COMPLEXITY: Physics of Life

John Krakauer Part 1: Taking Multiple Perspectives on The Brain

Episode Notes

The brain is arguably one of the most complex objects known to science. How best to understand it? That is a trick question: brains are organized at many levels and attempts to grasp them all through one approach — be it micro, macro, anatomical, behavioral — are destined to leave out crucial insights. What more, thinking “vertically” across scales, one might miss important angles from another discipline along the “horizontal” axis. For inquiries too big to sit within one field of knowledge, maybe it is time we resurrected the salon: a mode of scientific exploration that levels hierarchies of expertise and optimizes for more complementary and high-dimensional, egalitarian, communal discourse. As with the Jainist philosophic principle anekantavada — how many blind people does it take to grok an elephant? — neuroscience is perhaps best practiced as innately and intensely multiperspectival…

Welcome to COMPLEXITY, the official podcast of the Santa Fe Institute. I’m your host, Michael Garfield, and every other week we’ll bring you with us for far-ranging conversations with our worldwide network of rigorous researchers developing new frameworks to explain the deepest mysteries of the universe.

This week is part one of a two-part conversation with SFI External Professor John Krakauer, Professor of Neurology and Director of the Center for the Study of Motor Learning and Brain Repair at Johns Hopkins . In this episode, we talk about the history of different ways of studying the brain — in animals and humans — and how subjects as complex as brains invite a different way of seeing, one that synthesizes many different ways of seeing…

Thanks for your patience with the recent delays in publication — with InterPlanetary Festival and our Annual Symposium behind us, Complexity will now return to regular biweekly scheduling.

Be sure to check out our extensive show notes with links to all our references at, and stay tuned for part two — in which we talk about how learning is inherently a future-focused exercise, and what that means for education. If you value our research and communication efforts, please subscribe, rate and review us at Apple Podcasts or Spotify, and consider making a donation — or finding other ways to engage with us, including an open postdoctoral fellowship in Belief Dynamics — at

Thank you for listening!

Join our Facebook discussion group to meet like minds and talk about each episode.

Podcast theme music by Mitch Mignano.

Follow us on social media:
Twitter • YouTube • Facebook • Instagram • LinkedIn

Referenced in this episode:

Neuroscience Needs Behavior: Correcting a Reductionist Bias
John Krakauer, Asif Ghazanfar, Alex Gomez-Marin, Malcolm MacIver, David Poeppel

Two Views of the Cognitive Brain
David Barack & John Krakauer

On Beyond Living: Rhetorical Transformations of the Life Sciences
Richard Doyle

Simon DeDeo on Good Explanations & Diseases of Epistemology
Complexity Podcast Episode 72

Former SFI Fellow David Kinney, epistemologist (re: disciplines as levels of explanatory granularity)

Coarse-graining as a downward causation mechanism
Jessica Flack

Integral Ecology: Uniting Multiple Perspectives on the Natural World
Sean Esbjörn-Hargens & Michael Zimmerman

Carl Cranor, moral philosopher (re: causation)

The Learning Salon: Toward a new participatory science
Ida Momennejad, John Krakauer, Claire Sun, Eva Yezerets, Kanaka Rajan, Joshua Vogelstein, Brad Wyble

Brain Inspired Podcast
Paul Middlebrooks

eLife Journal


W. Brian Arthur on Economics in Nouns and Verbs (Part 1)
Complexity Podcast Episode 68

W. Brian Arthur (Part 2) on "Prim Dreams of Order vs. Messy Vitality" in Economics, Math, and Physics
Complexity Podcast Episode 69

Sand Talk: How Indigenous Thinking Can Save The World
Tyson Yunkaporta

Episode Transcription

John Krakauer (0s): It’s not just pluralism for us as humans dividing the world into disciplines each with its own vocabularies and conceptual frameworks and explanatory objects. I think, as I said before, the nervous system from the inside develops different objects for control as you go up the neuro axis. So I think that there's pluralism in how the nervous system sees the control problem and the representations it uses, and that is the ontological truth of pluralism. And then it has a kind of mapping onto the epistemological pluralism we have, and it may well be the reason why we have psychology and neuroscience, and we have psychiatry and we have neurology, is that those are as valid and separable as economics and philosophy are, or economics and sociology. 

We don't believe that those disciplines anytime soon are gonna collapse onto the more basic one. So it may just be that we are beginning to discover that our disciplines in their divisions are actually picking up on true ontological differences. I think pluralism is a deep metaphysical truth. I don't think it's just our limitations as humans having to divide the world up into these disciplines. But an alien species could just see them all as a kind of particle physics. I just don't think that's true. 

Michael Garfield (1m 43s): The brain is arguably one of the most complex objects known to science. How best to understand it? That is a trick question. Brains are organized at many levels and attempts to grasp them all through one approach, be it micro, macro, anatomical behavioral are destined to leave out crucial insights. What more thinking vertically across scale. One might miss important angles from another discipline along the horizontal axis. 

