David Papineau: How Philosophy Serves Science

Are philosophy and science entirely different paradigms for thinking about the world? Or should we think of them as continuous: overlapping in their concerns and complementary in their tools?

David Papineau is a professor at Kings College London and the author of over a dozen books. He’s thought about many topics — consciousness, causation the arrow of time, the interpretation of quantum mechanics — and in all of these he advocates engagement with science. The philosopher should take its cue from our best theories of nature. For example, a philosophical account of causation must pay attention to the way this concept is used in the sciences.

But the philosopher can also be a servant of science. Philosophers are undaunted, excited even, by apparent paradoxes and where such thorny problems pop up in science this is where philosophical tools can be brought to bear. For instance, when quantum mechanics appears to suggest cats are alive and dead, the philosopher’s interest is piqued (even as the physicist’s attention may wane).

David’s website


View full transcript

James Robinson: [00:00:00] David Papineau, thanks for joining me.

Very pleased to be here. Thanks for inviting me.

James Robinson: I wanted to start with a bit of a personal bugbear of mine. I’ve heard of such a thing as the scientific methods. And I don’t know what it is. Can you let me in on the secret?

David Papineau:Yes. When I started doing philosophy, that was the thing in philosophy of science anyway.

There was inductivism and hypothetico-deductivism and falsificationism and then Kuhn’s challenges to these. scientific methodologies and Lakatos’ methodology of scientific research programs. And the idea was that philosophy of science would figure out. The procedures the standards that scientists use in arriving at and accepting their theories and this would be a benefit [00:01:00] to perhaps scientists knew it already, but the world in general to be able to think like scientists.

And that program interestingly has just gone away.

James Robinson: Is it because we found out the answer?

David Papineau: I, I have a view about what happened. And it’s a kind of naturalism, reliabilism. You might talk more about this, but I take the aim of science and. investigation generally to be to get at the truth to arrive at true judgments about the world.

Because if you have true beliefs about the world, then your projects will Succeed. You choose your actions and your policies based on knowledge of how things are going to go. And if you’re, if your beliefs are true, then everything will work as planned. So truth is what we’re after and then I think we can [00:02:00] ask what procedures, what ways of investigating will be well designed to get at the Get at the truth and perfectly good question and there’s plenty of good answers, but they’ll be different in different areas of science.

And the scientists themselves are often the best people to tell you about what are the best methods to find out about the nature of viruses or the sequence of hominid species on the face of the earth. You’ll need very different methods to find out about these things. And the idea that there’s one unified method across the board seems to me.

Seems to me silly, but I think there’s a big philosophical shift behind this move away from worrying about the scientific method, which is

50 years ago when I started doing philosophy, the idea was that we [00:03:00] built up the world from the inside out. We started with Maybe knowledge of our own minds, or at least knowledge of our sensory deliverances and we had to somehow construct an account of the world on that basis using no materials other than what knowledge of our own mental states gave us and perhaps deductive logic.

And So it was a out about it but we don’t want to throw away what we already know about the world in deciding what’s the best wizard of finding out more.

And once you have that more naturalist view that we’re natural beings within a natural world about which we already know quite a lot, then it looks like you will expect to find different ways of getting at the truth in different areas of reality. Yeah, there’s

James Robinson: kind of arrogance, perhaps, reflected in the idea that philosophy needs to restart everything, perhaps.

And I [00:04:00] quite admire the logical empiricist program, which was started where they tried to, build everything up from the ground up. But it just didn’t work.

David Papineau: And it really is the last wave of the kind of Cartesian idealist. Descartes wasn’t idealist, but his way of doing philosophy pretty quickly. motivates idealism that you’re building up from the logical positive is very much part of that program. You have suggested that it’s ambitious.

It’s impressive that logical princess, we’re going to tell scientists how to do things. And there’s plenty of philosophers of science. Then and now, who think it was arrogant, presumptuous of philosophers to feel that they could teach scientists how to do things.

I’m not very much of that mind. I’m not the kind of science worshipper. I think there’s plenty of areas where philosophers can show scientists that they’re doing things wrong. But [00:05:00] we’ll talk about that more later.

James Robinson: Yeah. And I think, to give the logical empiricists their due, I think they were really, they, they were arrogant in some ways, but they were also very respectful of science.

And they thought actually this is the way that philosophy should be done. But I want to come back to the idea of the scientific method, perhaps just one more time. I think there’s something laudable in the idea that there might be a scientific method, or at least in searching for that, in as much as, if you could find that thing, it might be a good explanation of why science makes progress, which seems to be a difference between the sciences and other disciplines, perhaps.

But I also wonder if maybe the explanation for that is just simpler, and it’s just that science deals with the sort of subjects in which we can make progress. It deals [00:06:00] with structures of the physical world where you can build up knowledge, whereas other fields deal with perhaps human affairs and much more fluid states of things.

I don’t know if you have an opinion on that.

David Papineau: Interesting thoughts. It may well be that attractable subject matter is necessary. For the kind of scientific progress that is so striking over the last 300, 400 years. But subject matter isn’t sufficient. Many people across the world, serious sophisticated, literate numerate people in ancient Greece, China, the East very interested in the same subject matters and science didn’t really get off the ground in the way it did in Europe in the 17th century.

And so there’s a very interesting question which in the first instance I think of as [00:07:00] more historical, sociological what was it that led to the last 400 years of science in the Western tradition. And there’s many competing answers. Michael Strevens has just written a book and about, about this, what’s so special about science. And he has a very I’m going to oversimplify a very odd answer, but he says, for better or worse it’s part of scientific culture that if you are going to

have a dispute, you have to settle it by producing new data. And what if you noticed, to philosophers and people in humanity, social sciences generally it’s very odd. If you go to a scientific talk, [00:08:00] Unless the speaker is past 60 or a Nobel prize winner, they aren’t allowed just to theorize.

They have to give you some data. They have to show you what their experiments have come up with and somehow it’s built into scientific culture that, that they have to argue with. New empirical facts and Michael thinks that leads people to spend a lot of time Seeking out empirical facts ones that will be informative or decide between theories but you can’t just reason and speculate and why that became part of scientific culture who knows maybe it’s just A lucky break.

