Popper's Ratchet, right?

[00:02:39]  Unidentified speaker: Yes.

[00:02:39]  Peter: If the theory becomes more testable like a ratchet kind of a thing.

[00:02:45]  Unidentified speaker: That's the idea.

[00:02:46]  Peter: That's how we move closer to truth in this world, basically, and rather than something more like an ad hoc save would be making it less testable.

[00:03:01]  Unidentified speaker: That's right.

[00:03:02]  Bruce: Now, there is one caveat, though. You can add new auxiliary theories to save your theory.

[00:03:09]  Unidentified speaker: Okay.

[00:03:09]  Bruce: So long as the new auxiliary theory you add is itself non-ad hoc, that is independently testable, testable beyond the problem that you're trying to solve. This also increases the overall checkable, testable content of the collection of your theories.

[00:03:29]  Unidentified speaker: Okay.

[00:03:30]  Bruce: So if you keep doing this, the theory that remains at the end will be difficult to defeat using objective criticisms. Or put another way, a true theory must of necessity be both highly vulnerable, be it logically or empirically, yet no one can seem to find an actual objective problem with the theory. That theory, that sole remaining theory, must of necessity be closer to the truth than its discarded competitors. It's rational to then tentatively adopt that theory as correct. So that's the elevator pitch version of how I understand Popper's epistemology.

[00:04:08]  Peter: And to be clear, this isn't just like a helpful rule of thumb or something. This is more like the center of

[00:04:16]  Unidentified speaker: [unclear overlap: how it actually works.]

[00:04:18]  Peter: How humans create knowledge in this universe, according to our best, what at least you and I perceive to be our best theory.

[00:04:28]  Bruce: Yes, there could be normative claims that come off of this. This is a description of how it works. Once you know how it works, you could come up with, and we're going to talk about this in a future podcast, you could come up with rules that make it easier to not fall into traps and mistakes, something along those lines, right? And those could be suggestions on how to go about doing things. But up to this point, we're just describing how it actually works, right? So this is also, by the way, a description of what it means to be rational. So now I added an additional claim on top of this, and many might even agree with what I just said, right? But this additional claim is where people tend to disagree with me. And this claim is that this epistemology that I just outlined, contrary to popular belief, works with both empirical and non-empirical theories. In the case of an empirical theory, the objective criticism might take the former character of, say, yeah, an empirical test. But in the case of, let's say, a mathematical theory, the objective criticism might take the character of a logical contradiction.

[00:05:42]  Unidentified speaker: Okay?

[00:05:43]  Bruce: Or it may also be a counter-example. A lot of mathematical theories can be checked by looking for counter-examples, or refutations or falsifications. I've been using the word checkable instead of testable throughout this, even though those concepts are pretty much the same. The difference is that, as we'll see, even a philosophical theory can be framed such that it has consequences that can be checked. Since Popper already used the term testable to explicitly mean empirically testable, I'm going to avoid using the term testable, and instead I'm going to use the term checkable. But checkability is just a generalization of testability. It's a form of testability that includes types of tests that aren't empirical tests, such as in the case of a mathematical theory. This might be, say, looking for logical contradictions or finding counter examples. So it's not really that much different. In fact, I would argue that it's almost exactly the same or maybe even the same. So, for example, the Church-Turing thesis may not be an empirical, empirically testable theory, but we know exactly what a counter example to it would look like. It would be the design for a computer that has a different computational class for algorithms, either by making a non-computable algorithm computable or by moving an algorithm from an intractable class, like let's say NP-complete, to a tractable class, like let's say P. So, but checkability is identical to testability other than being a generalization that is more encompassing to include philosophical theories.

[00:07:27]  Bruce: It is admittedly a slightly vaguer concept, and that does lead to some problems, though far fewer problems than is generally supposed. And that's what I'm going to argue throughout this podcast. The amount of checkability of a theory is the amount of content a theory has, for exactly the same reason that the amount of testability of an empirical theory is the amount of content that theory has. So by implication, a theory with zero checkability has zero content.

[00:07:56]  Peter: Now, what if you have a theory and you go on X and you argue about it for an hour? Is that a form of checkability?

[00:08:06]  Unidentified speaker: No.

[00:08:06]  Unidentified speaker: No.

[00:08:07]  Unidentified speaker: Okay.

[00:08:08]  Bruce: I mean, it could be.

[00:08:08]  Peter: It's not that vague then. Okay.

[00:08:12]  Bruce: So, yeah, it could be. I need to probably give specific examples throughout this episode so that it makes more sense what I'm saying.

[00:08:19]  Unidentified speaker: Okay.

[00:08:20]  Bruce: So let me keep going. I will try to get to that question.

[00:08:24]  Unidentified speaker: Okay.

[00:08:26]  Bruce: So the reason this critical methodology works is because realism is a correct theory. That is to say, there is an external reality, independent of you and me, and it is logically consistent with itself. So a correct description of this reality or a correct theory of this reality, even if we're talking about mathematical reality, by the way, must of necessity have the character of at once being vulnerable, that is, it's easy to understand what a counterexample to it or a contradiction would look like, yet you won't be able to find any counterexamples or contradictions. What you do not want, at least you do not want it if you desire to find the truth anyhow, is a theory that can explain anything and cannot clash with reality and is thus not vulnerable or checkable. For a theory that can explain anything explains nothing. Now, every crit-rat I know would agree with everything I just said except for the part about applying it to philosophical theories as well. Yet I just pointed out that the Church-Turing thesis is not an empirical theory, yet it can be falsified by counterexample, just like an empirical theory can. You really have to stop and think about the examples I'm using. If you're trying to stay vague in your mind, yeah, you can escape what I'm saying. But look at the exact examples I'm using. Notice the precision with which I'm doing this, okay? And this is why it's examples like the Church-Turing thesis that drive me to realize there is something going on here, okay? And that there is such a thing as a theory that's not empirical that can be falsified.

[00:10:05]  Bruce: Now, it's not uncommon for people to want to declare their pet zero content theories a best theory or sole surviving theory by trying to put it into a contest with a more content-filled and, thus, more vulnerable theory. Now, when somebody does this miraculously, their pet theory always ends up the best theory, precisely because the competing testable or checkable theory always has real problems to deal with, whereas their content-less pet theory has no problems at all and can't even in principle have problems. This is the very mistake that I made when I was a boy and I was a creationist. I would debate evolutionists and I would easily, easily point out very real, large, huge gaping problems in evolution. Then I would declare that evolution was refuted, and creationism was the sole remaining option in my mind.

[00:10:59]  Unidentified speaker: Okay?

[00:11:00]  Bruce: This mistake, and it is absolutely a rational mistake, we're going to call it the creationist fallacy.

[00:11:07]  Unidentified speaker: Okay?

[00:11:08]  Bruce: It's therefore critical to understand that the above methodology that I just described, it always takes place as a contest between roughly equally vulnerable or equally content filled theories. Preferably ones making specific predictions at odds with each other. Now in episode 83, Popper's second axis, I argued that Popper's most valuable insight was the realization that you can measure theories in two ways. One, by how many implications it has that you can test or check, and two, by how well those tests and checks have failed to undermine the theory when you actually did them, corroboration versus [refutation?]. Now, folk epistemology, which is just my term for default human epistemology, not the more powerful Popperian scientific epistemology, has one of these two axes, but not the other. It has how well the theory has survived tests without first measuring how vulnerable the theory actually was in the first place, to such tests or checks. That's why, as a creationist, it always felt like I was beating the evolutionists in arguments, because they had a very vulnerable theory that had real problems to deal with. Creationism has zero problems to deal with because it has zero content.

[00:12:31]  Unidentified speaker: So it never could have problems.

[00:12:36]  Bruce: But what do you do if a theory is philosophical? This is really one of the biggest questions that comes to mind. And it's one of my biggest disagreements with the CritRat community. Maybe the single biggest disagreement that I have with the CritRat community is that they see philosophical theories as exception cases to the epistemology that I just outlined. Now, to be fair, their objection does in fact come from Popper himself, particularly in Conjecture and Refutation, Chapter 8. In that chapter, Popper, which we covered in detail in a past podcast, he says that when it comes to metaphysical theories, which he equates to non-empirical theories, that instead of trying to falsify them with an empirical test, we need to rely on criteria like, now this is a quote from Popper from page 269, does the theory solve the problem? Does it solve it better than other theories? Has it perhaps merely shifted the problem? Is the solution simple? Is it fruitful? Does it perhaps contradict other philosophical theories needed for solving other problems? Now these of course aren't bad questions. And for a carefully formulated theory, even a carefully formulated philosophical theory that is highly checkable, many of these questions might rightly be regarded as what I'm calling objective criticisms. But for a vaguely framed philosophical theory, these won't serve as a particularly helpful basis by which you can assess philosophical theories.

[00:14:07]  Bruce: In fact, many of these questions for a badly formulated philosophical theory become little more than just gut feelings or subjective opinions, what I'm calling subjective criticisms. So let's take an example. Free will. Does free will exist or not? Now, in episode 112, I pointed out that most people that do combat over this question are intentionally vague. They can argue all they want over which theory is more fruitful, whether free will exists or whether it doesn't exist, or does it solve problems better than the alternative? But at the end of the argument, it's really just a matter of subjective opinion. And I think people tend to be very comfortable with what I'm calling subjective criticisms. So I think here of Bayesians that dedicate maybe their whole life to AI-doomerism, despite that being a theory that has really no checkable or testable implications at all. Or I think of crit-rats doing the same with anarcho-capitalism, a theory that they themselves insist have no checkable or testable consequences. So indeed, people tend to be very comfortable with subjective criticisms even against whole empirical fields of science.

[00:15:19]  Unidentified speaker: We're talking about lay people here, of course.

[00:15:21]  Bruce: So for example, crit-rats, I've argued in past podcasts, will often dismiss whole fields of science based on vague criticisms like, well, that field is empiricism, or I don't believe in explanationist explanations, by which they almost always really mean that the explanation wasn't as deep as they are requiring, not that they would hold themselves to the same impossible standard when it's a theory that they like. Neither group of rationalists, Bayesians or crit-rats, feels much need to first be sure that their theories are at least as testable or checkable as the theories they are criticizing. When they do this, they are committing the creationist fallacy. And both groups are incredibly comfortable with criticizing theories that they don't like with arguments that really are kind of just a matter of opinion. When it comes to philosophical theories, I've argued that crit-rats utilize what I call the crit-rat loophole. The crit-rat loophole allows bad explanations to masquerade as good philosophical explanations using the following foolproof trick. Declare your theory to be philosophical. If someone points out that it can be refuted via testing, ad hoc save the theory. For it is always possible to ad hoc save any theory by introducing a newly wholly untestable auxiliary theory. If someone points out that you just violated Popper's no ad hoc rule, respond that you already said your theory was only philosophical. So thus, the no ad hoc rule does not apply to your theory. Then point out how your theory is fruitful and simple, and how it solves some arbitrarily selected problem.