For inquiries too big to sit within one field of knowledge, maybe it is time we resurrected the salon, a mode of scientific exploration that levels hierarchies of expertise and optimizes for more complimentary and high dimensional egalitarian communal discourse. As with the Jainist philosophic principle anekantavada, how many blind people does it take to grock an elephant? Neuroscience is perhaps best practiced as innately and intensely multiperspectival. 

Welcome to 


y, the official podcast of the Santa Fe Institute. I'm your host, Michael Garfield, and every other week we'll bring you with us for far ranging conversations with our worldwide network of rigorous researchers developing new frameworks to explain the deepest mysteries of the universe. This week is part one of a two part conversation with SFI External Professor John Krakauer, Professor of Neurology and Director of the Center for the Study of Motor Learning and Brain Repair at Johns Hopkins. 

In this episode, we talk about the history of different ways of studying the brain in animals and humans, and how subjects as complex as brains invite a different way of seeing, one that synthesizes many different ways of seeing. Thanks for your patience with the recent delays in publication. With Interplanetary Festival and our annual symposium behind us, 


will now return to regular biweekly scheduling. But be sure to check our show notes with links to all our references and our epic Twitter threads from the above mentioned events at

And stay tuned for part two in which we talk about how learning is inherently a future focused exercise. If you value our research and communication efforts, please subscribe, rate and review us at Apple Podcasts or Spotify and consider making a donation or finding other ways to engage with us, including an open post-doctoral fellowship in belief dynamics

. Thank you for listening. John Krakauer, welcome to 



John Krakauer (4m 41s): Thanks for having me. 

Michael Garfield (4m 42s): So I would like to start by giving people a bit of preamble about you and your life and an intellectual autobiography, how you came to be on the quest that you're on today. What motivated you to ask the questions that you're asking in your scientific career? 

John Krakauer (5m 2s): I think it came from me wanting to be a doctor and a scientist and not realizing that those were two careers that would intertwine. So I think that was the first thing that happened is that I ended up starting doing two things and from there I think a lot of the directions I took stemmed from the fact that I was a medical doctor doing neurology and a scientist doing neuroscience and the way that they cross pollinated could be blamed for almost everything that I'm doing now because I didn't realize that you had to be into science to do medicine. 

I was actually very interested in the humanities, literature, philosophy, history, and so I thought I'd be a doctor and do those subjects. And then I fell in love with science, chemistry and physics in particular. And it took me a while to realize that my love of chemistry and physics and my desire to be a doctor were a better match than being a doctor and doing philosophy, history and literature. My love for those never ended and I think they've continued to inform what I do, but I think it was really that double origin that has continued to dictate the kind of question that interests me. 

Michael Garfield (6m 14s): And so why the brain? 

John Krakauer (6m 17s): I think when I was an undergrad at Cambridge, I gravitated away from physics and chemistry and moved towards biology. I was interested in molecular biology and then I moved away from molecular biology. And in my last year at Cambridge, I was in the zoology department and while I was in zoology department, I did projects, for example, one on locomotion in the locust with a very famous biologist called Malcolm Burrows who had done seminal work on locomotion in insects. 

And so I did stuff with him. Gabriel Horn, who was the chair at the time with doing work on imprint, had done seminal work on imprinting and chicks. I just began to meet neurobiologists and got interested in developmental neurobiology. So my interest just across my entire life as an undergraduate at Cambridge went from maths, physics, chemistry, molecular biology, immunology, neurobiology to behavior in almost a tight concertina of experience. 

So that stayed with me. And then I went to Columbia and went to med school, did a rotation on a pediatric neurology ward. And it began to dawn on me that I was interested in neurology as a clinical specialty and neuroscience as a scientific one. Anyone looking from the outside who said it was obvious that that was my trajectory, but I didn't realize it until it was staring me in the face really After that I began to think, well, what sort of research should I do? Because I realized that I wasn't gonna just be a doctor. 

I even then I thought I was gonna do more molecular neuroscience, but I didn't really have a taste for that. It was like cooking. I wasn't interested in being in a wet lab. I started to ask system level questions on the wards. What does the cerebellum do? What do the basal ganglia do? What does the motor cortex do? I began to try and derive almost from first principles what those structures were doing based on what I was seeing in the patients. And then I realized, well, this isn't gonna do. I'm gonna have to do some systems level neuroscience to compliment what I'm seeing on the wards. 

And I joined a motor physiology lab and then the rest is history really. I joined, my mentor was Claude Ges. He was a very brilliant experimentalist with theoretical intuitions and he took me, had done seminal work in the cat, then move to the human. He taught me how to do psychophysics. I wasn't really aware that psychophysics was a discipline, you know, that you could analyze behavior in that quantitative way. And so then what I was seeing on the wards and the system level questions I was asking on the wards and the type of neuroscience you can do at the systems and behavioral level fused and that was it really. 