But there’s also the whole

institutions of science are a lot to do with it. In the 17th century, there was a big problem. They, so what one thing that happened with science is nullus inverbia don’t respect authority. Authority has no standing in these matters. [00:09:00] Everything has to be figured out a new, I think that’s clearly part of the, what’s different about modern science and you might think of it as associated with various kinds of Christian Reformation, back to the text itself, don’t listen to the, don’t listen to the clerics.

But at that point you have a problem about, about what are the facts? And the philosopher Daniel Garber once gave a lecture, it was very funny, about how the, 17th century struggled with this issue and he compared Francis Bacon, who at the end of the November War Game says, send me all your facts, send me all your facts.

I’m going to collate them and I’m going to figure out the right theory on the basis of all these facts. But of course, there’s a worry there about who, how good are these facts, who are the people who are sending them, can we trust them and he [00:10:00] compared them with Descartes, who at the end of the Discourse on the Method, doesn’t say send me your facts, he says, send me your money.

So I’ll have enough resources to do experiments myself. And so there’s this tension in early modern science. How do you get well accredited data? And you can think of one part of modern science is a compromise that sure. Nobody has to, it’s not the case.

Everybody has to be like Descartes and do all the experiments themselves. That would be crazy. You have to trust other people but their experiments Have to be replicated is quite a system of accreditation who are reputable scientists whose findings we can at least at first pass take seriously and so that’s a lot of What’s special about science?

But it doesn’t sound like the kind of scientific method that logical empiricists were after does it? No,

James Robinson: I think [00:11:00] if there is such a thing As a scientific method, it’s going to be the sort of thing that you can’t write down very concisely. It’s going to be a collection of lots and lots of practices. Yeah, I think.

And even this idea of having to present new evidence or data. This was one of the things that came out of the whole logical and persist episodes. There’s no clear distinction between data and theory. There’s a kind of continuum. Your data is understood through theory.

It’s collected with instruments that have been devised based on theoretical specifications. So it’s very hard to untangle those things, even in the simplest of cases.

But there does seem value in, even if we can’t find this overarching method, there’s value in, in philosophers looking at what? Science does. Can you maybe speak to that and where philosophy adds some unique [00:12:00] insights or helps science?

David Papineau: So one question Is what’s the difference between philosophy and some philosophers think they’re very different. The science is based on empiricist investigation, whereas philosophy is all a priori and perhaps it’s an analysis of concepts and so on. But That’s not how I think of it at all.

I think, no, analysis of concepts. If that’s all science was, if that’s all philosophy was, I would quit philosophy. That’s very boring and uninteresting. I think philosophy and science are very much in the same line of business. They’re both after substantial, synthetic, informative theories about the real world.

But I think they use different methods or rather. They come in when there’s different problems. Often we are stuck in science. We don’t have enough data to decide between theories. But in other cases, we’re stuck because we’ve got plenty of data, but [00:13:00] our theories are in a tangle. And different assumptions we have lead us to Commits to inconsistent claims and we’re in a tangle.

Something’s gone wrong in our theorizing and we’ve got to somehow figure it out. And that happens, it happens in non scientific areas. Free will, we’ve clearly got free will. We act on the basis of our character and our motives. No, we don’t have free will. Everything is determined.

There’s some kind of. Inconsistency in our thinking, but it happens within science too. Know about this. Quantum mechanics the realities often in in a superpose state there’s no factor of the matter. But when we observe, we get a definite result, but how can we get a definite result?

That means that something other than the Schrodinger equation is [00:14:00] governing the behavior of matter. And how can that be? Isn’t everything made of small particles that are governed by Schrodinger’s equation? So there’s a, there’s an inconsistency there in in, in biology. Darwinian natural selection theory seems to imply that behavior that’s good for your conspecifics and bad for you will never evolve.

There’ll be no altruism in the animal world, but look, here’s lots of experiments, lots of data showing there is altruism in the animal world. And in these cases, it’s not that we’re short of data. It’s that, that our theories lead to the inconsistency, the paradox. And scientists aren’t generally very good at dealing with that.

Quantum mechanics is a classic case. The physicists want to Brushed that difficulty under the carpet and led by bore they succeeded in doing so for nearly a hundred Hundred years. Not all physicists are [00:15:00] like that physicists have philosophical inclinations and they want to unpick the tangle, but most of them say That argument isn’t going to be resolved by no more data, it’s therefore not a scientific question and shut up and calculate.

So in those areas it’s not, I’m not interested in demarcation disputes that I’m a credited philosopher and the scientists aren’t any good at it, but you need to have a philosophical turn of mind to deal with those kind of problems. And I’ve just got very interested when we’re talking about the scientific method in the replication crisis, about replication crisis that, yeah,

James Robinson: so I guess this is the,

the many false positives that you might get.

I suppose if it’s probably best known within psychology and medicine. And the standard is just, you have to have your P values above. 5 percent, which means only a 5 percent chance of your results having been produced by [00:16:00] chance.

But of course, if you run a hundred experiments, you’d expect to get five interesting looking results just by chance.

David Papineau: It should go straight into the journals. So my supervisor was Ian Hacking, and his first book was The Logic of Statistical Inference, and he could see, and so could plenty of philosophers, that the so called logic of significance testing was a load of cobblers, to be honest, and and philosophers have been saying that for quite a while, and it’s surprising in a way to me that it’s taken So long for the chickens of significance testing to come home to roost.

It’s, as you just put it the whole methodology designed. To if we keep testing rubbish hypotheses, it’s designed to commit us to one in every 20. And yeah.

James Robinson: And you don’t publish the problem [00:17:00] is you don’t publish the 95, which the 95 out of 100

David Papineau: that, but do you say produce interesting things?

Problem. But the pump, that’s just a corollary of supposed logic of significance testing. That’s what Fisher and Neyman and Pearson said so long ago. How can we since. Any data are logically consistent with any probabilistic hypothesis. How can we ever know that a probabilistic hypothesis is true?

And they said okay wait until you get some data that would only happen Very rarely, if it weren’t true, and if that’s what happens, then you can take it to be true. That’s what they urged. That’s what significance testing is. And if we if we publish all the results, which of course we should nowadays we can on the internet then you’re not doing significance testing anymore.

You can note how likely or unlikely the data are on various hypotheses, but [00:18:00] you ought then to leave it to good judgment to decide on that basis how seriously you ought to take different hypotheses. The point is obvious. If somebody comes along and says, I’ve done a really quite extensive survey and look, these data about voting patents I’m just trying to think of a crazy hypothesis there’s

James Robinson: classics like, ice cream sales in Italy or something, and,

David Papineau: And you’d only get data like this one time in 20 if this hypothesis weren’t true. And I say Yeah, but that’s still a completely crazy hypothesis.