[00:16:59]  Bruce: At a minimum, whatever theory you are against can itself be declared a moral problem, because everything can be declared a moral problem. And your theory is therefore fruitful by solving this moral problem. All the while, the falsifiable and vulnerable alternative theory can be maligned, perhaps even accurately, just like I did with evolution when I was a creationist, because good explanations have problems. And all you have to do now, since your theory has no problems, because it has no content, is you just need to point out the problems with the alternative theory. Now, literally any theory can be declared a best or sole surviving theory, using this simple formula. Or, to put another way, this whole line of argument is a low quality, or if you prefer, easy to vary, theory, because it can never clash with reality in any way. So can falsification be falsified?

[00:17:52]  Unidentified speaker: So let's talk about that.

[00:17:53]  Bruce: My argument is that even philosophical theories must follow the no ad hoc rule, and must, or at least if you care to make progress, follow the epistemology I just laid out. Now if I'm right, then the crit-rat loophole is an outright mistake. Now to show my point, let's use Popper's own epistemology, which is definitively a philosophical theory, as our example. We'll delve into this across several episodes, but for now we're going to focus on one specific thing. Is falsification itself falsifiable? Now John Horgan writes in his article The Paradox of Karl Popper, here's a quote from the actual interview he did with Karl Popper. He says, I decided to launch into my big question to Karl Popper. Is his falsification concept falsifiable?

[00:18:42]  Unidentified speaker: Popper glared at me.

[00:18:44]  Bruce: Then his expression softened and he placed his hand on mine. I don't want to hurt you, he said gently, but it is a silly question. Peering searchingly into my eyes, he asked if one of his critics had persuaded me to pose the question. Yes, I lied.

[00:18:58]  Unidentified speaker: Exactly, he said, looking pleased.

[00:19:02]  Bruce: The first thing you do in philosophy, says Karl Popper, in a philosophy seminar, when somebody proposes an idea, is you say, it doesn't satisfy its own criteria. It is one of the most idiotic criticisms one can imagine. His falsification concept, he said, is a criterion for distinguishing between empirical and non-empirical modes of knowledge. Falsification itself is decidedly unempirical. It belongs not to science but to philosophy, or meta-science, and it does not even apply to all of science. Now, based on this quote from this interview with Karl Popper, Popper clearly thought his theory of falsification wasn't falsifiable. So there you go. Popper has answered the question. Is his falsification concept falsifiable? Popper has answered, no, it is not. And the reason why is because it's a philosophical theory, not an empirical theory, and falsifiability is only something that applies to empirical theories. What I'm asking you, though, is was Popper actually correct or not? Let me ask a different question first to [soften?] your defenses on this, because I know I'm going against amazing amounts of defenses on this.

[00:20:15]  Unidentified speaker: Has induction been falsified?

[00:20:18]  Bruce: Every crit-rat I know claims induction has been refuted or falsified, and that Popper refuted induction ultimately by solving the problem of induction, by showing there was no problem of induction, because induction did not exist. But isn't induction a philosophical or meta-scientific theory too? How can it be that a philosophical theory was refuted or falsified, as all crit-rats claim is true of induction, that it's been refuted? And isn't that by definition impossible? Surely induction is also decidedly unempirical too, just like Popper said of his falsification, right?

[00:20:59]  Unidentified speaker: So we have a problem.

[00:21:00]  Bruce: It is apparently possible to falsify or refute, at least in some sense of those terms, a philosophical or metaphysical theory.

[00:21:07]  Unidentified speaker: At least in some cases.

[00:21:10]  Bruce: If induction could be falsified, and Popper's epistemology, his falsificationism, let's say, couldn't be, wouldn't that be a very big problem for Popper's falsificationism? In fact, wouldn't that be such a big problem that it might even be grounds for abandoning Popper's epistemology in favor of inductivism? Some crit-rats might try to play a word game here, and say philosophical theories can be refuted but not falsified. Whatever that means, since those obviously normally would be considered synonyms. Or I might here anticipate an objection to something like this. It's true we say induction was refuted or sometimes even falsified, but we don't literally mean that. Only empirical theories can literally be falsified. It's more like an analogy.

[00:21:59]  Unidentified speaker: Hmm, okay.

[00:22:01]  Bruce: So you're saying there is some sort of generalization of the concept of falsification that applies to philosophical theories.

[00:22:08]  Unidentified speaker: I'll accept that.

[00:22:09]  Bruce: What I'm asking is, what is that generalization of the concept of falsification, or refutation if you prefer, that apparently applies to both empirical and philosophical theories? And how do we recognize a philosophical refutation slash falsification? Is it a purely subjective thing? When we say induction is refuted, do we actually mean something like, in my personal subjective opinion, induction is refuted, but your mileage may vary? Or do we mean objectively induction is actually refuted, and that someone who says otherwise is mistaken? So is this the battle between [induction?] and critical rationalism or falsification? Is it more like arguing who's the better superhero, Superman or Batman? Or is it more like arguing which is the better theory, Newtonian physics or general relativity? Which of those two analogies comes closer in your opinion when we speak of induction being refuted? So I surely thought, I thought I was under the impression that when we said induction is refuted, it was pretty much identical to claiming Newtonian physics was refuted, that if someone disagreed, they were mistaken. If you, if in reality, arguing over which is better, induction or critical rationalism is more like picking a favorite superhero, then honestly, I think I just lost interest in critical rationalism. Another claim I've heard many times is something like this. You can't falsify or refute a philosophical theory, but you can criticize it. But there's a problem with this formulation. It's pretty vague what we mean by criticism. Practically anything can count as a criticism.

[00:23:49]  Bruce: And isn't it obvious that Popper's refutation of induction was a pretty hardcore objective criticism, very reminiscent, if not maybe even identical to a falsification, whereas many other kinds of criticism are not? For example, there's probably very real but ultimately subjective criticisms that can be leveled at either Batman or Superman. For instance, we might argue Batman is better than Superman because Superman is overpowered and that's boring. Or we might argue Superman is better than Batman because Batman's all dark and gloomy and isn't bright and hopeful. If you've ever heard comic book fans argue over stuff like this, you know they take subjective criticisms like this very, very seriously. And you will probably note that some of their criticisms make good sense to you. But ultimately, which is the better superhero really is a matter of taste. And this is what I mean by subjective criticisms. A subjective criticism is a criticism that isn't intersubjective. It's something that you may deeply feel in your heart, but there is no way to convey that feeling to someone else. You can explain the feeling or your reasons for the feeling or something along those lines, but if the other person isn't able to feel it as well, that's where the criticism ends. This is in contrast to an objective criticism, where the criticism in question is something more analogous to an empirical test, where everyone can do the experiment or look at the criticism for themselves and repeatedly get the same result. So think of the Church-Turing thesis example here again.

[00:25:26]  Bruce: Anyone can go try to design a computer that is more powerful than a Turing machine, and in principle, therefore, falsify the theory. They just aren't able to, right? Because just they don't happen to exist. But we all know exactly what it would be like if the Church-Turing theory—[thesis?] were false and what a falsification of it would look like, even though it's not an empirical theory. Subjective criticisms might be very real, and they might even have a real basis in how you picked your favorite pet theory to work on in the first place. But subjective criticisms aren't ultimately what we most care about in the realm of choosing a best or sole surviving theory, in the critical rationalist sense. Because no one other than yourself ultimately has access to that subjective feeling of boredom that you may very well possess for Superman and thus prefer Batman. Note that if someone else happens to possess the same feeling of boredom that you do toward Superman, that likely they will wholeheartedly agree with your criticism. And for them, the criticism will land. And they will maybe even mistakenly claim that Batman is objectively the better superhero. But there is no way to transfer those feelings to someone else that doesn't happen to have that feeling. And this is what I mean by subjective criticism and how it differs from an objective criticism. Let me make my point clear. If you are taking falsify as a term, the term falsify, to explicitly mean empirically falsify, as Popper usually does, then clearly philosophical theories tautologically can't be falsified.

[00:27:06]  Bruce: But is that what the term falsify normally means in normal use in the English language?

[00:27:12]  Unidentified speaker: No, it's not.

[00:27:14]  Bruce: I'm taking the stance that that is an anemic understanding of the term falsify or refute, at least in terms of how it gets used in regular language. Okay, not when we're getting really super specific about Popper's scientific epistemology. And I'm claiming that there is a completely valid generalization of the term falsify that can include philosophical theories such as induction. We just need to tease out better what such a quote unquote falsification would look like. Now, recall from episode 92, Popper on philosophical theories, that Popper, right there in chapter 8 of Conjecture and Refutation, gives a refutation of induction, and his refutation is not some mere attempt to subjectively argue that it is less fruitful than his epistemology, or something along those lines. Which, by the way, probably wouldn't even be true, because induction has been an incredibly fruitful theory, okay? So, no, Popper, what he does is he objectively shows that induction has a logical problem that it cannot possibly solve. That is to say, he shows induction has a logical contradiction that anyone can go check for themselves. So Popper outright falsifies induction in a way entirely comparable to how we falsify scientific theories empirically. No subjective feelings or opinions were required. Now, let me make a comparison to an argument that I know a lot of people buy, and that may be rightly so, who knows, right? But compare this to Sadia or Philip Goff or Lee Cronin or Pinker or Nagel or many very smart people that claim that qualia is a logical contradiction to computationalism.

[00:28:57]  Bruce: Yet, there is no clear logical contradiction here that would require to have a clear logical contradiction that would require a logical theory of qualia, a theory that's very precise as to what we mean, which does not exist yet, okay? What they really mean is that they subjectively feel like it's an impossible problem to solve. So this is really, even though it may be a very, very convincing argument to a lot of really, really smart people, it may even be a correct argument for all I know, it's really still a subjective criticism, not an objective logical criticism. This contrast between an objective criticism, such as a logical contradiction, and a deeply felt subjective criticism based on a feeling, in this case that qualia is something that's impossible to explain via computation, is a really good example of the objective-subjective criticism divide that I have in mind. I'm arguing that a good philosophical theory makes itself vulnerable to objective criticisms, not merely subjective ones. Which by the way, makes induction a good philosophical theory, albeit a false one.

[00:30:09]  Unidentified speaker: Or does it?

[00:30:11]  Bruce: The issue here is that induction, like all theories, can be easily vague-manned such that Popper's refutation is no longer valid. Vague-manning is my term for formulating a theory to be vague for the purpose of removing content from a theory so that you can immunize it from criticism. So hopefully you're seeing my actual point here. Popper could only refute one very specific version of induction that was formulated in such a way that it potentially could and ultimately did clash with reality. Popper could not refute a vague overall concept of induction because that was impossible, which is why inductivists still claim induction isn't refuted. Falsification requires specific details that have implications that may clash with reality. With this in mind, let's go back to my question of if Popper's falsification can be falsified. Let's first admit that there is a certain formulation of Popper's epistemology that cannot be falsified. It goes something like this. Popper's epistemology is that you are open to criticism and then you error-correct your theories based on that criticism and thereby you get closer to the truth. We'll call this the invite criticism correct errors version of critical rationalism for short. You hear this all the time as a formulation of Popper's epistemology. Now, I gather that for some people, they're just giving a quick overview of Popper's epistemology, in which case there is absolutely nothing wrong with what I just said.

