Hi, and welcome back to The HPS Podcast, the place for discussing all things, history, philosophy, and social studies of science. I am Samara Greenwood, and my very special guest today is Hasok Chang. I'm sure for many listeners, Hasok needs no introduction as he has been a leading scholar in HPS for many years.
Hasok is the Hans Rousing Professor of History and Philosophy of Science at the University of Cambridge. In looking up his profile online for this introduction, I came across this little gem. “Professor Chang's research focuses on taking the most obvious items of scientific knowledge and asking how we came to know such things. Usually, such a line of questioning reveals that even the most mundane piece of knowledge was hard won through the most challenging and fascinating investigations and debates.”
As well as being well known for his books, including Inventing Temperature, Is Water H2O?, Realism for Realistic People, and his upcoming work, How Does a Battery Work?, Hasok has also taken a leading role in the discipline, particularly as a founding member of both the Committee for Integrated HPS and the Society for Philosophy of Science in Practice, both of which have previously been mentioned on the podcast. As an aside, I've been lucky enough to present at conferences run by both organizations, and they were truly wonderful experiences, and Hasok was always a very generous presence.
In today's episode, we discuss Hasok’s notion of ‘epistemic iteration’, the idea that we do not often start our inquiries from a solid foundation. Rather, we knowingly start from an imperfect position and then use the outcomes of our inquiry to refine and correct that original starting point. In Realism for Realistic People, Hasok quotes Pearce as saying, inquiry “is not standing on the bedrock of fact. It is walking upon a bog, and we can only say, this ground seems to hold for the present. Here I will stay till it begins to give way.”
Samara Greenwood: Hi Hasok and welcome to The HPS Podcast. It is great to have you as our guest today.
[00:02:04] Hasok Chang: Hi Samara. Thank you for having me on the podcast.
[00:02:07] Samara Greenwood: First, I would love to know how you found your way to history and philosophy of science.
[00:02:11] Hasok Chang: I was an undergraduate student of physics at Caltech in the States. I was very gung-ho about physics, particularly theoretical physics. I mean, yes, I was going to be the next Feynman, or whatever.
When I went to university to start studying physics seriously, what I found was that, well, basically physics wasn't what I imagined it to be, and everything I was excited about, my physics professors would say, ‘that's a philosophical question.’ Meaning, ‘don't ask, don't worry about this stuff, do your problem sets, and maybe after you get your PhD, you could start thinking about things like interpretations of quantum mechanics, or what happened before the Big Bang’. So, I was struggling with that and then I discovered that there was such a thing called the philosophy of science and gradually it became clear to me that's what I needed to do. So, I went to get my PhD in philosophy and then history of science came slightly later in the course of doing philosophy.
But one thing I have to add to that story is that I got a really excellent humanities education at Caltech, which is not why I went there. But as it turns out, they do have an excellent division of humanities and social sciences. I was taught by great people like Jim Woodward and Daniel Kevles, and it was just such a draw because every time I expressed any genuine interest in the humanities, the professors over there just loved it. They were so encouraging. I could get a reading course on whatever topic I was interested in that they didn't have a course on. So, I ate all of it up and ended up actually with a self-designed major, which I called ‘Physics in Context’ and then PhD in Philosophy and research in History of Science. That's how it went.
[00:04:34] Samara Greenwood: A topical question for you next. Recently on the podcast, we covered the relatively recent ‘turn to practice’ in philosophy of science. Interestingly, it provoked some strong opinions on social media about the positives and the negatives of this turn. I know you've played some part in this shift, I was wondering what you think the strengths and weaknesses are of the practice approach to philosophy of science?
[00:04:59] Hasok Chang: Yes, sure. I think a clear possible weakness is that it can go kind of vague or wobbly in philosophical terms. So, we have to take care that our thinking is systematic and precise, but I think we can do it and there are many excellent practice oriented philosophers doing precisely that.
Through organizations like SPSP, the Society for Philosophy of Science and Practice, we are really promoting the highest quality work. I know you've had a recent podcast with Rachel Ankeny. She and I were among the co-founders of SPSP. It's been a wonderful enterprise to be involved in, but for me, the practice orientation is almost a given. It's not even really a choice. That's the only way I can see doing philosophy of science.