My credence in it has gone up from one in 10 million. It should go up if you get data that are more likely on that hypothesis than the alternatives, but it’s gone up from one in 10 million to one in 1 million. That’s what’s happened.

James Robinson: So what you’re describing there is a kind of Bayesian updating where I do think that has to be [00:19:00] part of the answer, if not the answer, but there’s an obvious problem with that, which is how do we set our priors?

How do you, for that to get going, you need to have an initial idea of the probability of something. And that’s not always easy, but maybe.

David Papineau: While that raises its own problems. That’s not sufficient reason for saying, Oh, but so we don’t want to be Bayesians because it involves arbitrary prize.

We want to be, we want to be frequentist significance testers because frequentist significance testing is in effect, take any crazy hypothesis but to put to one side that it’s crazy and try and treat it as we treat any other hypothesis in deciding how credibility is given certain data. And that’s just, Fairly absurd.

The truth is, in reasonably developed areas of science, people in those areas have a pretty good sense of which [00:20:00] hypotheses are likely to be true and which aren’t and they have They use that to decide what attitude to take when they see some new study. I’ve got quite a lot of medical friends and I say, you think about this stuff about mad car disease being caused by prions.

Look here, here, these data is strongly supported. And they’ll say, I’m not I’m not so sure. It doesn’t, runs counter to my understanding of how infectious diseases work. They need more convincing than just, a strong p value because they know from the start that the hypothesis is dodgy.

And I don’t think you should try, of course there’s a worry, there’s a standard worry that traditional prejudice white old male doctors have come up with these kind of assumptions that have got no basis in fact at all. But I don’t think one should Throw out all scientific judgment because of that kind of worry.

James Robinson: So coming back to the original point, which was something like [00:21:00] the philosopher being a kind of theoretical scientist without a portfolio. So samurai roaming around the academic landscape, looking for problems where paradoxes arise. This is an attractive, this is an attractive view of things.

And I’m just wondering is this something that’s just confined to science? Do philosophers, can philosophers help with other disciplines as well?

David Papineau: Oh yes. I’m not sure I want to push this kind of samurai for hire. You philosopher’s biology, they work in biology, and then you need to know quite a lot of biology in order to be an effective philosopher of biology and certainly philosopher of physics and yes so this account of philosophy as Trying to solve theoretical contradictions, paradoxes in our thinking. That applies across the board. My point was only that, that these kinds of paradoxes arise within [00:22:00] science as well as elsewhere. But of course they arise elsewhere. They arrive in everyday thought.

All the time. And that’s why we have philosophical puzzles of outside philosophy of science. I mentioned free will, but you could take all kinds of things. What’s the nature of, Persons, philosophers say persons, I don’t know why, what’s the nature of people, what kind of things are people what’s morality it’s easy enough to show that we have strains in our thinking that, that lead us into conflicting ideas in all those areas and philosophy’s job is to figure out, common sense plus science kind of leads us into.

theoretical puzzlement and something is wrong in our thinking and we’ve got to figure out what it is. That’s the job of philosophy.

James Robinson: And I’m wondering as well, when things stop being philosophy. So the interpretation of quantum mechanics is a good example. I also think of other interpretational [00:23:00] problems within physics, like understanding the nature of space and time, which as you point out are commonly picked up by philosophers, but there are physicists who work on these.

I feel often the reason it’s philosophers who work on them is it’s just, they’re the ones who are permitted to. So yes, David Wallace, my teacher at Oxford was. Initially a physicist or his first PhD was in, in physics. And then he did a PhD in philosophy because he just realized he wouldn’t be able to work on the particular physics that he wanted to within a physics department.

And I wonder if these problems of paradox and things once, at a certain point, they Once they’re resolved, I suppose they just become subsumed within the field. They’re no longer part of philosophy. So if everyone accepts, as I think they probably should, and I think you’ll agree, the Everett interpretation of quantum mechanics, so the many worlds interpretation, [00:24:00] That would no longer be something that philosophers thought about.

It would be taught in physics textbooks and be something that physicists think about and use to understand the work that they’re doing. Does that sound right or do these things just never get resolved?

David Papineau: I mean there’s an issue about philosophy doesn’t seem to settle things, the, the way that things get settled in science. And I’m just wondering whether even the,

the multiverse interpretation, Everettian interpretation, I’m hesitating to say many worlds because it’s the vulgar version of every, but let’s just say many worlds. I don’t think it will become to be all cut and dried and settled probably because it conflicts with common sense.

And so new generations of Students, young people coming up will, will struggle with how to make their ordinary, [00:25:00] naive everyday view of the world consistent with Everettianism. It might be all settled for the physicists, but they’re still going to be awkward. Awkward problems.

I have no very clear views on this. Why? I’ve been doing philosophy long enough that I’m now getting a sense within philosophy. Let alone within the wider world that problems don’t get settled because we sorted this out 40 years ago and now all young people are coming back and want to sort it out again.

And and yeah. I’m not sure I regard this as an unhealthy thing. I think that, and I’m thinking out of the top of my head, because I haven’t really thought very hard about this, but I think there’s something about this tangle business, this, what’s particularly about phosphate, that means that people do need to Re engage with the issue, every so often, you need to do it for [00:26:00] yourself.

Partly because, issues about everything is physical, there’s nothing over and above the physical world. Where does that leave the everyday world? It’s, metaphysically necessitated by the physical world when we have kind of reduction and supervenience and now there’s a whole new wave of tools for thinking about reduction and supervenience and I think that part of what goes on in philosophy is that if you look back 40 years the way people were thinking about it strikes the new generation as not quite the right way premised on certain assumptions that have now disappeared and we’ve got to we’ve got to do it again look

So the world is bifurcating all over the place all the time. Don’t worry about the flabbergasting. Multiplicity of the bifurcation. Just, just focus on one, right? So you are going to turn into these two of you in a second or so. So here’s a problem. [00:27:00] And when you talk about yourself and now you are gonna be.

Two what are you referring to? What kind of thing is a person? And, if you’ve done this personal identity business, you’ll realize that there really aren’t any good answers. Derek Parfit thought pretty much that when you split, you cease to exist. He thought it didn’t matter too much, provided there was continuity between you and your various descendants.