[00:31:44]  Bruce: The invite criticism correct errors version of critical rationalism probably works very, very well as a brief quick overview of Popper's epistemology. But I also gather there are many in the CritRat community that actually see that formulation as the whole substance of what they think Popper's core epistemology is, as if Popper's big discovery was that we should seek criticism. I mean, surely that existed before Popper, didn't it? But notice the inherent problem with this formulation. It's very, very vague. We might want to ask questions like this. Wait, what if somebody goes around seeking criticism, but debating by debating and arguing with everybody, but they mostly end up relying on insults? Though perhaps the person sincerely believes those insults when they happen are warranted because they think the other person was being scoffing or intentionally dense. Or what if the person doesn't even bother to take the time to understand the alternative theories that he or she is criticizing? But perhaps they sincerely believe they are. But when asked to summarize the other viewpoint, they always make an excuse why there isn't time or how it would be pointless. Or what if they are actually seeking criticism, but via confirmation bias? What I have in mind here is that they seek the best criticisms of their theory. Let's say Mises in economics as an example, by say reading Mises, who lays out in detail potential arguments against anarcho-capitalism and then carefully refutes each one.

[00:33:18]  Bruce: But let's say this person never bothers to actually engage actual criticisms from actual competing views. Is that person consistent with Popper's epistemology? Maybe they sincerely don't know of better potential counter-arguments. They tried to find some by reading Mises, but perhaps... So they honestly think they went out and found the best criticisms and then debunked them all, right? So maybe they really and truly believe that they've done their best to look for the best criticisms, but in reality, they didn't really bother to look, right? They were really just reading about Mises' own views of his own theory. This practically forces, the questions I've just asked above, practically forces a crit-rat to take a stance something like, well, what we mean is that you have to really seek criticism, which just leads to the obvious question, what do you mean by really? So hopefully you're starting to see the problem. If Popper's epistemology or his falsification really does boil down to nothing more than seeking out criticism and error-correcting your errors, then literally everyone qualifies, at least in their own minds, as sincerely doing Popper's epistemology. Every religious person you've ever met qualifies. Every suicide bomber qualifies. Every troll on the internet qualifies. This then calls into the question the actual value of Popper's epistemology. What's the point of Popper's epistemology if everyone claims they are doing it?

[00:34:49]  Bruce: And that, in fact, everyone for the past 200,000 years plus claims they were doing it inside of every single static society that has ever existed.

[00:34:59]  Peter: It seems to me that there's kind of like two ways that critical rationalism connects with people. And, you know, there's people like me who are drawn to it, who kind of like it is sort of life advice that makes sense.

[00:35:18]  Unidentified speaker: Yeah.

[00:35:19]  Peter: And, you know, and it does. I mean, it's not like these critical rationalists' assertions, I may be wrong, you may be right, but together through an effort, we get nearer to the truth. Kind of ideas don't make a lot of good sense. I think you would say that too. But, you know, there's probably a minority of people even within critical rationalism who are more like you, who want to get it down to like a machine learning level. I want to program it kind of a thing, which obviously is a very valuable enterprise too. But maybe these two groups kind of talk past each other a little bit sometimes.

[00:36:07]  Unidentified speaker: Yeah.

[00:36:08]  Bruce: You know, you're making a good point. And let me actually emphasize this point because I feel like it's really important, okay? I am not claiming that this invite criticism correct error formulation is wrong. So let me read it again. Popper's epistemology is that you are open to criticism and then you are, then you error correct your theories based on that criticism and thereby you get closer to the truth. That statement, without a doubt, is a completely true theory. The problem is, is that it's not formulated in a high-quality way that can clash with reality. In other words, it's too vague. This is what I mean by vague manning a theory. No matter what the truth turns out to be, once we get critical rationalism down to the level of machine learning and we've got an AGI algorithm, there will almost assuredly be some way we can map the final detailed precise truth to that very vague statement, right? Thereby making it a true statement. This is one of the main reasons people often prefer low quality theories over high quality ones. Low quality theories are far more likely to be correct. This is something I think people have missed, right? Is I think they want to see bad explanations, low quality explanations as wrong. But in fact, they're actually more likely to be correct, okay, than high quality theories precisely because they can be easily fit to later truths discovered. So, human beings tend to love like, quote, low quality formulations of theories precisely because they can't be refuted.

[00:37:37]  Bruce: And low quality theories tend to feel true because they are often, in some limited sense, actually true, okay? So, they feel true because they are true. So, people love to quote Popper saying this from All Life is Problem Solving. A rationalist is simply someone for whom it is more important to learn than to be proved right, someone who is willing to learn from others, not by simply taking over another's opinion, but by gladly allowing others to criticize his ideas. So, again, this is a clearly true statement. But, like, what does it really mean, okay? Now, I've never in my life met someone, no matter how irrational, dogmatic or foolish, that didn't sincerely believe, sincerely sincerely believe, that they were, quote, willing to learn from others or, quote, gladly allowed others to criticize their ideas. And everyone I've met claims that they agree it is more important to learn than to be proved right. So, I mean, literally, everyone I've ever met claims that. When I saw this one in a past podcast, you said, what about somebody who's, like, intentionally being violent? Okay, I'm talking about people I've actually met. Yes, there are counter examples to this, right? But in terms of every person you've had a discussion with, and it's at the level of discussion, they're all going to claim that this is what they're trying to be.

[00:39:01]  Unidentified speaker: Okay.

[00:39:03]  Bruce: So, a few years back, I had a long chat.

[00:39:06]  Unidentified speaker: This is kind of a funny story.

[00:39:07]  Bruce: I had a long chat with a member of my church congregation, and they went on for over an hour to me about how the earth was actually hollowed out and full of water. And he knew this because he had read some, quote unquote, science book that he liked. He kept telling me, I mean, for like an hour, he went on saying stuff like this, how other people just aren't critical enough of scientific theories and they're so dogmatic about them. And if they just looked at the evidence, the evidence that he read in this book, they'd realize how wrong modern science is about the nature of the planet. And he wished everyone was more like him. Clear thinking, carefully critical, letting the evidence guide him instead of just dogmatically accepting what we've been told by scientific authorities. It took me all of ten seconds to Google the book and to find a complete debunking of it. The quote unquote science book turned out to be some non-scientist in his basement that was doing experiments in his basement that had read the Bible and worked up a quote unquote theory based on his personal interpretation of what he read in the Bible and then worked up in his basement a few experiments that he felt favored his theory. The site debunking him was by the way done by a Mormon professor of geology at BYU. So these guys, either the one in my church that read the book or the guy who wrote the book, are they rationalists or not? They are at least in their own minds. They are completely following Popper's definition of what it means to be a rationalist.

[00:40:41]  Bruce: So I mean like the guy who wrote the book, he actually took his theories to a BYU professor that was a geologist and asked for feedback, right? Now by which I really mean he tried to convince that professor that geology, the entire field had it all wrong. Then when the professor tried to explain to him and tried to explain the counter evidence and the counter explanations, he didn't understand them, he wasn't interested in them and that's where things got left.

[00:41:10]  Peter: Can I say one thing about that?

[00:41:12]  Unidentified speaker: Yes.

[00:41:12]  Peter: I just had to ChatGPT it. And interestingly, even with that, there's a spectrum, it seems what I get from this of theories. There's the primary water theory, which is not mainstream science, but then seems like it has some kind of science-like validity from what it says. There's the water world theory, which is that the earth was used to be covered in water, debated in science, I guess. And then there's hydroplate theory, which is creationist pseudoscience, where the earth has to do with Noah's flood, is the earth, there's a vast reservoir of water.

[00:41:55]  Bruce: That comes up when it floods the earth, right?

[00:41:58]  Peter: It sounds like maybe that's more like what your friend is advocating, hydroplate theory.

[00:42:05]  Unidentified speaker: Yes, yes.

[00:42:07]  Bruce: So it's tempting here to say, well, by rationalist, I mean, you have to really seek criticism. But this guy really thinks he's seeking criticism. In fact, if you were to hear his side of the story from, I'm hearing the side of the story that comes from the BYU professor. If you were to hear the guy who wrote the book side of the story, I suspect that what he would say is that he tried to carefully lay out evidence that this BYU professor became combative over, and he wouldn't even hear him out.

[00:42:35]  Unidentified speaker: I'm sure.

[00:42:36]  Bruce: I do not doubt that's exactly what he would say.

[00:42:39]  Unidentified speaker: Okay.

[00:42:40]  Bruce: So who gets to decide if he did or didn't seek criticism? And by what means can we know? So what we want to formulate is Popper's epistemology such that it can clash with reality, and we can tell if someone really is or is not following his critical methodology, or at least have a better chance of figuring it out. I'm sure it's always somewhat subjective.

[00:43:04]  Unidentified speaker: I'm not denying that.

[00:43:06]  Bruce: So instead of saying merely, we seek criticism and then correct our theories, we should get more specific on what we're actually saying should happen. Sure, you should criticize your theories as much as possible. Sure, you should correct your theories. But what does that even mean? So let's try to get more specific and formulate things such that they might clash with reality. So for example, falsification, let's get specific on what we mean by falsify in the first place. Now, generally speaking, people tend to understand the word refute or falsify to mean something like we have evidence that shows that a specific theory is false. This seems like a natural enough interpretation, doesn't it? This is how most people would understand the term falsify. But in fact, it is impossible to ever use an experiment to falsify or refute a theory like this due to the Duhem-Quine problem. So Duhem-Quine is the idea that you never test a single theory directly. You always test a theory plus a slew of auxiliary theories. So for our purposes, we'll simplify. Suppose Theory X can be tested with exactly one auxiliary theory, let's say Theory Y, which is the theory of how your instruments work.

[00:44:18]  Unidentified speaker: Okay.

[00:44:18]  Bruce: Now, that's pretty unrealistic, I admit. We're just trying to keep it simple for hypothetical purposes. To make it slightly more realistic, we're going to throw in a Theory Z, which is our ceteris paribus Clause, i.e. all things being equal, which we take to mean we know of no other force that could impact this.

[00:44:35]  Unidentified speaker: Okay.

[00:44:36]  Bruce: So we've got three theories now. Theory X, the theory we're interested in, Theory Y, a theory about our instruments, and Theory Z, our assumption there's no other forces. Now suppose we find Theory X missed a prediction. Was Theory X falsified? If we're sticking with how regular English works, the answer is no, it was not. Because of course, the problem might instead be Theory Y or even Theory Z. So people that read Popper read about his asymmetry between verification and falsification, and they think about it for a moment, or they hear about the Duhem-Quine problem, and then they realize, hey, you can't actually falsify any theory with a single experiment either. So there is no asymmetry. Then they think they've falsified Popper.

[00:45:24]  Unidentified speaker: Or ironically enough.