As I mentioned, I had come from science. My philosophical questions were always really about the details of science. So, the practice orientation, I think, comes naturally. Then, because I got so interested in the history of science, I just almost always found the more, let's say, ‘proposition focused’ philosophy of science really quite useless for most work in history of science. I think that's been responsible for this great gap people often see, at least in the Anglophone world, between history of science and philosophy of science. If there's blame to be had, I think much of it is with the type of philosophy that we have been practicing in mainstream philosophy of science.
So, I think the practice orientation for me is the natural way to bring the ‘H’ and ‘P’ together in HPS.
[00:07:01] Samara Greenwood: Today we're discussing the concept of epistemic iteration, a term you introduced in your 2004 book, Inventing Temperature. Could you first tell us the backstory of how you came to formulate this particular concept?
[00:07:17] Hasok Chang: Yes, as you mentioned, the concept came up in my book about the history of temperature and thermometers. It was my answer to a real dead end in my thinking that actually gave rise to the book in the first place. The book began when I started thinking a bit casually about how thermometers work. Easy, right?
But when I started thinking about how we know if our thermometers are correct, then I realized there was a really serious circularity there. So, we might want to say - if you're looking at the standard mercury thermometer - we need to know that the expansion of mercury as temperature goes up is actually linear or uniform in some discernible way.
How can we test that assumption? We'd have to know enough thermal physics to be able to test that assumption, because the obvious way to test it would be with a thermometer, which we don't have. What you'd love to do is make a plot of the volume of mercury against real temperature, see if that's a straight line, but where do we access the real temperature without a thermometer that we already trust? There's circularity one. Failing that, what we'd like to do is have a good enough theory of thermal physics which can tell us for sure how mercury or any other liquid fluid would expand. There is circularity number two. So, you need to have a thermometer that you trust in order to be able to have a trustworthy thermophysics and it goes round and round.
So, I was just completely stuck. I said, ‘how do I get out of this?’ And epistemic iteration was the answer. The answer was, we start from something that is quite uncertain, like the thermometer that you rig up without a complete assurance, either by experiment or theory, that it is correct. But you take that as your provisional temperature standard, which allows you to go and make investigations in thermal physics.
Then the idea of iteration is that ,having made such investigations, hopefully you learn something that allows you to come back to your starting point and correct it or refine it. That's exactly what I then began to realize had happened in the history of thermometry physics in general.
[00:10:10] Samara Greenwood: Excellent. I was wondering, from that starting point, did you use other case studies to also think about this, or it was just purely looking at that particular case study?
[00:10:20] Hasok Chang: So, in that 2004 presentation, there were multiple examples, all from thermometry, right? Starting with, in fact, the use of bodily sensation as a starting point. Where our notion of temperature comes from in the first place is the sensation of hot and cold. Initially we say, ‘yes, there's something there about our world that can be captured by this notion of warmth or temperature’. Then, what happened next, is people made these little devices with a glass tube filled with a liquid, and they realized, ‘Ah, when it gets warmer, the liquid goes up. When it gets colder, liquid goes down’.
The term of art for that is a thermoscope, right, rather than a thermometer, because you don't actually have numbers attached to the readings. Once you invent a device like that, what you realize is actually that device can correct my sensations in a way. Because I say things like, ‘Oh, the thermometer's not changing at all, but I'm feeling really cold and I'm shivering, so I must be running a fever.’
You start doing these things, which looks like a repudiation of your original starting point. Because why you said this thermoscope was a useful thing at all was because it agreed with your sensations of warm and cold, and that's fine. When you get down to it and start seriously using the thermoscope, you realize it doesn't always agree with your sensations, and then scientists made the decision that the thermoscope was to be trusted more than the original standard, which is the human sensation. Why? Because the readings of the thermoscope had better qualities by certain standards, like consistency.
There's a famous old experiment in which you get three buckets of water, hot, cold, and lukewarm in the middle. You plunge your left hand in the hot bucket. Plunge your right hand in the core bucket. After a while, you take them both out and put them both in the lukewarm bucket. Your left hand says, ‘Oh, this is really cold’. Your right hand says, ‘This is really warm’. Even though they are in the same bucket. So, then you realize, my hands can't be trusted.
But, the thermoscope doesn't suffer from that problem of inconsistency, it just gives you one reading in that middle bucket, right?
It goes like that at every stage. You start with some prior standard, which you believe or trust for, let's say, some historical reasons. Because that's what you've inherited, either from your biological evolution or from your previous stage of development of science. So, you start on that basis, you go on, and you come back, hopefully, to be able to improve or refine the starting point. I say improve because one thing that happens in the development of measurement standards is higher and higher precision.