But the thing you were referring to as you now ceases to exist when you split. And David Lewis says, no they’re two of you and you’re, they overlap in you now. And I prefer a yet third theory, John Perry’s theory that when you talk about yourself now, you’re referring to the Hydra.

that consists of all your future selves. But in fact, most philosophers working on [00:28:00] Everettianism don’t really engage with this issue. I think if Everettianism becomes generally accepted, philosophers are going to have to think about it quite hard. But As I said, it’s the kind of thing that needs to be redone each generation.

So that’s why. In some ways,

James Robinson: the Everett interpretation just makes sharper some problems that were already there in philosophy. Parfitt had thought experiments about putting people through splitters and transporters and all sorts. And he didn’t really care about whether those things were real or not, but I suppose it provides extra motivation for thinking about those problems.

I would say one other

David Papineau: thing on it, I think. It’s all very well for him to say if you split you cease to exist, but there’s continuity and that’s what matters. If it’s a thought experiment about something that might happen [00:29:00] Occasionally now and then, but if it’s happening constantly all the time, the idea that, you don’t exist ever for more than a millisecond that looks like you’ve got to start saying something more than that.

Yeah, I

James Robinson: think that’s fair. Here’s another take on. Maybe what’s going on philosophy in terms of it, working on things that, that never seem to get resolved. You’ve talked about philosophy being continuous with the sciences and other disciplines, but if we think more. precisely about the places at which it’s continuous.

So where it meets those meets the sciences tends to be at foundational issues. I remember Harvey Brown commenting that, if the philosophers of physics get stuck on all the terms on the first page of a physics textbook, frame of reference, they’re like, Oh, what’s that? And that never gets explained anywhere else in the physics book.

It’s just taken as read.[00:30:00] And so philosophy is well, physics is a good example because, physics was natural philosophy. There was a kind of pre theoretic stage of physics where people were struggling to find the right ways of thinking about things. And I wonder if what philosophy does is just lets us revisit that.

The assumptions and the models of the pre theoretic stage so it takes us back to the foundations and of course those things are always up for some kind of interpretation or maybe even questioning.

David Papineau: When I said that the philosophical issues need to be revisited. I didn’t mean to be denying.

I think there’s two thoughts here. The thought that you aired earlier that that when some of these theoretical tangles are resolved, they become part of established science. They cease to be philosophy anymore. And I think that’s right for certain problems. If [00:31:00] you think about Newton versus Leibniz about.

Absolute space is zero. Absolute rest. I take it that’s resolved. We’ve got Neo Newtonian space time. We can understand this all. Nobody wants to go back to that. So certainly the problem’s moved on. Exactly. My thought was that, but there will always be problems at the interface of these puzzling scientific views and common sense.

And for that reason, people are going to have to keep revisiting them just to deal with the issues that come up at that. interface. I wonder if that’s just some of the foundational problems within science are horribly long persisting. A lot of space and time. Is resolved with the move to, to special generativity or even before, though there’s other problems there.

But think about the other thing David Wallace works on a lot the asymmetry of thermodynamics. And I’ve been working on that a lot because I’ve been working [00:32:00] on the asymmetry of causation, which I think of as very closely related. And it’s a complete nightmare, the asymmetry of thermodynamics. People have been banging their heads against this for, what, 150 years now.

And and not only does in philosophy, there’s plenty of different views, people just don’t agree. It’s almost within statistical mechanics, there’s all kinds of different approaches. There’s no clearly agreed. Foundations, there’s different ways of approaching statistical mechanics.

James Robinson: Maybe I don’t think this is something we’ve discussed before on the podcast.

So perhaps we can give a little bit of detail on this issue. My gloss is something like, the laws of physics as best we know them are symmetric in time. Yeah. At least in the sort of. Most foundational laws, thermodynamics.

David Papineau: The basic dynamic laws. The

James Robinson: basic dynamic laws.

You can think of, yeah, [00:33:00] billiard balls hitting each other. You can rewind that video. It still looks absolutely physical and normal.

David Papineau: The way the movie backwards what happens is still in accord with the basic requirements of

James Robinson: dynamics. Exactly. And all our kind of most fundamental laws. Or exactly like that.

They all, you can reverse the arrow of time in them and it’s the same equation. But when we look at stuff in the world, it doesn’t look like that at all. And it’s very obvious that there’s a a sort of one way nature to the way that things happen. Yeah. And I suppose that, the explanation that.

I’m sold on this is to do with the past having started in a very special state that explains why all that happens, because obviously what is not just a product of physical laws, but it’s a product of physical laws acting on things in the world and things were set off, our universe began in such a rare state that it’s progressing.

It’s bound to progress in such a way [00:34:00] that there appears to be an arrow of time. It’s going to take a lot more cashing out to explain what that means. I don’t know if you want to, I don’t know. So why

David Papineau: did I say it’s still a mess? I think there’s probably consensus that the

asymmetries, the asymmetries we see in time are an upshot. of the fact that down one temporal end of the universe is a very low entropy state. And and not, no reason to suppose it’s so at the other end of the universe. And that’s why we call that in the universe the past, the origin. So there’s consensus about that, but exactly then how it works out, because standard view is that The

laws of thermodynamics, which are extra to the basic Dynamical laws, yeah. [00:35:00] Dynamical laws are probabilistic in nature. That if you have a isolated system in a low entropy state, it will probably very probably increase in entropy. And you need some statistical postulate. to add to the low entropy initial condition.

And then there’s horrible questions about what’s the nature, what’s the truth maker of that statistical postulate. There’s a strong, powerful thought that, that reality consists of what actually happens. And and therefore there shouldn’t be any kind of, that doesn’t include the probabilities.

So where do the probabilities come from? Yeah, it’s probability is very strange thing. And so a lot of the puzzles about thermodynamics come from that element of it. And and you’ve somehow got to tie together

probabilistic aspects of [00:36:00] thermodynamics with this low entropy initial condition. So we have a, an effectively isolated system in a low entropy state and you can say Nearly all the ways in which it might be in low entropy state will be ones in which later on you’ll get a high entropy state just because there’s an awful lot more high entropy states than low entropy states.

But that thought tells you that this isolated system in low entropy state almost certainly came from a higher entropy state as well as is about to evolve into it, and that’s not what we want. We want it to be low entropy increasing all the way along. And so somehow you’ve got to tie up the initial condition with the probabilistic assumptions you apply to particular systems.