[00:45:26]  Unidentified speaker: And they're done.

[00:45:27]  Bruce: They dismissed the theory without further consideration. Now CritRats today will give a response that I've seen that's somewhat questionable to this approach, to this criticism. They might say, well, of course, we didn't mean you absolutely knew for sure that Theory X was falsified. We meant you could tentatively falsify Theory X. And the Popper skeptic scratches her head and thinks about it for maybe five seconds and then responds, okay, but maybe when a test passes, I can tentatively verify a theory instead based on that. And the CritRat response tends to again be somewhat questionable. He might say, no, no, no, no, no. Didn't you know that verification is always absolute and certain and falsification is always tentative and uncertain? And the Popper skeptic thinks for another five seconds and then says, I'm okay that I'm introducing a brand new idea called tentative verification. That is the mirror image of tentative falsification. And I'm now claiming that I can tentatively verify theories with tests. And the CritRat at this point probably moves to insults or attacking on what words were used because they're not really sure what to say because they've backed themselves into a corner. And the Popper skeptic takes this as proof that she was right all along and skips off on her merry way. And this is, I'm going to assert, not too far from the truth of why Popper does not gain ground philosophically speaking. By the way, I call this above CritRat argument the absolute verification fallacy.

[00:47:03]  Bruce: It's the idea that Popper's asymmetry is explained by the idea that verification is always certain and falsification is always tentative.

[00:47:12]  Unidentified speaker: Let's ask this a different way.

[00:47:14]  Bruce: What is the CritRat misunderstanding that is causing them to fall into the absolute verification fallacy in the first place? Or in other words, what is the problem that they are trying to solve? The problem, if we get a little more specific, is that it's impossible to falsify or refute a theory with an experiment due to the Duhem–Quine problem. But this is actually a distinct problem from the fact that nothing is ever certain, although that's true too. So the CritRat is trying to slap the word tentatively in front of the word falsified to acknowledge the problem, and then they end up confusing two distinct problems. In their mind, they imagine they've solved the problem now. They are saying, sure, you can't actually refute a theory, but you can tentatively do so. But this weakens the asymmetry of refutation and verification that Popper promised us, or maybe even eliminates it, thus undermining one of the main planks of Popper's epistemology. Or to put this another way, the CritRat has solved the problem by quote-unquote, solved the problem, by vague-manning Popper's epistemology, which, as it always does when you vague-man a theory, removed valuable content from the theory.

[00:48:25]  Unidentified speaker: To be sure, the absolute verification fallacy is a fallacy.

[00:48:29]  Bruce: It's a misunderstanding of Popper. So Popper never intended the absolute verification fallacy in the first place. His actual asymmetry between refutation and verification lies elsewhere. If you're curious about that, check out episodes 41 and 42, Popper Without Refutation, for a discussion of Popper's actual intent here. It's not going to matter for our purposes today.

[00:48:50]  Unidentified speaker: Okay, so but the simple truth is, the CritRat isn't entirely off base here.

[00:48:55]  Bruce: They are just literally interpreting the term falsify to mean literally show a specific theory X to be false. The problem is that falsification in English is used to mean specifically proving theory X wrong via experiment. Which is literally impossible due to the Duhem–Quine problem. So if by falsification we mean showed the specific theory X to be wrong, that is exactly what you have not shown due to the Duhem–Quine problem. Indeed, you haven't even probably shown the theory to be wrong. Now, of course, most CritRats know not to use the term probably here. But you haven't even tentatively shown the theory X to be false.

[00:49:41]  Unidentified speaker: Why?

[00:49:42]  Bruce: Because the problem could be one of your auxiliary theories that is false. So there is not even a sense in which you have tentatively falsified theory X. What you've really shown is that the combination of theories, theory X, Y and Z is false. One of them is false.

[00:50:00]  Unidentified speaker: At least one of them is false.

[00:50:02]  Bruce: And that is all you've shown. Duhem–Quine is a harsh mistress, unfortunately. So we need a totally different approach if we're going to solve this problem. Now, I previously said that people that make this argument against Popper's theory, that think that they've now quote unquote falsified Popper, that they tend to just be kind of done. And they say, okay, I've disproven Popper. I don't even need to consider him any further. Popperians will often here try to claim that this is a failed or incorrect falsification. But I don't think that's quite right. I think it's a correct falsification of one specific possible way, incorrect but possible way, to read Popper's theory. Today, we now call this falsified version of Popper's epistemology, naive falsificationism. Now, naive falsificationism is, in fact, a falsified theory. It is refuted by the Duhem–Quine problem because it's a contradiction to it, and you can easily find counter-examples. This means that naive falsification, even though it's false, has one big thing going for it that the more popular but vaguer invite criticism correct errors version of Popper's theory did not have, and that is that naive falsificationism can clash with reality and thus be falsified, whereas the invite criticism correct errors version can't. So in a limited sense, the critic was correct. A certain specific interpretation of Popper's epistemology has indeed been falsified.

[00:51:40]  Peter: Sorry, I just wanted to review for our listeners. I get confused about some of this terminology too, but the Duhem–Quine thesis is just the idea that our ideas exist in interwoven webs of theories.

[00:51:57]  Unidentified speaker: Yeah.

[00:51:57]  Peter: So there's always something that you would need to, some kind of assumption you would need to make, I guess, to believe in practically anything.

[00:52:11]  Bruce: So, the absolute verification fallacy is clearly not a correct way to solve the problem that we're trying to solve, because it simply is a word trick, where we vaguely introduce tentatively in front of falsified and then don't really explain what it means in practice, thus allowing the same word trick to the critic, which is why this is a self-undermining argument.

[00:52:34]  Unidentified speaker: What might be a better approach?

[00:52:36]  Bruce: Enter the much-hated Lakatos. Lakatos, who was a student of Popper, his apostasy from Popper started with this fairly obvious problem with naive falsificationism, except he took seriously that you can't solve this problem by merely slapping the word tentative in front of the word falsified. Lakatos wanted a way to resolve this problem more convincingly, by which I mean he wanted a way to resolve it that was still, in principle, falsifiable. So Lakatos developed this idea, later independently developed by David Deutsch in his remarkable paper The Logic of Scientific Experiments, that to truly falsify a theory, you need a second theory. Here's the Deutsch version of it from the paper Logic of Scientific Experiment, page 8. In this view, a scientific theory is refuted if it is not a good explanation, but has a rival that is a good explanation with the same or more explicanda. So another consequence is that in the absence of a good rival explanation, an explanatory theory cannot be refuted by experiment. At most, it can be made problematic. Now, let's dwell on this idea a bit, and let's see why Lakatos and Deutsch both felt that this error correction to Popper's epistemology was necessary, as well as let's make sense of why Popper regarded this as a straw man of his epistemology, and literally eventually leading to their boy band breaking up, right? Because this was something that became a wedge between Popper and Lakatos. So to Deutsch, as Deutsch puts this, we could think of a counter-example to a theory as a problem to be solved, rather than as a falsification of a theory.

[00:54:23]  Bruce: So earmark that, that's a really golden idea for later, and it's not a bad idea, okay? Now, under Deutsch's and Lakatos' version of falsification, that problem, this counter-example that we're now calling a problem, will get upgraded to a falsification if and only if a new replacement theory is found. Under this way of thinking, the perihelion of Mercury, or for that matter, Lorentz contraction of light, did not falsify Newtonian physics, because you can only falsify—refute a theory once a replacement theory, in this case, general relativity, is invented. So let's call this new theory of falsification, the, quote, two theories falsification, quote, view, because it requires a new explanation, a replacement theory to falsify an old one. In addition to a refuting counter-example. This is, I admit, a smarter idea. But still, I'm going to argue a false interpretation of Popper's epistemology. The idea deserves some credit, though. What it does is it accepts the word falsify or refute in English means something like we falsify Theory X specifically, as opposed to falsifying some giant collection of theories. So, say you find a misprediction to Theory X, that this new version of falsification that Deutsch and Lakatos are putting forward, accepts that due to Duhem–Quine that a misprediction does not refute, quote, or falsify, quote, the specific Theory X that you're interested in. But say you later have a better theory that subsumes the original Theory X, including explaining why it missed the prediction. Those that make this argument claim something like this.

[00:56:11]  Bruce: Now that you have an explanation of what that prediction was missed, sorry, of why that prediction was missed, as well as a new explanation or theory that matches that prediction, you can now safely, though still tentatively, afford to consider the original Theory X as refuted or falsified. So it tries to solve, this approach tries to solve the problem by creating a much higher standard than a mere missed prediction. You are now required to actually have an alternative explanation as well.

[00:56:42]  Unidentified speaker: And I admit this much.

[00:56:44]  Bruce: In most cases, if you got to the point where you have an alternative explanation, say general relativity compared to Newton, you probably do have a very good reason to tentatively now consider the original Theory refuted. But it probably is mostly a pretty safe, so it probably is mostly a pretty safe standard for most purposes. So if the goal of this whole exercise was merely to decide if we can now use the word refuted or falsified with respect to a specific theory, with a high degree of confidence, whatever that means to you, and of it not being overturned later, then this isn't the worst possible answer.

[00:57:26]  Unidentified speaker: It's at least a decent heuristic.

[00:57:29]  Bruce: But what I found, though, is that most people, including most critical rationalists, what they do with a theory like this, that Lakatos' or Deutsch's version of falsification, is they become positivists. They look at the fact that general relativity was not considered falsified when the perihelion of Mercury or the Lorentz contraction was discovered. They then note that Eddington's expedition and experiment during the eclipse is often cited as the point at which Newtonian physics got quote-unquote refuted or quote-unquote falsified. So never mind that that's actually historical revisionism. We'll talk about that in a future podcast. This whole narrative of the story fits the Deutsch-Lakatos view of two-theory falsification. Having found a positive example of the theory, most people declare it obviously true and they move on and they're done. But not so fast, because as Popper explains, positivism is a bad epistemology. I call this mistake, by the way, narrative-izing, where we tell ourselves a story about how certain observations are explained well by a theory without considering what problems our theory might have. It's equivalent to counting white swans. However, I can, it can, if done correctly, at least show a theory is not ad hoc, but that's a discussion for another time. What we really want to try to do is to seek a counter-example to Deutsch–Lakatos two-theory falsification. Can we find an example of a theory that is understood to be falsified even though no replacement theory has yet been found?

[00:59:09]  Bruce: Now, once I phrase it this way, notice how the fact that I phrase it very specifically makes it so much easier to deal with. It now becomes easy. General relativity itself, as well as quantum mechanics, are both considered falsified theories today, even though neither has a replacement theory, which would be quantum gravity, as of yet.

[00:59:29]  Unidentified speaker: So guess what?

[00:59:30]  Bruce: We just falsified Lakatos–Deutsch’s theory of two-theory falsification. So we falsified this error-corrected version of naive falsification. Now, I know at this point, many crit rats will likely try to ad hoc save the theory. They might say something like this. But there is a second theory, Bruce. For general relativity, the second theory is quantum mechanics. And for quantum mechanics, the second theory is general relativity. Now, of course, this isn't what Deutsch or Lakatos originally said. When they spoke of needing a second theory, they specifically had a replacement theory in mind. In that case, we have no replacement theory, which would be quantum gravity.