Anyway, this is a long answer to your question. Initially, all the examples were from thermometry. When I put out the concept of epistemic iteration, I didn't claim that this was a very general thing. I presented it as an idea that helps me make sense of these specific instances, which I was treating in the book. But later, I did find examples of epistemic iteration almost everywhere. Maybe you'd like me to say something about that?
[00:14:23] Samara Greenwood: That would be great, if you can talk about some of the specific ones that have jumped out to you.
[00:14:28] Hasok Chang: So, an easy extension was almost automatic, right? Because it wasn't anything particularly about temperature that warranted this iterative way of working. This would happen and did happen anytime anyone wants to create a measuring instrument or observational instrument. Something I learned from Nick Rasmussen, who's been teaching in Sydney for a long time. Nick taught me all about how electron microscopes were validated at first. At first, they had to be validated by a match of output with the high-powered light optical microscopes. But once you've done that, then you push on with the electron microscope, you reach much higher resolution. You claim to see things that you definitely can't see with an optical microscope. Anytime you have to establish a new measurement or observational apparatus something like epistemic iteration needs to happen. So that was an almost automatic extension.
But then I also started thinking about categories, kinds. In other words, going into ontology, metaphysics. This was occasioned especially by Catherine Kendig asking me to put a chapter in to her book on natural kinds after the practice turn. I ended up writing an article about how what we usually consider natural kinds also evolve through a process of epistemic iteration. So that was a sort of a leap from measurement to ontology.
As a very quick example, take the notion of an atom whose defining feature in the ancient times up to John Dalton and beyond in the 19th century was indivisibility. Then what is an atom in the 20th century? It's definitely not indivisible. It's that kind of looping development that happens all over the place, so we get a continual change in the meaning of concepts in the furniture of the universe that we presume in our investigations, and that's also an iterative process.
[00:17:02] Samara Greenwood: They're excellent examples. Just as you were talking about that one thing that has always struck me with epistemic iteration is also how I see it in my own process. Just personally, as I'm working through research problems. I was interested to know if you find that, can you see yourself doing some epistemic iteration in your own work?
[00:17:23] Hasok Chang: I think so. I mean, in a way, this is a danger. I think epistemic iteration is the kind of concept that you could apply everywhere. Once you recognize it, you tend to see it everywhere. But, is it now me treating everything as a nail because I have a hammer? Or is epistemic iteration really a very pervasive feature of human research and thinking?
I'm inclined to think the latter, and this is where the philosophical picture gets big. What I was struggling with in the original work with temperature was the kind of foundationalism people in current philosophy tend to absorb without even being taught it explicitly, right? I mean, we all go through Descartes meditations and whatever. We get this instinct that true science must start from a firm foundation. Whether you consider that some kind of pure observation or some platonic intuitions or whatever, we have this notion that we must start from a firm foundation. And time and again, that's what I see not happening in the practice of science.
When you think about it, that's obvious, right? We know that Descartes had to resort to God in order to get his certainty. And if we have rejected God, at least in the realm of empirical science, and if we have also rejected some sort of infallible intuition, then we have got nothing left that would give us certainty. Then, any investigation we make has to be on the basis of something like an iterative method. We have to start from where we stand, and then if we're lucky enough, we will be able to improve our foundation. But the foundation's never going to be indubitable and eternal. The foundation is going to be provisional.
So I think, yes, I do see it everywhere, that includes philosophy. This is a too long a story to tell within the 30 minutes, but I see it connecting deeply with the pragmatist tradition, with, for example, John Dewey. Saying, ‘look, even logic is an empirical discipline, we're going to improve it as we go on the basis of our experience of inquiry.’ So, I'm taking an empiricist or pragmatist attitude towards philosophy as well.
How do we do philosophy? In philosophy, we can only start with a plausible provisional starting point and let me see how the thinking goes and I'll probably have to come back and revise my basis.
[00:20:31] Samara Greenwood: That's fabulous. So how is the concept of epistemic iteration useful for thinking about the practice of science?
[00:20:38] Hasok Chang: There's an overall perspective I’d like to offer in that regard, which is that if we want to talk about the practice of science, then we have to think in terms of actual places where scientific investigation can start. That is indeed what practicing scientists do help themselves to. There are occasionally practicing scientists who claim to have the word of God <laughter>. But, on the whole, they go with what they’ve got in terms of what seems like the most reliable basis on which to make their observations and builds their theories.