It’s very difficult. I’ve been reading a lot of David Wallace papers about this, and it’s very interesting.[00:37:00] He’s taken issue, David Albert, along with Barry Lower have developed a way of thinking about thermodynamic asymmetry and David Wallace is rather sceptical that they brush over a lot of details that are important.

James Robinson: Is this a case actually that we’re just going to continue revisiting? Or is it the sort of thing where we can get a consensus because some argument is so good or some evidence comes along that changes the state of things? I can’t really see there being any new evidence to add to this, unless there’s a

fundamental change in physics.

David Papineau: No, we’re not.

I wish I knew more of the,

the theories within physics. But my understanding is that, Even people who theorize in physics aren’t really in agreement about how to think about these matters. There’s various effective [00:38:00] mathematical techniques that you can apply to, to analyze and predict how systems out of equilibrium will move towards equilibrium.

But they’re pretty kind of Nancy Cartwright y. They’re ad hoc. They’re mathematical techniques. We know they work. They give us the right predictions. But exactly what’s the underlying basis for them? People aren’t sure. And at the theoretical level, there are different approaches within statistical mechanics.

There are people who use the Gibbs framework. There are other people who think the Boltzmann framework is better. There are people who say they aren’t really different, these two frameworks. So while they’re there, effective techniques used by more practical scientists within the science. There’s no clear agreement at all about the foundational assumptions you need to make in thermodynamics.

James Robinson: Yeah. Interesting. All up for grabs. Another thing that just struck me recently was that it makes sense [00:39:00] to talk of the physics of biology. The physics of chemistry, in fact those fields exist, you have like biophysics and quantum chemistry and things, but one doesn’t really talk about the chemistry of physics.

But with this framework, the something of something, you can almost say the philosophy of X. So philosophy is this reflective thing there are clearly cases where it applies better because there’s still open questions like the case of the arrow of time in physics.

I’m wondering if it also allows us to look at any discipline and open Pandora’s box and say, okay, I don’t know. The philosophy of law, the philosophy of engineering, right? You can ask questions which you can’t ask within those fields. Okay.

David Papineau: So philosophy of law it’s a pretty weird area actually because it’s often done by lawyers and they have special training anyway, but philosophy of law is a very live.

area full of [00:40:00] problems, difficult. Many people work on it. Philosophy of engineering, not so much. Maybe it should

James Robinson: be. I think there should be a philosophy of software engineering. I don’t think there’s anything, I

David Papineau: mean, that things differ in this way. Some subjects are more philosophically puzzling than others.

Philosophy of chemistry. I have friends who work in philosophy of chemistry, but. There aren’t many of them. There’s not a lot to do there. Philosophy of engineering, certainly, I know one person works in philosophy engineering and yeah, not a very exciting area if you

James Robinson: ask me.

And why is that? Why do you think philosophy then tends to look at the, is it because philosophy is at the bottom of the stack and physics is at the bottom of the science stack as it were? Yeah it’s,

David Papineau: as I said, it’s just that philosophy comes in when we have some nasty intractable puzzle.

There, look so here we are. The basic laws of dynamics are time symmetric. The physical world is not time symmetric. That’s a puzzle. It ought to be time symmetric, but [00:41:00] look, it isn’t. And that’s puzzling. There’s nothing very puzzling in engineering. I’m just trying to think, is there philosophic, can we think of an engineering philosophical?

James Robinson: I don’t know. You’ll have to ask your friend.

David Papineau: Yes.

James Robinson: What do you think are going to be the fruitful subjects for philosophy over the next decades?

David Papineau: Ooh, that’s a hard one. That’s a hard one. It’s not so hard because there’s one obvious one right in face artificial intelligence.


Already? It’s already huge amounts of philosophy going on about. Artificial intelligence and every philosophy department is trying to hire artificial intelligence philosopher and they aren’t enough to go around, to be honest, but no it’s the hot, it’s the hot area and I

James Robinson: mean, it is something that, philosophers have thought about artificial intelligence for some time. The Turing [00:42:00] test, Turing wasn’t a philosopher exactly, but I think we can think of that as a piece of philosophy, what he was doing in trying to define what what it is that makes a, or when we should say that machine is intelligent.

And I think people have now probably casting around for a new Turing test .

David Papineau: Because GPT 4 clearly passes the Turing test. Yeah. And I that’s quickly becoming. A real social problem that you’re talking to a device, you don’t know if it’s a human or not. That’s what it is to pass the Turing test.

Yeah. It’s going on all the time already. I mean it, it shows that Turing test is not a very good test either for intelligence or consciousness. I see to very few people in my admiration for GPT 4, and I’m very ready to think that artificial machines Can be conscious and probably will soon, but I don’t think GPT 4 is, and I don’t think it’s particularly intelligent either, though it’s very impressive.

James Robinson: I think it is [00:43:00] exciting because we have seen emergence of particular abilities with these LLMs. We’re starting to actually get some data, which maybe philosophy can use particularly in the realms of complexity and emergence, of course. But no one is able to say if consciousness will emerge. I’m with you. I see no reason that machines can’t be conscious. But we still have very little understanding of what it would take to produce consciousness in a machine. Is this a purely scientific problem or is this really at the coalface of philosophy?

David Papineau: I think this is clearly a philosophical issue. This is another case where you might feel philosophers can guide the scientists. There are many people who think consciousness is a scientific problem. And there are many reputable scientists who [00:44:00] work on consciousness and develop theories of consciousness.

And I’m afraid, I myself think the whole program is misplaced. I don’t think, sometimes I say people think consciousness is a thing. They’ve got to realize consciousness isn’t a thing. Partly, is it a thing in the slang sense? It’s become a thing in science, but it oughtn’t to be.

But also, people are too inclined to reify consciousness. And I think a lot of what’s going on is that

pretty much everybody has an overwhelming inclination to think of consciousness as something separate, extra to the physical world. Overwhelming inclination to, to intuitively think in terms of dualism. And so they think there’s this extra stuff that arises in certain places. And the job of the scientist is to figure out.

where it arises. And I think that whole program is [00:45:00] misguided. There’s no extra stuff. There’s just a whole lot of physical processes. And because of that, I think that when it comes to other animals and even more so artificial intelligences, it’s not that we find it hard to know if they’re conscious or not.