[01:00:12]  Unidentified speaker: It doesn't exist.

[01:00:14]  Bruce: Now, if you doubt that, consider this quote from Deutsch. Page 9 of his paper. A test of a theory is an experiment whose result could make the theory problematic. A crucial test, the centerpiece of scientific experimentation, can on this view take place only when there is at least two good explanations of the same explicandum. Good that is apart from the fact of each other's existence. Ideally, it is an experiment such that every possible result will make all but one of those theories problematic, in which case the others will have been tentatively refuted. Note how Deutsch is clearly placing refutation within the context of a crucial test between two competing theories. Quantum mechanics and general relativity are not refuted in that we did a crucial test between them and one of them survived. They're not even directly competing theories, right? They're, that's exactly the problem. We need quantum gravity to be able to fit what Deutsch is saying here on page nine. So at a minimum, we can still consider the original Lakatos-Deutsch theory now falsified and replaced with the generalization of what a second theory means, and that now that's that quote, second theory does not require—is not required to be a replacement theory.

[01:01:35]  Unidentified speaker: Let me restate that.

[01:01:36]  Bruce: I didn't state that well. What I'm trying to say here is that Lakatos and Deutsch had in mind the second theory being a replacement theory. At a minimum, I have falsified that. Yes, you could ad hoc save the theory and say, well, second theory can mean just any second theory, not necessarily a replacement theory. What you're doing when you do that is you're generalizing. You're not really saving the original theory from falsification. You're coming up with a new generalized theory that isn't subject to that falsification, okay, by being vaguer about what you mean by second theory. But this approach has a serious problem. It's now itself unfalsifiable. Once you relax the requirement that second theory has to be an actual replacement theory, it becomes very vague, indeed entirely subjective, as to what you can or can't count as a so-called second theory. But, note that this is good news for Deutsch and Lakatos' theory. Because it means their theory was falsifiable in its original formulation. We know that is the case because it's now been falsified. It's the ultimate proof, right? This means the original two-theory falsification view can be error-corrected, whereas the vaguer version of meaning second theory just to mean some second theory, that can't be falsified.

[01:02:58]  Unidentified speaker: So, here's my key point.

[01:03:00]  Bruce: There was a specific falsifiable reading of Popper, naive falsification. It was falsified and thus could be error-corrected. Lakatos and Deutsch both independently offered a falsifiable improved theory, what we're now calling the two-theory falsification. It was falsified and thus could be error-corrected. But in each case, a decision to formulate the theory such that it could be falsified had to be made. We could have just stopped with the vaguer version of Popper's epistemology, invite criticism and correct errors. This version is unfalsifiable due to its vagueness. Or we could have ad hoc saved two-theory falsification by simply being very vague about what counted as a second theory. Or, but this would cost our theory its falsifiable status. Now, if you are after a theory that feels right to you, that is, that it lands for you, the unfalsifiable versions of these theories will work just fine. But if what you're after is an error-correctable theory that will help you get closer to the truth, the falsifiable and indeed false versions of Popper's theory are much better theories, and you should prefer them. So let's formalize this idea. If you want to get closer to the truth, if that's your goal, a falsifiable and maybe even false theory is much better than a true but vague theory. Now I know this seems truly counterintuitive. It probably even sounds outright weird. Shouldn't you prefer a true theory over a false theory? It seems like that should be the case.

[01:04:40]  Bruce: But I hope I've now shown why you should actually prefer the falsifiable and false theory, because it allows you to see real problems that can be solved and error corrected, thus giving you a fighting chance of getting to an even truer theory. The true but vague and thus unfalsifiable version of these theories feel intuitively correct, but actually can't be error corrected or improved. So they are what we might call epistemological sinks. They satisfy the intuitions, but at the cost of being able to correct errors. What we really want is a way for us to correct errors, the errors of both naïve falsificationism and two-theory falsificationism, while still keeping our theories falsifiable.

[01:05:26]  Unidentified speaker: But how can this even be done?

[01:05:29]  Bruce: By the way, if you don't like that I'm referring to some high quality philosophical theories as falsifiable, I'm going to continue to refer to them as that, because I think that's the right term. Maybe I can at least get you to admit that, as I've shown, some philosophical theories are far easier to criticize than others. So most crit rats will insist philosophical theories can't be falsified, even Popper says that, so they're going to have a problem with me saying that. But most of them will admit, well, some theories are more criticizable. So let me pull that thread just for a second and prove my point here. We can refer to these higher quality philosophical theories that are highly criticizable. We can call it that instead of falsifiable, if you wish. It doesn't matter what we call it because it's the same concept either way. What I want you to note, and this is the thing I'm hoping you'll get, that those theories that are more exposed to strong objective criticisms, even if they're philosophical, take the form of something similar to, if not identical to, what Popper called falsifications. That is, that these stronger, high-quality philosophical theories expose themselves to counter examples or logical contradictions, just like an empirical theory does. Now, why is it so difficult to come up with a falsifiable way to solve this problem? I'm going to suggest that it has to do with a blunder on Popper's part. It's because at some level, you really badly want the word falsified to mean exactly what it seems to mean.

[01:07:02]  Bruce: That in some sense, we have shown a specific theory X to be false, at least tentatively. Not that you showed some large collection of theories to be false, but we don't know which one.

[01:07:13]  Unidentified speaker: I mean, that makes sense, right?

[01:07:15]  Bruce: Like really stop and think about this just from a lay person's perspective. The word falsifies just misleading here. How can falsificationism not mean that we're showing that the theory of interest is false? Frankly, at least in the English language, if you say Theory X can be falsified, that phrase just carries too strong of a connotation for Popper's purposes. It implies that Theory X can actually in some sense be by experiment shown to be false, which is never actually true due to Duhem–Quine’s problem. If you are going to hold on to the idea that falsify as a word means in some sense that Theory X is, quote unquote, shown to be false, even just tentatively, then falsification becomes literally impossible period, end of story. But what else could falsify as a word even possibly mean if not shown specifically to be false? What makes this worse is that Popper's wording seems to suggest this is indeed what he meant. So consider these quotes from Popper. This is why I call it Popper's blunder. These quotes I'm about to read, I don't think they're true statements, right? So can, this is a quote from Popper, can the claim that an explanatory universal theory is true be justified by empirical reasons, that is, by assuming the truth of certain test statements or observation statements, which it may be said are based on our experience? My answer to the problem is the same as Hume's.

[01:08:43]  Unidentified speaker: No, it cannot.

[01:08:45]  Bruce: No number of true test statements would justify the claim that an explanatory universal theory is true. But there is a second logical problem, which is a generalization of the first problem. It is obtained by merely replacing the word is true by the word is true or that it is false. Can the claim that an explanatory universal theory is true or that it is false be justified by empirical reasons? That is, can the assumptions of the truth of test statements justify either the claim that a universal theory is true or the claim that it is false? Keep in mind, we are talking about the truth of test statements. Okay, to this problem, my answer is positive. Yes, the assumption of the truth of test statements sometimes allows us to justify the claim that an explanatory universal theory is false. That's all from Objective Knowledge, page seven. Now tell me, how do you reconcile that statement to the Duhem–Quine problem?

[01:09:45]  Unidentified speaker: I assert you can't.

[01:09:47]  Bruce: Without any further context, if you look at full context of Popper, we give him credit. Like, obviously, I'm stealing words out of context, and he gets this right in all sorts of places, and I'm not claiming otherwise. I'm claiming that this statement is desperately misleading because it happens to be false. Okay, even if in context we can maybe interpret it in a more positive way, all right? Without any further context, Popper is wrong here. You cannot solely have simple empirical reasons, such as the outcome of an experiment, to justify a theory as false. Note that even if you believe the two-theory falsification version from Deutsch, that this is still a false statement, since you really need an empirical reason plus a second theory to show that it's false. So let's be honest, that last quote really does sound a little too much like naive falsificationism. I can give you other quotes from Popper that sound just like this.

[01:10:46]  Unidentified speaker: How about this one?

[01:10:48]  Bruce: In the first level, there is a logical asymmetry. One singular statement, say, about the perihelion of Mercury, can formally falsify Kepler's law. That's Conjecture and Refutation, page 54.

[01:11:01]  Unidentified speaker: Or this one.

[01:11:03]  Bruce: We have seen that theories cannot be logically derived from observations. They can, however, clash with observations. They can contradict observations. This fact makes it possible to infer from observations that a theory is false. Infer from observations, not observations plus a second theory, that a theory is false. The possibility of refuting theories by observation is the basis for all empirical tests.

[01:11:27]  Unidentified speaker: That's from Conjecture and Refutation, page 260.

[01:11:30]  Bruce: These are technically false statements, aren't they? You really and truly can't formally falsify Kepler's laws, or Newtonian physics for that matter, with one singular statement. Nor can singular observations, or any set of observations really, allow you to infer a theory is false. Strictly speaking, you can only really falsify the sum total collection of theories related to that prediction including all the background knowledge and auxiliary theories. If you want to infer a theory is false, something more is clearly needed. But what is that something more? Statements like this are precisely why people come away from reading Popper and they feel like Popper is preaching naïve falsificationism. I think if you read Popper carefully and you don't just take a quote out of context like this, like I just did, you can tell Popper never really believed in naïve falsificationism. He's really relatively clear on this point in total.

[01:12:28]  Peter: And just to be clear, are you saying that he would not have accepted the two theory version of his idea?

[01:12:37]  Bruce: No, he did not accept it. He rejected Lakatos' version of that.

[01:12:41]  Unidentified speaker: Okay.

[01:12:42]  Unidentified speaker: Okay.

[01:12:42]  Bruce: So we know he didn't accept it.

[01:12:45]  Peter: There's no ambiguity there.

[01:12:46]  Unidentified speaker: Okay.

[01:12:47]  Unidentified speaker: Okay.

[01:12:48]  Bruce: I'm going to explain why. There's actually a pretty good reason why.

[01:12:51]  Unidentified speaker: Okay.

[01:12:52]  Bruce: When later accused of being a naïve falsificationist by including his own student, Lakatos, Popper would get angry and point out to all the places where from the beginning, and it was true, it was from the beginning, he did not say you cannot falsify that. He did not say you can falsify a theory with certainty. That is, Popper tried to jump to the tentative falsification strategy, just like crit rats do. Even though this is really, for the reasons I just explained, a pretty bad response to the very real problem that's being raised. Yet Popper never really seemed to have fully understood why he himself left this impression of naïve falsificationism. Popper also never understood why merely saying, I've never said you can falsify a theory with certainty, was not really an appropriate response to the concerns people were raising with his theory of falsificationism. More to the point, Popper never realized he had made some technically false statements, like the ones I just read, that needed to be corrected and that would inevitably cause people to misunderstand him. This is why I say it's Popper's fault that this problem exists and holds back the success of critical rationalism. Consider that Popper has to, late in his career, attempt to explain that by falsifiable, he didn't actually mean you could, via experiment alone, show a theory to actually be false. So here is Popper in Realism and the Aim of Science, that's a 1983 book, so rather late in his career. So, Logic of Scientific Discovery, first published in 1934, in German. So this is 50 years later.