I think in that regard, epistemic iteration is probably the best overall framing device for the study of the practices of science.
[00:21:34] Samara Greenwood: One thing that's come up a little bit is, okay, what happens when science goes wrong? Is there anything epistemic iteration could bring to this?
[00:21:42] Hasok Chang: There are two things that can happen.
One is, if the iteration goes well, then you will be able to revise your starting point without completely discarding it. And there would be some sort of convergence happening in the successive steps. We don't need to pretend that what we're converging to is the absolute truth or anything. But convergence means stability, it means something we can build something else on. So that we can have - if we're lucky. And that should be helpful to the practicing scientists as well as to us observers of scientific practice.
But it can also happen that epistemic iteration gets nowhere. You actually discover a dead end. Then you just have to go back to the drawing board, start with a new provisional starting point, see how that runs.
So it could be that we mess up the earth so much that we do have to go with Elon Musk to Mars, hopefully not, right? Hopefully we can work with what we've got, but some things are dead ends, and I think that's also happened in the history of science, right? I mean, alchemy, as we know it, was kind of a glorious dead end. A lot of wonderful things came out of alchemical investigations, but if alchemy means a crude thing about gold making or creation of medicine to give you eternal life or something, that turned out to be an unproductive starting point, and that's fine, too. We learn on a large scale in that way.
[00:23:35] Samara Greenwood: So I'm interested to know what kind of reception you've had to the concept. And, do you think it's been taken up by others in any interesting ways?
[00:23:43] Hasok Chang: So, the reception has been quite pleasing. I think anyone who considers the issue tends to agree that yes, something like epistemic iteration has to be happening in order for us to make credible progress in science or in other realms of life as well.
On the other hand, I think a lot of philosophers, particularly in metaphysics, have just paid no attention to this kind of idea because they're not, in the end, interested in any sort of evolution of concepts. In that regard, I've actually been making some really good connection with people who are working on conceptual engineering these days, because epistemic iteration fits almost exactly the kinds of developments they're looking for.
So yes, even in metaphysics, those who are inclined towards conceptual engineering tend to be very much in favor of the iterative perspective, but that's certainly a minority within metaphysics.
[00:24:54] Samara Greenwood: We talked a bit about how you feel the concept of epistemic iteration may be useful for practicing scientists. Was there anything else that you wanted to mention on that?
[00:25:03] Hasok Chang: Yes, one of the most interesting experiences I've had in terms of the reception of the idea of epistemic iteration was an approach from psychiatrists. This was particularly through the very philosophical psychiatrist Ken Kendler who had also been talking a lot with the philosopher Ken Schaffner. So, the two Kens, along with Katie Tebb, who had been working with Schaffner.
They invited me to speak at one of the annual conferences they had been running in Copenhagen about psychiatry and philosophy. That year's theme was taxonomy or psychiatric nosology, disease classifications in other words, and Kendler and others were despairing about how the diagnostic categories were always shifting around within psychiatry, right?
So, going from DSM 3, 4, 5, they couldn't see stability, the categories kept getting revised and they thought, does that mean we're not talking about anything real? Kendler picked up the notion of epistemic iteration, so I went and gave a talk.
My main message was that it's okay to have shifting categories as long as you're managing the shifting in this iterative fashion. Another thing they were worrying about was whether they are getting stuck in, as it were, local minima. Whether they were getting stuck in a stable place that's not the optimal solution. My message about that was, well, that may be, but don't knock local minima so much. Because if by iteration we get to a stable place, that is worth something.
So, that was also a pluralist message. There isn't just the one right answer, especially when it comes to things like taxonomy, kinds. There are many ways of cutting the universe. There may be multiple joints, even if you believe in the joints. And it's fine to try out different methods and try to improve each one of the starting points that you may take.
[00:27:38] Samara Greenwood: I'm also interested if anyone has challenged you on the concept at all.
[00:27:41] Hasok Chang: Not very much. Again, there are people who have completely ignored it. But in terms of a challenge, I don't think I've had very much. But people do raise serious questions, including, well, what happens if your iterative process doesn't get you anywhere stable? And that's the point I tried to cover earlier. You may just have to abandon it and come to a new starting point. From the strongly scientific realist side, people have questioned the value of a nicely constructive iterative process, because to them it doesn't give you enough, doesn't get you to truth, and therefore it's not really that exciting. So I've had some skepticism in that regard, but I think that's part of a bigger dispute, not really about epistemic iteration.