It’s that there’s no fact of the matter about whether they’re conscious or not. I think our notion of consciousness is grounded in introspective awareness of many of our mental states. And in effect, the notion of consciousness is states that are like these. And now we as it were, point to all the states that we are introspectively aware of when we have them.

And if that’s what the notion of consciousness is, like these in what respects? And [00:46:00] and then you suddenly get people. arguing about whether a system that was like us, structurally behaved in rather the same way, but didn’t have biological wetware inside, would that system be conscious?

It’s got states that are like our states in terms of their causal interactive structure, not like our states in terms of the way they’re, Biologically realized and I don’t think there’s a fact of the matter about whether those systems are conscious. I do think there’s a fact of the matter about how we ought to engage with them.

Are they deserving of moral respect and so on. But I think that comes first, the question of whether they’re conscious.

James Robinson: It’s interesting that there’s one thing which I really can’t doubt, which is my own consciousness. So if I can say that there’s a factor of the matter about one thing, it’s that [00:47:00] I’m conscious and it feels odd to admit that perhaps there is no test that can be done that would adjudicate where whether someone else is.

It is conscious and and that, and that aligns with the, with your idea that there is no fact of the matter

David Papineau: about that. One doesn’t want to move too quickly from there’s no test that will tell us the answer to there’s no fact of the matter. So it’s not just that there’s no test. It’s that think about the, how we give meaning to the term conscious.

Do we do it in a way that will draw a line? in fancy artificial intelligent machines or in octopuses as to whether they’re conscious or not. Do they have states that are like these ones, the ones that we know are conscious in us, [00:48:00] and they’re like ours in some ways and not like ours in other ways?

And Peter Caruthers has a nice analogy. He says, imagine I say that This is a nice neighborhood, not thinking very much about what nice neighborhood means, right? There’s a number of respects and now here’s another, and I say, look, there’s another neighborhood that’s a nice neighborhood too.

And you say, no, it’s not. And I say look, it’s got, the streets are clean. And you say, yeah but look a lot of the houses aren’t really well painted. And now we start arguing about. which of those it’s the right criterion for being a nice neighborhood and clearly there’s no factor there’s no fact of the matter it’s a loose concept that’s what i want to say about the concept conscious of course that sounds wrong but that sounds wrong because it’s so natural for us to think that consciousness is some kind of extra Aura, inner light, cloud of stuff that arises in [00:49:00] certain systems and not others.

But that’s a wrong way to think. I think,

James Robinson: yeah, I mean it comes, I’m remembering, was it Bertrand Russell who said, just because I have a beetle in my box, doesn’t mean you have a beetle in your box. But what you’re saying is something different, which is that

David Papineau: It was Wittgenstein.

Wittgenstein, was it? Yeah. You did too much physics and not enough Boston style philosophy.

James Robinson: But what you’re saying is slightly different because, saying there’s a beetle in a box or not is still talking about facts of the matter. And you’re saying no, it’s just not a question of there being, whether there’s a beetle or not, it’s a value judgment or it’s similar to a value judgment, at least if we buy into the analogy of the neighborhoods.

I just don’t know if that’s right. So

David Papineau: in the philosophy of mind, philosophy of cognitive science is a long standing debate about do mental states generally pain say, but consciousness [00:50:00] more. Specifically, does it depend on the causal structure of what’s going on in your brain at some level abstracting from the detailed physiology, physics of your brain, or does it depend also on the way that causal structure is realized?

And people talk about role and realizer views about pains or consciousness generally. And,

and it’s a longstanding and irresoluble philosophical dispute. Okay it’s got quite definite consequences. So there’s. Commander Data, do you watch Star Trek The New Generation at all? No. He’s an android. He’s an android. He behaves, looks just like a human being. He’s slightly funny and cold in some ways.

Yeah. I know

James Robinson: he’s quite humorless. Humorless,

David Papineau: like Commander Data. And now That role realizer issue[00:51:00] is the issue where the commander data is conscious. Is he conscious because he behaves and talks and walks just like us? Or is he not conscious because he’s made of silicon chips? And my view is there’s no fact of the matter.

This is a Indeterminate issue. But I don’t think there’s no fact to the matter about Commander Data’s moral standing. It seems to be pretty obvious that he, if, if he walks and talks like us, he deserves to be treated morally, just like one of us. And the idea that, no, you oughtn’t to treat him morally because he’s not conscious.

It’s a bit like, Descartes saying you can be nasty to animals because they’re not conscious. And In fact, it’s worse. Okay, so I think the issue of how something like that should be treated should be addressed independently of prior to the question of whether they’re conscious and I also [00:52:00] think this is all in fact very Wittgensteinian and driven by the idea that, there aren’t really Beatles in these boxes, even if people say so that

I’m sure that we’d come to think and treat Commander Data as conscious because we’d included him in our moral circle. He’s somebody whose moral interest is deserving of consideration as ours. Look, this is going to become a live issue. There are going to be, very soon actually, there are going to be Large language models, which are more than large language models, and are designed to be social companions to people.

It’s already happening, but, they were doing it before there were good large language models. I’m sure it’s happening very fast. And if you ask me if there’s some aspects of artificial intelligence that should be regulated, it’s the development of social companions. It’s going to be weird [00:53:00] and strange.

And I think quite soon, many people are going to think that these entities need, their welfare needs to be taken into account. Certainly people will start thinking that about their own social companions, but they’ll start thinking about other people’s social companions or, and they’ll start thinking it about systems that are used as slaves and and then there’ll be other people.

Who say no, all these things are just machines. You people are crazy to be worrying about their welfare. And we’re going to have to address this issue. Do we need to be concerned of, about the welfare of these entities? And at that point, it seems to me, ah, but first of all, we have to figure out if they’re conscious or not.

I think that’s just going to be the wrong way to go. And I hope that people will see that’s just a, that’s not going to help to start. Let’s wait for the scientists [00:54:00] to come up with a good theory of consciousness and it will help us to solve this issue. That seems to be the bad thought.

James Robinson: Yeah, I think that’s fair enough.

I do wonder though if one takes the alternate view that there is such a thing. as consciousness, which could be correct, right? That we can actually start, could this be a place where you can do experiments, right? You could say, Oh, let’s try the Mary, the color scientist experiment.

You can actually try that.

David Papineau: So look, hang on. I don’t want to say okay, this is it. I didn’t want to say there’s no consciousness. I never said that you wanted to say, look, in your own case, one thing I’m sure you’re going. In fact, there are philosophers. Illusionists who say, look, it’s an illusion to suppose there’s.