[01:14:32]  Unidentified speaker: Okay.

[01:14:33]  Bruce: Popper is doing his best to clarify that by falsifiable, he didn't actually mean you could actually show a theory to be false. Here's the actual quote from Realism and the Aim of Science, page XXIII. We must distinguish two meanings of the expression falsifiable and falsifiability. Falsifiable one, as a logical, technical term, in the sense of demarcation criterion of falsifiability. This purely logical concept, falsifiable in principle, one might say, rests on a logical relationship between the theory in question and the class of basic statements, or the potential falsifiers described by them, versus falsifiable two, in the sense that the theory in question can definitively or conclusively or demonstratively be falsified, demonstratively false. I've always stressed that even a theory which is obviously falsifiable in the first sense is never falsifiable in the second sense. For this reason, I have used the expression falsifiable as a rule only in the first technical sense. In the second sense, I have, as a rule, spoken not of falsifiability, but rather of falsification and its problems. I mean, like, no wonder people are so confused at this point.

[01:15:50]  Unidentified speaker: Does that clear things up?

[01:15:51]  Bruce: I mean, I understand what he's saying.

[01:15:53]  Unidentified speaker: But here he is.

[01:15:54]  Bruce: He's still trying to explain this technical difference that by falsifiable, he had this really technical, logical meaning in mind. I think that's true. Like, I think if you understand that that's true and you kind of just dismiss out of your mind that the word falsifiable just does not mean to Popper what it normally means to people. I think it's possible to read Popper from the beginning, from logic of scientific discovery on, and to realize that this is what he was saying all along, okay? But man, it's confusing. It's not a little bit confusing.

[01:16:26]  Unidentified speaker: It's overwhelmingly confusing.

[01:16:28]  Bruce: It's a terrible, terrible choice of words. Here he's admitting that falsifiable is merely a logical term that has nothing to do with actually showing Specific Theory X to be false. Falsification is a bad term for what Popper actually had in mind, which was more of a logical idea. What everybody else wants to know is how do I tell if Theory X specifically is false? And Popper is claiming to not even be trying to answer that question. This is one of the main reasons why trying to explain Popper's epistemology in terms of the word falsification is truly unfortunate. Inevitably, people understand the term falsification to mean show specific Theory X, at least in some sense, to be false. People just have a really hard time wrapping their minds around this expanded idea of falsification that ultimately is going to involve falsifying a collection of theories instead. But this less conventional sense of falsification, this logical sense, and the fact that it deals with a collection of theories instead of a specific theory, that is the only kind of falsification that the Duhem–Quine thesis or problem is going to allow if you're just looking at observations that come from an experiment. It doesn't feel like falsification at all. This is why people get so confused. Now, I got a trick question for you before we continue. I need a little bit of an aside because it's going to be important.

[01:17:58]  Peter: Okay.

[01:17:59]  Bruce: Consider this quote from Peter Medawar, okay, from The Philosophy of Karl Popper, page 283. One of the strongest ideas in Popper's methodology, that the only act which the scientist can perform with complete logical certainty is the repudiation of what is false.

[01:18:21]  Unidentified speaker: Okay.

[01:18:23]  Unidentified speaker: Is that a true statement?

[01:18:24]  Bruce: Like, stop and think about that for a second. Can you with complete certainty have a repudiation of what is false? And by the way, doesn't that sound like naive falsification to you? I mean, like, couldn't we maybe argue that Peter Medawar has misread Popper as a naive falsificationist here? And isn't it, like, technically incorrect because falsifications are tentative and not logically certain?

[01:18:50]  Peter: He's literally says it seems like it contradicts fallibilism in a way to say complete certainty, right?

[01:18:57]  Bruce: Okay, I think most people would agree with you. So here we have Medawar falling, seemingly falling into naive falsification due to Popper's poor wording, right?

[01:19:10]  Unidentified speaker: Or maybe not.

[01:19:10]  Unidentified speaker: Let me explain.

[01:19:12]  Bruce: Here's a question for you to consider. Can you with complete certainty tell by a counterexample that one of your theories is false?

[01:19:22]  Unidentified speaker: Be careful here.

[01:19:24]  Bruce: You've likely heard over and over and over that there is no such thing as absolute certainty. So you may be tempted to answer here, of course not, that would violate fallibilism. But in fact, the correct answer is yes. You can tell with complete certainty that one of your theories, including your auxiliary theories, the background knowledge, the assumption there's no other forces, one of them must be false. Now why must that be true? Because the very fact that you missed a prediction definitively does mean that something is wrong with your collective theories. Think about this carefully. It must of necessity mean this. It's literally a logical necessity. Now it may be that the theory that you have wrong is something totally mundane, like my instrument is working correctly. Or it might, maybe something like, I've not correctly thought through these implications, or even just I'm currently confused about what I even mean. Like, there's all sorts of different ways we might use the word theory here, okay? Maybe sometimes somewhat vaguely. But it must be the case that something is wrong with your overall collection of theories and ideas if there is a misprediction. But it seems really weird to call this a falsification. That was why Deutsch preferred the term problem. If we want to adopt Deutsch's language, we would say it like this. Problems are objective. You have real problems that need to be solved, and they objectively exist. There's really no such thing as a problem not being a problem. Now, I anticipate here a possible objection. What about a pseudo problem?

[01:21:02]  Bruce: Isn't that by definition a problem that wasn't a problem? So aren't pseudo problems examples of problems that turned out to not be real problems after all? After all, thus you were mistaken about a problem being a problem. Well, yes and no. Okay, and let me be, I'm trying to be as clear here as I possibly can, because I'm making a point, it is important, but I acknowledge that a problem can be a pseudo problem. Here's the thing you have to keep in mind. Even a so-called pseudo problem is still a case of some idea you are holding being wrong in some way, or that some misunderstanding took place that needs to be error corrected. Typically, when we talk about a pseudo problem, the most common example would be that your use of language was vague and you got confused and you accidentally equivocated in some way, thereby creating a problem that didn't exist with the theory. One thing you may have heard me say in past podcasts is that it's impossible to tell the difference between a problem and a pseudo problem until you have the solution in hand. A pseudo problem has to be solved just like any other problem. The difference has to do with the nature of the solution. If the problem was that the theory in question, Theory X, let's say, has to be modified, then we in retrospect call it a quote unquote real problem. But if the problem turned out to be that you misunderstood Theory X's implications, perhaps due to a misunderstanding of language, then we call it, in retrospect, a pseudo problem. But for our purposes, a pseudo problem will be considered a legitimate problem.

[01:22:37]  Bruce: It is in this sense, and really only this sense, that I'm claiming that it's possible to be certain that a problem exists. And let's be clear, I'm really just doing a tautology here. Okay, like there's not a lot of content with what I'm saying. It may sound really shocking or really at odds with fallibilism.

[01:22:53]  Unidentified speaker: It really isn't.

[01:22:55]  Bruce: The reason why is that if you have to solve a problem, even if you later realize it was a pseudo problem, by definition, it meant that you were mistaken about something and you had a problem to solve, right? It's a little like saying the very fact that you have a problem means you have a problem. And that's why I say it's closer to a tautology, right? And yet, it does mean you always have something wrong with the way you're thinking, and therefore there's something in your ideas that must be error corrected. So let's, in that context, look at the Peter Medawar statement again. One of the strongest ideas in Popper's methodology, that the only act which a scientist can perform with complete logical certainty, is the repudiation of what is false. But if you actually meant you can know that something is wrong, that could be a false statement, I admit. But if he just means you know something is wrong, I actually think he's technically correct. This is a key insight that's going to help us come up with a version of falsificationism, that is falsifiable, but that isn't yet proved false. The last idea we need is of course the idea of ad hocness, or rather the no ad hoc rule. The idea of ad hocness or the no ad hoc rule is actually pretty simple. Think of the no ad hoc rule as a normative methodological convention that scientists all, perhaps implicitly adopt, that says something like this. You are only allowed to save a theory from refutation due to one reason.

[01:24:21]  Bruce: If you can offer a new auxiliary theory that is not ad hoc, where ad hoc is defined by Popper as having independent tests, or for our purposes, having independent checkable consequences, not related to the problem being solved. This is also known as reach in Deutschian terms. Or put another way, you may only save your theory from empirical refutation by increasing the overall empirical content of the collection of theories involved. What I have called Popper's Ratchet, okay? Because it ratchets towards a higher overall theory content. Moreover, in episode 93, as well as in this episode, of course, I made the case that Popper's Ratchet or the no ad hoc rule was easily adapted to even philosophical theories. And indeed, our whole discussion so far of Popper's epistemology bears this out. It is entirely possible, indeed necessary, that we formulate Popper's epistemology such that it has checkable consequences. If we find that naïve falsification fails in real life, we can put aside that version of Popper's epistemology as false. And if we find that two-theory falsification doesn't always apply in real life, i.e. that there are counter examples to it, we can conclude that it's not a wholly correct theory. So there really is no reason why we can't apply Popper's Ratchet and the no ad hoc Rule to philosophical theories. So we must modify and generalize Popper's Ratchet to be something like, you may only save your theories from refutation by increasing the overall checkable content rather than testable content of the collection of theories involved.

[01:26:02]  Bruce: It's a very minor change, but it allows us to apply the no ad hoc Rule to philosophical theories such as induction and Popper's own falsification. Why does it matter so much that a theory has to be formulated such that it has independently testable or if you prefer, checkable consequences distinct from the problem being solved?

[01:26:21]  Unidentified speaker: Popper explains this.

[01:26:22]  Bruce: Let me give the explanation. Suppose you have a Theory A and it misses a prediction. So suppose you come up with a Theory B, which has exactly the same empirical content as the first theory, except the single missed prediction. Would you now consider Theory B the better theory? Many people would say yes here, by the way. Understand how infinitely easy it is to create a Theory B once you already know what the missed prediction was, and understand that there is in fact an infinity of such Theory Bs to consider. So from just a pragmatic basis, it doesn't make sense to consider Theory B the better theory, until not only can it explain the one missed prediction, but also makes its own independent separate predictions due to the content of its own theory. Now here is the big idea. While it is easy to dream up an infinity of Theory Bs that explain exactly one missed prediction, it's very difficult indeed to explain that missed prediction in a way that also makes some other unique and preferably unexpected prediction. In fact, if your Theory B does make an independent but unexpected prediction, there is simply no way to make sense of what just happened, so long as you're following the no ad hoc Rule, but to assume that Theory B must have something to it. This means that the no ad hoc Rule is in some sense recursive. Only theories that make unique or unexpected predictions are allowed into the debate in the first place.