[00:28:43] Samara Greenwood: And I guess one potential line of challenge would be what we were talking about before, is that you could tend to see it everywhere when perhaps it's not always the most appropriate way to describe a process, you could have other processes.Would that be true as well?
[00:28:58] Hasok Chang: I think that is true. People haven't specifically made that point to me very much, but I think that is true. I mean, one thing I think we have to consider carefully is whether there are any ways in which something like foundationalism is the right framework, at least for analyzing certain situations. I haven't really seen them, but it's not to say maybe there is.
[00:29:29] Samara Greenwood: What about people beyond research and academia? What kind of value might the concept have for the general public?
[00:29:35] Hasok Chang: I think I would channel John Dewey again to say when it comes to things like epistemic iteration, I don't think there's a sharp boundary between scientific research and other kinds of intelligent action, right?
So that's the continuity that Dewey saw between everyday life and science, and I see it as well. When we're trying to do anything at all, when we're trying to learn anything in life, I think it has to be something like an iterative process. Now, there will be another model, which is the dogmatic one, in which you just get told something and you believe it.
But in the big picture, I mean, how did even dogmas arise, unless you really think that God handed down the tablets to Moses? Right, okay, possible, but most likely we learned anything at all in life by sort of an epistemic iterative process. And there's something that Popper said once that's similar. When people asked him, ‘well, where does that endless process of conjectures and refutations begin?’ He thought, well, most likely the initial conjectures have to be biological instincts, something we are born with. And I think that's probably right.
[00:31:06] Samara Greenwood: Also, from what you're talking about, the lived experience of everyday life. Just trying to understand how do I get through the day? That's the starting point for that inquiry, isn't it?
[00:31:17] Hasok Chang: Yes, and when we think about how do babies learn language? How do they learn bodily coordination? And starting from that, how do we learn any kind of skill in life has to be based on some other skills that we already have and so on.
[00:31:37] Samara Greenwood: That's a really nice connection to the bigger picture. Finally, I wanted to ask if you think there's any untapped potential in the concept. So, what kind of further research would you like to see on epistemic iteration?
[00:31:52] Hasok Chang: That's a fantastic question to which I don't have a very good answer but let me just throw out a couple of vague notes. So, one limitation that I'm very aware of in my own work is that I'm always working from episodes in the physical sciences, simply because that's what I know.
What I haven't seen worked out very much, and I'd love to see that done by some other good people, is epistemic iteration in the biological sciences, especially. I know that there are people trying to apply it in areas like measurements in the social sciences. There's good work happening there. But I think we're pretty thin on the life sciences. So that would be a really nice area in which we can push the concept further.
Another direction would be something more systematic in what we had just been talking about, right? Going outside science, how does learning in general rely on a kind of iterative process?
Then there are also directions of work that would add philosophical precision to the idea. This is a little contrast with mathematical iteration, which I was inspired by initially. In mathematical iteration, there is a definite algorithm to get you from the initial guess to the next step, and it's the same algorithm that is applied at every step. That's definitely not true in epistemic iteration. So there's sort of a question, like the question about the context of discovery. How do we do this? How do we start from the unnumbered thermoscope to the numerical thermometer, for example?
That, to me, seems a much underexplored area. There must be something other than, ‘well, we just apply what we've learned at stage 2 to make stage 3’. I've left it vague like that, and something more precise, I think, needs to be said.
[00:34:14] Samara Greenwood: Oh, that sounds like a wonderful project for one, or many, people to take up.
[00:34:20] Hasok Chang: Someone should do it.
[00:34:22] Samara Greenwood: Thank you Hasok for being on the podcast. It's been an absolute pleasure.
[00:34:26] Hasok Chang: You're most welcome Samara. I look forward to hearing the final product.
[00:34:33] Samara Greenwood: We hope you're enjoying season two of The HPS podcast, where we discuss all things history, philosophy, and social studies of science. To learn more, you can check out our website, hpsunimelb.org. You can find links to our blog, Bluesky, Twitter, and Facebook. I'm Samara Greenwood, my co producer is Indigo Keel, and we would like to thank the School of Historical and Philosophical Studies at the University of Melbourne for their ongoing support.