Consciousness in truth, my position isn’t very different from theirs, but I don’t think I’m forced to say there’s no consciousness. What’s

James Robinson: the difference between saying

David Papineau: this? Yeah,

I don’t think there are no good neighborhoods and no bad neighborhoods. I think [00:55:00] it’s a loose concept, but it works as well. It works well enough for everyday purposes in thinking about neighborhoods. And, sometimes people are conscious, sometimes people aren’t conscious. So I think it’s a perfectly useful notion, but I don’t think it refers to a particularly determinant. property. It’s hard to know how to put the, there’s a definite thing that science could identify and discover the nature of. I don’t want to think of consciousness like that. So if that makes me an illusionist, then okay I’m with the illusionist but I don’t think I’m forced to say there’s no consciousness about it.

Yeah. Our conscious lives are terribly important. When it comes to. To human morality, how things are going consciously for a human is the main part of it. And I don’t want to deny that.

James Robinson: Yeah. I think, I certainly agree that there’s no such thing as a binary, conscious is not binary.

It’s not a [00:56:00] light bulb, which switched on at some point in our evolution. That makes it even more interesting in some ways, particularly if with these AI topics, if we accept that it’s an emergent thing. Of some sorts if we can get some kind of view on where it comes out of the pre conscious void, I don’t know how to call it that would be fascinating.

And I think there’s another reason for being interested in consciousness in AI aside from the moral obligations. And I think you’re right, like for certain purposes, we should just put consciousness aside and ignore it. Stuart Russell has also said, we shouldn’t care whether AI is conscious or not, if it’s something that can kill us all.

 Who caress , right? Yeah. What its kind of feelings are. But on the other hand I do think there is additional risk with consciousness because it just complicates the whole thing so much more when something starts to reflect on [00:57:00] itself. And I do take that as being a part of consciousness.

James Robinson: It becomes even more difficult to debug . If we look at LMMs already, it’s so different to any kind of computer program that you might have. And I’ve spent a lot of time trying. Did he buy computer programs, but good luck trying to understand, trying to trace out exactly why chat GPT has produced a certain response for you.

Yeah, and consciousness will just add an extra layer of unpredictability in my opinion. And actually that may be the best marker of looking for it.

David Papineau: But you think of. Consciousness as something

that will make a difference to how the system behaves, you say that, over

and above how the

structure of connections within the system leads it to [00:58:00] behave? No, I

James Robinson: think It’s a way of categorizing that structure of connections, right? It’s something that appears once you hit a certain kind of complexity which we don’t know how to define yet.

David Papineau: Okay. Yeah. I

think you need to watch it. I, that you’re still thinking in a dualist way. There’s some special extra thing that

James Robinson: I don’t think it is. I think it’s the same as, arguably, one of the GBTs at some point figured out how to do maths, right? And it was able to solve novel arithmetic problems.

Not maths in general, but it can do arithmetic. And that was just an ability that emerged. And it’s not, that’s not a new, and I think of consciousness in exactly the same thing. If we build complex enough things within them, the ability to reflect on themselves will emerge.

And that’s not a property. I think just looking at the way that the models work at the [00:59:00] moment, I don’t believe that is a property that they have this sort of a single pass circuit. I don’t think it even structurally allows for that at the moment, but I’m not entirely sure to be honest.

David Papineau: Large language models are very simple.

And so they, certain things might start that weren’t designed into them emerging. But I think a lot of the stuff that we’re looking at we’ll need to have more complex machines with a different dimensions to their design. But think about the theories of. What makes for consciousness in humans and maybe other high level animals.

So there’s some theories that say you’re conscious of what’s in the global workspace. So we have a bit of the system that processes various bits of information and then distributes and makes them available to all other bits of [01:00:00] the system. And then there’s theories that say consciousness comes with tension that when the system devotes all its processing resources to one issue then it’s conscious of that issue.

There are systems that say consciousness comes with certain kinds of

resonance, certain kinds of feedback loops between perception and other things. So there’s in, in the business of scientific theories of consciousness, there are different accounts of which bits of our cognitive processing constitute consciousness. Now, all these bits of cognitive processing Make a difference to what the system can do. But I don’t see that. If somehow, and I don’t see how this would happen we have evidence that one of those systems, there’s supposed to be incompatible theories about. What constitutes consciousness? One of those [01:01:00] systems, that’s the conscious bit, and it’s going to make any difference to what it does.

That’s why I’m suspicious of the whole program, as if there’s a question that will have some explanatory significance. Yeah.

James Robinson: I think that’s enough.

David Papineau: All these overlapping processing powers constitute consciousness.

James Robinson: Yeah, I think that’s right. It may be more an issue of by the time that thing emerges, it’s already so complicated that it’s possibly beyond our control, maybe not but certainly behaving in a way that’s very difficult to understand.

Yes. So maybe it’s more of a way marker rather than a a phase change. I don’t know.

David Papineau: Yeah. I’m very interested in what large language models show us about the nature of human intelligence and consciousness [01:02:00] because it seems to me we’re very analogous. We do all this talking but if you try and figure out what gives rise to it, what’s going on. It’s inside our cortex that accounts for this.

It’s very difficult to explain. In fact neuroscientific theories don’t even really start getting to grips with where is thought, how is thought being produced and Expressed in the human mind, in the human brain. Yeah.

James Robinson: Yeah. There there’s a lot of mysteries, a lot of mysteries to pick up. Yeah, I know you, you are on sabbatical right now. Perhaps this is my final question. I’m just curious what you’re working

David Papineau: on. I’m working on the moment on. causation. And I’m

still got one paper to get out the way. And then I hope I’ve got a book. I’ve done lots of papers. I, nearly all the bits are ready to go. I hope I’m going to be able to [01:03:00] write it straight out. And this is another case where science and philosophy, as I see it, need to get together.

But in fact, in this case, it’s not so much that the philosophers have some stuff to bring to the scientific table. It’s the other way around.

I have some background in mathematical statistics, so I’ve always known. In fact, in my first job was in a, before I had a job in a philosophy department, I was teaching philosophy of social scientists, social science to sociology students. So I’ve always known about the business, what’s now called causal inference Bayesian networks, stuff that developed by computer scientists in the 90s, Judea Poole being a leading figure.