[01:28:01]  Bruce: If such a theory ever makes a misprediction, they must then only be saved via a new explanation or theory that is itself independently checkable, and thus not ad hoc. But let's go back to our toy example, Theory X and Theory Y, with the one auxiliary theory, Theory Y, which was the theory of how the instrumentation works. How would the no ad hoc Rule play out here? Okay, this is, we're now getting to the answer to the problem of how to understand falsificationism so that it doesn't just get falsified, but it yet is still falsifiable itself. So say Theory X misses a prediction. Duhem-Quine is often understood as saying, you can't know if a failed experiment falsified your theory, but really that isn't what it's saying. What Duhem-Quine really says is something a bit more specific. It says, a misprediction actually does falsify the combined collection of theories involved. That's why I spent some time making sure it was clear that was the case. That is to say, you know for sure either Theory X or Theory Y is false. Or realistically, maybe it's Theory Z or ceteris paribus Clause that's false. So you then theorize, my instruments are broken. And you come up with an experiment to check that.

[01:29:25]  Unidentified speaker: That's Theory Y.

[01:29:27]  Bruce: Note that this is itself an independent test you come up with to check your equipment. So it does not violate the no ad hoc rule. Maybe you find your experiment doesn't work. Your equipment doesn't work when you do the second experiment on it.

[01:29:44]  Unidentified speaker: Hooray!

[01:29:45]  Bruce: Theory X is not refuted after all. Theory Y is refuted. And you move on with a new Theory Y. That is, you fix your instruments to be correct so that they really work. But suppose Theory Y passes its independent tests. You make sure the instruments are working on an experiment with a known outcome, say.

[01:30:06]  Unidentified speaker: Now what?

[01:30:07]  Bruce: Maybe you are able to creatively devise another non-adhoc theory about what is wrong with your equipment. So long as your theories are not ad hoc, you are allowed to attempt as many saves as Theory X as you wish by testing new Theory Ys. For the sake of argument, let's say you exhaust every single non-adhoc theory you can devise about Theory Y. That is, every single non-adhoc theory you can devise about your instruments being working correctly. Suddenly, you are starting to eye Theory X. Could Theory X actually be wrong? You're thinking to yourself. Maybe you next challenge Theory Z, the ceteris paribus Clause. You try out crazy ideas about some of the force you didn't account for that is somehow involved that you just didn't know about. So long as your crazy idea is not ad hoc, that is, it has independent checkable consequences not related to the original problem, that's okay. You're allowed to attempt to save Theory X in this way. But let's say that you exhaust all your crazy ideas on Theory Z also. Again, you start to eye Theory X and wonder could Theory X be wrong? This idea of exhausting non-ad hoc alternatives is probably one of the single most important ideas of Popper that he never happened to state clearly, that I can recall. It is implicitly part of Popper's no-adhoc rule, and thus an implicit part of his whole epistemology, but he seemed to see it as so obvious that it didn't need stating. But we're starting to really understand now what Popper really had in mind when he spoke of falsification.

[01:31:53]  Bruce: The key here is that non-ad hoc explanations are hard to come by, so you can exhaust them. But ad hoc explanations are cheap, so they are impossible to exhaust. But here is the key insight. If you are assuming realism to be correct, then any true theory will always have a non-ad hoc formulation. So you actually lose nothing whatsoever by simply disallowing non-ad hoc theories, even though they may well be true in some [loose?] sense. This is what we call an admissible heuristic in machine learning or in AI, and it means it drastically reduces the size of the search without losing any correct answers. A* algorithm is famously an admissible heuristic, where we use the direct length between two points as an approximation of the length through the maze. It will remove many different possible paths, but never remove a correct shortest path. Of course, if you doubt realism, then you would be skeptical of this claim. That's a podcast from another time though, or better yet, check out episode 99, where we already address this in detail. But for now, I'm going to assume that you are a realist. The key point here is that since the no ad hoc rule is an admissible heuristic, it's completely valid to always apply it. This is the sense in which all science relies on the assumption of realism being true, a claim that David Deutsch makes in Beginning of Infinity and that sometimes gets challenged. Now, if you exhaust all alternatives, does that guarantee you that Theory X is the falsified theory? No, of course not. All falsifications are still tentative.

[01:33:46]  Unidentified speaker: Why?

[01:33:49]  Unidentified speaker: That's the question you should be asking here.

[01:33:51]  Bruce: Notice that we now mean something far more specific when we say it's a tentative falsification. When we say this falsification is tentative, we are not talking about a stance on if it's possible to be certain or not. In fact, it is possible to be certain, at least in a very limited sense that we just talked about. No, what we're actually meaning now by tentative is we don't know if we just failed to be creative enough to come up with an alternative that solves the problem in another way. Now, I've likened this to the halting problem. You can exhaust all the alternatives you currently know about, but you can't exhaust alternatives that haven't yet been thought of. It is always possible a previously unthought of alternative will be discovered that forces you to rethink everything. But until then, you literally can't consider an alternative that doesn't exist. So don't bother. Instead, stick with whatever collection of alternative theories you do currently have available, as long as they're non-ad hoc. If there is one theory in particular that has no known problems, it's rational to go with that theory regardless at this point. That's the concept of the sole surviving theory or the best theory. The Duhem-Quine problem poses no problem at all, so long as you situate it within this process of exhausting non-ad hoc theories, rather than to try to see if a single observation or experimental outcome falsifies a theory. That is, I think, the correct way to understand Popper's falsification, in terms of a process of exhausting non-ad hoc theories.

[01:35:33]  Bruce: This is one of the big differences between Bayesianism and critical rationalism. Bayesianism purports to be a way to decide which theory is probably correct now, whereas critical rationalism describes the process you will follow to eventually eliminate alternative theories via the process of non-ad hoc theory exhaustion. They're not even talking about the same thing. Bayesianism and critical rationalism are entirely different things.

[01:35:57]  Unidentified speaker: Okay.

[01:35:58]  Bruce: Let's call this version of falsification non-ad hoc theory exhaustion, or maybe just theory exhaustion falsification for short. In previous podcasts, particularly 83, Popper's Second Axis, I tried to describe Popper's epistemology as having two axes. One axis was strict falsification, which theory has survived falsification the best, and the second axis was the no ad hoc rule, where we measure a theory as having potential falsifiers or not. By this point, I hope it's obvious that by falsifies, we actually mean the theory can clash with reality and thereby force the exhaustive search I've just described. But maybe two axes is the wrong way to think of it. Maybe it's more like two filters. The first filter is what is required to even get considered in the first place. Specifically, the theory must have independently checkable consequences that can clash with reality. This is a hard requirement even for philosophical theories. One possible exception would be if all the theories under consideration, none of them had checkable consequences. That's going to be a future podcast for discussion. My key point here is we prefer philosophical theories that have checkable consequences that make them vulnerable. But normally, if a theory is simply ad hoc, it doesn't even make it to the falsification process that follows. Then once you survive the first filter, which would be that you have checkable consequences, you then are allowed to participate in the actual falsification process of exhausting non-adhoc theories, including the auxiliary theories.

[01:37:36]  Bruce: This idea of two filters matches well with Deutsch's idea that most theories never make it to testing because they get criticized out of existence first. He was suggesting a version of this two-filters version of critical rationalism that I am suggesting. Now a fair question is if this non-adhoc, exhaustive search version of falsificationism is itself falsifiable. We first considered three problematic forms of Popper's falsificationism. The first was the invite criticism, correct errors version of falsificationism. We dismissed that one as being too vague to be falsified. We then considered naive falsificationism, which we dismissed by counter-example and by invoking the Duhem–Quine problem. We then considered the two-theory version of falsificationism where you have to have a replacement theory. We falsified that version by counter-example. We also considered inductionism, by the way, specifically a form where you can induce generalizations from only specific observations and showed it was logically impossible, thus falsifying it. In past podcasts, we also considered a vaguer version of induction where we relaxed the requirement of only specific observations, and we found that it was so vague that this form of induction may indeed actually just be talking about Popper's epistemology. So we dismissed it as vague and unfalsifiable. But what about our new exhaustive search version of falsificationism? It might seem unfalsifiable at first, for there is no counter-examples to it that I can think of either.

[01:39:16]  Bruce: Well, actually, I do have one possible counter-example that we'll consider in a future podcast, but let's ignore it for now. But it's still pretty clear what a counter-example to this theory would look like if one existed. If someone can offer a specific example of how to progress, how progress is made that requires an ad hoc theory, that would falsify this specific theory. In fact, indeed, recall in episode 107, Joseph Agassi claimed that he had falsified this theory. Now, I haven't looked up his actual argument, so maybe he has, maybe he hasn't, I don't know. If he has, that's okay.

[01:39:54]  Unidentified speaker: It would be a chance to error-correct the theory.

[01:39:57]  Bruce: Having a false theory is just not a bad thing, as long as the theory is actually error-correctable. But it's a good sign that this theory is far harder to falsify than the ones we've considered up to this point. This is not unlike how we talked about the Church-Turing thesis. This theory that we're talking about, this non-adhoc theory exhaustion version of falsificationism, it still has counter examples that we could recognize, that someone could come up with. You just can't find them in real life. This is very similar to the Church-Turing thesis thing, where we know that if you were to find an example of a computer, a design of a computer that had more power than a Turing machine, you would falsify the Church-Turing thesis. But nobody can find one. Nobody can even conceive what one would look like. I think this is where we're getting to with this non-ad hoc exhaustion version of falsification. It can't be falsified in the sense that you can't find actual counter examples, but you know what a counter example would look like. And that's what matters, okay? That it could in principle clash with reality. We could also falsify this exhaustive search theory by showing that there exists a form of induction and abductive achievement that does not require this theory. More on that in the future podcast. But for now, recall our discussion about animal learning and Donald Campbell's theory. If we can create an adaption to an environment without some form of falsification, blind variation and selective retention going on, then Popper's epistemology is refuted. It's falsified, okay?

[01:41:48]  Bruce: That's what I'm talking about, right? Like literally, it's not even that hard to conceive of what a falsification of this theory would look like. It's just hard to find actual examples of it. We could also falsify this theory if we could show that some true theories are in fact ad hoc, but that no non-ad hoc formulation existed for them. So there seems to be quite a few ways we could go about falsifying this theory in principle, but actual counterexamples are much, much, much harder to come by than the other forms of falsification that we've discussed today. This suggests that we have made some real progress here, even if this form also turns out to be false, ultimately. So did Popper accept theory exhaustion falsification? Now, surely he never spelled it out the way I just did, but it does seem to follow naturally from his own no ad hoc rule. Most would agree that my formulation is correct, at least for empirical theories. They may take more issue with me applying it to philosophical theories, perhaps. Let me make an assertion here. I believe that this theory exhaust falsification is what Popper had in mind, but didn't explain it very well. I might be wrong here, of course. Of course, arguments from silence aren't good arguments. But consider why Popper felt strawmanned by Lakatos by claiming falsification required a second theory, just as Deutsch did. Also, to Popper, falsification had nothing to do with showing a specific theory to be false. We saw that he is quoted saying that, right? But did Popper feel you were required to have a second theory?

[01:43:30]  Unidentified speaker: No, he didn't.

[01:43:31]  Bruce: Consider this quote from Popper, Conjecture Refutation, page 324. It is possible in quite a few cases to find which theory is responsible for the refutation or in other words, which part or group of hypotheses was necessary for the derivation of the refuted prediction. Note how Popper's claiming you can often figure out which theory in the collection was false without a second theory to jump to. While Popper may not have spelled it out as well as I would have liked, I think I can understand why Popper felt Lakatos was strawmanning him. Because if you simply follow the no ad hoc rule, if you simply take the no ad hoc rule seriously, you can and will come up with a best theory as to which of your collections of theories is false. Having an alternative theory is very helpful, but not necessarily required. All right, quick summary and then we're done. In summary, it is possible to falsify a theory of falsificationism, the theory of falsificationism, despite Popper's claim to the contrary. At least for some fair understanding of the term falsify that includes logical contradictions of any sort, not just empirical tests. Surely it should be obvious that naïve falsification has been falsified despite being a philosophical theory. And surely we can agree that induction has been falsified despite being a philosophical theory. And I've argued that the Lakatos–Deutsch idea of two-theory falsification while [a?] step forward from naïve falsification is also ultimately false because we can show counter examples to it.

[01:45:12]  Bruce: So at least in some cases, it is possible to falsify falsificationism. In addition, I've offered my own improved version of falsificationism, rooted in a slightly generalized version of Popper's own no ad hoc rule that I've called theory exhaustion falsification, where you come up with to come up with tests for all your theories, including auxiliary theories, and eliminate them as a source of the problem. Now, let me set expectations here. This is just an early look at how this works. I'm surely not claiming that I've formalized this logically to the same degree that Popper did with empirical tests. And it should be obvious that empirical theories are much easier to work with. With empirical theories, it's much easier to work out what counts as a counter example than a philosophical theory. And the goal in science is to squeeze out and remove subjectiveness from our theories so that we can rely on intersubjective, that is, objective criticisms. I admit that this is harder with philosophical theories, and it is unclear how to formalize this, to me at least. So there is clearly still a subjective element there that I haven't removed. But it should be obvious that naïve falsification and induction are both far easier to understand what a falsification of those theories would look like compared to the invite criticism correct errors version of Popper's epistemology, due to how vague it is. So for now, my main point is only that we can squeeze out a lot of subjectiveness from our philosophical theories, and thereby make them easier to criticize objectively and error correct.

[01:46:48]  Bruce: For those uncomfortable calling this process falsification, because that term now specifically to you means empirical falsification, that's fine. I don't mind calling this an attempt to get specific and remove a lot of subjectiveness from what it means to make a philosophical theory easier to criticize versus immune to criticism. The key insight I'm offering is that a high quality philosophical theory is a high quality theory for exactly the same reasons that an empirical theory is considered high quality, because it has lots of content that might clash with reality and thus allow us to see objective problems that can be error corrected. This is just an early sketch of how we might go about generalizing falsificationism to apply to all theories, not just empirical ones. In fact, I suspect that my theory exhaustion falsification is probably incorrect, but if so, who cares? It's clearly a step forward over the alternatives we considered and that is really all that matters. And if it's false, then let's get a specific false example and then let's error correct from there and improve the theory from there. And that is my argument for why I do think falsificationism is in some sense falsifiable.

[01:48:02]  Peter: Okay, so I'm on I'm on my next bumble date, and I meet a woman who has some philosophical interests, and I tell her I'm on a podcast about Karl Popper, and she says, Oh, well, wasn't he a naive falsificationist? Isn't what do I say? And it's the simple language as possible.

[01:48:34]  Unidentified speaker: Oh, I can give that to you.

[01:48:36]  Peter: I have to prepare for this.

[01:48:38]  Unidentified speaker: Yeah, absolutely.

[01:48:39]  Unidentified speaker: Absolutely.

[01:48:39]  Unidentified speaker: Here's the answer.

[01:48:40]  Unidentified speaker: Ready?

[01:48:41]  Unidentified speaker: Yeah, I can do it in one sentence.

[01:48:43]  Peter: Okay.

[01:48:44]  Bruce: Here's a link to the podcast episode that discusses that.

[01:48:48]  Peter: Okay, I don't know about that.

[01:48:50]  Unidentified speaker: Okay, we'll see.

[01:48:53]  Peter: I don't know that I might not get a second.

[01:48:56]  Bruce: And then tell her, don't worry, it's only three hours of discussion from one person.

[01:49:01]  Peter: Okay.

[01:49:02]  Unidentified speaker: Okay.

[01:49:04]  Peter: Here's one more one more one more quick question. You know about the the hundred men versus one grizzly thing, something that popped up all the all the all the high school students were talking about it for a couple of weeks.

[01:49:19]  Unidentified speaker: It was online.

[01:49:21]  Peter: Who would win one grizzly bear versus a hundred men?

[01:49:25]  Bruce: I'd heard that there was this thing going around the internet where they asked a woman, would you rather be with a grizzly bear or with a man alone?

[01:49:31]  Peter: Oh, that's a different bear thing. No, this is, I think, a more interesting question. So a hundred men versus one grizzly bear. I had my friend, my math teacher friend, who's a jujitsu expert making the case that there's absolutely no way that a grizzly bear could beat a hundred men, because even if some of the men are going to die, they can poke out his eyes or something. He made a convincing case. But assuming you're not going to actually test this, and assuming that you can actually come to a definitive, pretty, I would think, imagine, I mean, if you talk to a hundred combat experts and got different takes, you could probably come to a different take on this. Is this a falsifiable assertion?

[01:50:26]  Bruce: I would say it's not in its current form. So here's what I would do. So first of all, you have to ask, how important is this question to you? Because it may not be, right? It may just be a fun question.

[01:50:38]  Unidentified speaker: It's fun, yeah.

[01:50:40]  Peter: But pretend it is.

[01:50:41]  Bruce: And in its current formulation, probably subjective criticisms are the best you can do. So you would probably want to do exactly what you just did. Go to experts in combat and have them say, well, you know, a bunch of the men would die, but you could like poke out the bear's eyes, right? And there's, who knows, right? Notice that what the guy is doing is he's trying to make the formulation a little more specific, right? So that it is falsifiable, at least in principle. He's trying to say, well, I think that what you could do is you could get a hundred men that are willing to completely throw their lives away and they just don't care about how much they're wounded. And I think that in that case, because they're all zealots that are trying to do this, that some of them could poke the bear's eye out while he munches on the other 50 and eventually the bear would win. Notice how this is a way more falsifiable, way less subjective formulation, right?

[01:51:39]  Unidentified speaker: Oh, that's interesting.

[01:51:40]  Bruce: The falsification has gone up.

[01:51:42]  Unidentified speaker: Okay.

[01:51:43]  Unidentified speaker: Okay.

[01:51:43]  Peter: And that's increasing the empirical content. Oh, yeah.

[01:51:48]  Bruce: The checkable content, let's say.

[01:51:50]  Peter: The checkable content. Okay. No, that's interesting.

[01:51:54]  Bruce: And so that's what I have in mind, is that there's this choice to make the theory testable or checkable, right? And you either make the choice to get specific, and therefore your theory can be error corrected, or you don't make the choice, and you stick with the really vague formulation of the theory, in which case the theory just, there's just not much you can say about it, right? But it has no content.

[01:52:19]  Unidentified speaker: That's the problem.

[01:52:21]  Bruce: And so I think in most cases, I won't say it's all cases. I mean, for me to say it's all cases would be silly, because that would mean I have literally checked every idea out there, right?

[01:52:33]  Unidentified speaker: Yeah.

[01:52:33]  Bruce: But I think in most cases, when you have an idea or a theory of any sort, it can come in vaguer forms or it can come in more checkable forms. And the vaguer form is kind of encompassing, you know, a large group of more checkable versions.

[01:52:53]  Unidentified speaker: Okay?

[01:52:54]  Bruce: And so what you want is you want to drive down into the checkable forms, and then you want to check those and you want to exhaust them.

[01:53:03]  Unidentified speaker: Okay?

[01:53:03]  Bruce: So the way you would go about this is you would try to come up with these less vague formulations as what it means to beat a grizzly bear and how you would go about doing that. And then you want to go check each of those as best you can. And let's say you keep finding good counter examples and eventually you exhaust them all. And you go, you know what? I think the grizzly bear would win no matter what.

[01:53:26]  Unidentified speaker: Right?

[01:53:26]  Bruce: And that's the way you would reason about this.

[01:53:30]  Peter: Okay.

[01:53:31]  Unidentified speaker: Yeah.

[01:53:31]  Peter: I mean, are they are they men selected randomly on the street or are they combat experienced?

[01:53:37]  Unidentified speaker: You know, there's tons of questions like this.

[01:53:39]  Unidentified speaker: Yeah.

[01:53:39]  Unidentified speaker: Right?

[01:53:40]  Peter: No, that's good.

[01:53:41]  Unidentified speaker: Okay.

[01:53:41]  Unidentified speaker: Yeah, that's actually a good answer.

[01:53:43]  Peter: I wasn't sure how that question would go.

[01:53:44]  Unidentified speaker: But okay.

[01:53:47]  Bruce: So I honestly think you can do this with almost anything.

[01:53:49]  Unidentified speaker: Right?

[01:53:50]  Bruce: Like every theory that you're out there having discussions with people on the internet and you're arguing with people and they're on critical rationalist boards or Bayesians are arguing with critical rationalists or whatever is free will true. The whole secret is, are you willing to drill down and get more specific and then reformulate your theory so that it's got actual content? And if you're willing to do that, you can make progress with almost any question.

[01:54:17]  Unidentified speaker: Okay.

[01:54:19]  Unidentified speaker: Okay.

[01:54:20]  Peter: Well, thank you very much for this, Bruce.

[01:54:23]  Unidentified speaker: I learned a lot.

[01:54:24]  Peter: This is an interesting investigation here. And I hope someone else gets something out of this. So I'll look forward to re-listening and editing.

[01:54:34]  Unidentified speaker: All right.

[01:54:34]  Unidentified speaker: Thank you very much.

[01:54:36]  Unidentified speaker: Okay.

[01:54:36]  Unidentified speaker: Take care.

[01:54:37]  Unidentified speaker: All right. Bye-bye.

[01:54:45]  Peter: Hello again. If you've made it this far, please consider giving us a nice rating on whatever platform you use, or even making a financial contribution through the link provided in the show notes. As you probably know, we are a podcast loosely tied together by the Popper Deutsch Theory of Knowledge. We believe David Deutsch's Four Strands tie everything together. So we discuss science, knowledge, computation, politics, art, and especially the search for artificial general intelligence. Also, please consider connecting with Bruce on X at BNielson01. Also, please consider joining the Facebook group The Many Worlds of David Deutsch, where Bruce and I first started connecting.

[01:55:29]  Unidentified speaker: Thank you.