I’ve always known about the business of taking in a pile of correlations and [01:04:00] conditional correlations and figuring out on the basis of those what the causal structure is. There are problems and issues with that kind of program, but it’s clearly

effective. Look, take a little example private schools, better exam results correlation between private schools and exam results. Prima facie indication that private schools cause better exam results. But then we do a bit more statistical analysis and we discover that When we separate the population into rich parents and poor parents, private schools don’t make any more difference.

Within the rich parents, the private schools don’t include, within the poor parents, the private schools don’t include the exam results. It’s all due to home background. So the two things are joint effects of the common cause of parental income. Very natural thought. A lot of, social science, epidemiology, [01:05:00] marketing research, people use that kind of technique all the time.

Philosophers have lots of theories of causation, kind of actual theories, powers theories, process theories regularity theories. None of them Give any account at all of why those inferences work. It’s something of a scandal. Philosophical theories of causation do not engage with the large scale scientific business of causal inference.

James Robinson: I think that’s true. although I have to say the kind of counterfactual theories of causation, you can see how that is relevant to, to this case or how they could touch. They can work with within this kind of a system because you say, look if the parent had been poor , private school wouldn’t have mattered.

So there’s what you’re doing with the blocker or the the additional piece of information that actually wants you control for,

David Papineau: This. I wouldn’t deny that [01:06:00] there’s a close link between counterfactual and causal claims, but. So David Lewis is a person who pushed counterfactual theories of causation.

I think it’s got it completely the wrong way around. I think that we want a causal theory of counterfactuals that once we causation on the table, then it’s easy to explain counterfacts. That’s why you think that if this kid who’s at a state school had been sent to a private school, they would have got better exam results.

No, that’s not true. The reason we make that. counterfactual judgment is because we know the causal structure. Yeah, I

James Robinson: think that’s, I think that’s right. Yeah. I guess Lewis’s program was another another many worlds program, nothing to do with Everett, but he was saying, yeah, what you’re doing when you talk about causes is you’re talking about additional possible worlds.

And it was never completely clear with Lewis, if these were real worlds or not,

David Papineau: at least. He thought they were real. [01:07:00] And, it was a bit of a glitch in his program. Yes. On the face of it, he was making causal claims depend on how things were in this other space of alternative realities. But you might think that causation is something that does its work in this actual world.

How can facts that are constituted by what’s going on in other worlds make a difference in this actual world? And Lewis didn’t. His followers have said things about that issue, but that was a puzzle for Lewis. But as I see it, the way to do this is to, what comes first are the statistical patterns used by causal inference people.

What comes next is the causal structure in this world. And what comes after that is what you want to say in terms of counterfactual language. That’s the direction to do the analysis. [01:08:00] But it is counter to how most philosophers do it. But most philosophers, I said, they really don’t engage.

Yeah, there’s connections between causation and counterfactuals. But I would challenge you to start with a counterfactual theory of causation and explain why you can infer. Absence of causation from the disappearance of a correlation when you control for something else. Yeah I, maybe it’s a challenge that can be met, but nobody’s even, nobody in mainstream metaphysical philosophy so much as asks the question.

Why do the Bayesian network causal inference techniques work? They don’t even ask it. Anyway, so that’s what my book’s going to be about. And it will relate to the asymmetry of thermodynamics as well. I have a theory of causation that I hope will happily explain why causation is asymmetric in time. My causes always come [01:09:00] before their effects.

James Robinson: That’s something to look out for. That sounds really fascinating.

David Papineau: That’s why I’ve been reading lots of David Wallace on the foundations of statistical mechanics. I’m giving a talk in Stirling in a couple of months, which is supposed to, I haven’t done it yet. I’m still thinking about it, showing how close an analogy is between the temporal asymmetry of causation and temporal asymmetry of

James Robinson: I’m not aware of accounts of causation, compelling ones at least where they show why there is a temporal arrow.

It’s not clear. That’s certainly something that is not apparent in the counterfactual Ah. Explanations. What does Lewis say about that?

David Papineau: Ah. So no, but Lewis, when he first did the counterfactual theory, the very first paper, took it for

James Robinson: granted. He thought [01:10:00]

David Papineau: that a natural measure of similarity between possible worlds would show you that if I jump out the window.

If I had jumped out the window, I would have fallen to the ground, is true. But if I had jumped out the window, I would have put a parachute on first. That’s false. He wanted his counterfactuals to be asymmetric in time. But he came under pressure quite soon after the original causation paper from people saying, come on, show us why the counterfactuals are asymmetric in time.

And being Lewis, he then sat down and thought very hard. and amazingly came up with something which is very close to a kind of thermodynamic asymmetry to explain why counterfactuals are asymmetric in time. He, counterfactuals are supposed to come first, causation is going to come later. He had to explain whether [01:11:00] counterfactuals are asymmetric in time and he appealed to the fact that The present needs many traces of present events in the future, right?

There’s a million records of the fact that the Battle of Waterloo happened in 1815.

But supposing there’s going to be a battle of whatever, Battle of Volgograd in, in 2025, there may be facts around now. That determine that’s gonna happen. But they aren’t as rife as the facts that, as it were retrospectively determined the Battle of Waterloo happened. So Lewis talked about the asymmetry of over determination.

The present vastly over determines the past, but doesn’t over overdetermine the future. It might just determine it, but it doesn’t overdetermine it. And he said that’s why [01:12:00] it’s right to think that changing the present. Can change the future. They’re not the part, but it can’t change the past because there’s a million other things that you aren’t wiggling that already fix what the past was like.

So he pulled out of the air this, what I think of as a pretty thermodynamic fact that the asymmetry of traces, asymmetry of records and use that to explain asymmetry account of actuals. So he did address this asymmetry question, but only

James Robinson: later on. Okay. Okay. And it sounds like thermodynamics is going to be key here as

David Papineau: well.

Yeah. Yeah, exactly. Okay. Brilliant.

James Robinson: Yeah I think

David Papineau: we better stop. We better stop.

James Robinson: If anything just because my brain is exploding, there’s so much things to think through here.. Thanks so much, David. I don’t know if you have a final. comment or mic drop. I think the looking out for your book is certainly something

David Papineau: to it’s still got to be, it’s still got to be written, but I hope we talked for a bit about all the things we [01:13:00] might cover.

And I think we probably have covered them all. I think so. Yeah.

James Robinson: And more things as well.

David Papineau: Yeah. Brilliant.

Leave a Comment