But it's much easier to tell it in retrospect than the sort of zig-zaggy path it was at the time. And then suddenly you find yourself good at it. And other words, that's the other very interesting thing is you go from being a good student to going to be a bit lost because you have to be creative as a scientist, not just good at taking exams. And then I began to realize that I actually quite liked trying to be creative in this space. And then you begin to develop a little bit of confidence. You develop, you publish a few papers that get attention, you get properly mentored and then voila, the souffle has been made. 

And from that foundation, that intellectual style that you build early on, you go with it. I hope that wasn't too long an answer. 

Michael Garfield (9m 47s): No, you've got this piece that you wrote on the Journal of Neurophysiology, on the neuro scene, essays at the interface of neuroscience in the world, arguing for more sort of popular long form science essay stuff in scientific publications. And you and your brother both have a fondness for the essay that I share. Let's make this conversation kind of an essay. 

John Krakauer (10m 7s): Yes, absolutely. 

Michael Garfield (10m 8s): And so just out of the seven or so publications that I felt worth highlighting, one of the ones that came up in what you said for me now was another piece that you wrote 

Perspective in Neuron

that led this piece with a number of co-authors on 

Neuroscience Needs Behavior: Correcting a Reductionist Bias

. There's this, this relationship between the way that these things are understood. And so if you could just take a moment to unpack that, cause I think that will give us kind of a nice bed within which we can grow some other things. 

John Krakauer (10m 43s): Yeah, I mean that piece which came out on February 2017, arose from a Gordon Conference that happened I think two to three years before where I gave a talk and some editors said, John, you should really think about turning that into an essay. And it turned out that David Poeppel, you know, a prominent neuroscientist at NYU, was at that Gordon Conference. And we immediately struck up a friendship. I mean, in the same way that the editors came up to me to write something, he just came up to me and said, we should be buddies. 

And we promptly went off in his convertible to get good coffee, which certainly wasn't available at that hotel. And we started to plot a piece where we would try and make a case that there was still a need to take a psychological stance. In other words, take a look at behavior, decompose that behavior into computations and algorithms, test hypotheses at the level of the behavior and experiments on behavior. And then when you have a framework, when you have a model, then start looking for the neural correlates. 

So the argument really was that you take the David, the classic David Marion framework of the computational, algorithmic and implementational levels and actually make a case that when you look at behaviors, especially simplified behaviors you can quantify, you're really doing work at the algorithmic level. You're saying, okay, here's a task analysis. How do you decompose that task into component algorithmic processes? Can you test either/or hypotheses? Can you compare models? 

And then once you've gotten that decomposition, you are much better position to go into the brain and look at the neural correlates. Now what are the neuro correlates gonna tell you? And this was the deep point about this article, which is, do you break ties? Is one idea that you have these different behaviors at the level of the task analysis and the only way to break the tie is to actually look at neural correlates. So in other words, you go into the brain almost for confirmation, not for discovery. So the thing I think that upset a lot of people and inspired others is most of the time the conceptual advances at the level of the task analysis. 

And then let's say that you posit something like a reward prediction error or a sensory prediction error and your experiments suggest that such an error should exist, then you go into the brain and you find a correlate of that error and you go, Ah, we were justified in positing such an entity because we could find a neural correlate for them. Now there are many, many pitfalls of coming to the conclusion that your variable exists inside the brain because you find a neural correlate for it, which we shouldn't discuss today. But nevertheless, it was basically the directionality of discovery should be task analysis theorized about that posit algorithms and computations and then go into the brain.

And what we felt, what the reductionist bias was, is things were kind of going the other direction. You would have a behavior. The behavior wasn't really very well thought. It wasn't constructed in a very sophisticated manner. It was really a behavior just to get the neurons going. So you just wanna get the neurons going and then you can just dive into the brain and the rest of your work will be a neuron land. So the behavior was really to initialize the system rather than being the thing that you wanted to explain in the first place. 

So you'd find papers that had over 50 figures, 49 of them would be neural and one would be of the behavior. So in other words, we thought this was all backwards. So that was the main idea. Also, we thought that you have to be very careful that you don't devise such artificial tasks and experiments that you are taking the nervous system with this poor animal for a ride that has nothing to do with what it was ecologically designed to do. It would be like using an iPhone as a hammer, you can use an iPhone as a hammer. So it was that idea that you could get yourself into trouble forcing neural responses out of unnatural behaviors that just lead you in the wrong direction. 

So the sort of second point with make sure that you go for ecological behaviors that are in the in velt of the animal. So it was really just a plea to go back to thinking about evolution, behavior tasks and not get carried away by methodologies where after you've gathered all the data, then you start to try and tell a story after the fact. So we were very much against these after the fact stories that you tell after you've got all these data you've collected. 

Then we made fun of things like filler terms. You do a correlation. The correlation you feel isn't sufficient. So what you do then is you say, this neural activity underlies this behavior or is the basis for this behavior or represents this behavior. And all you've done is found an English word, what we called a filler term. That is just another way to state the correlation that you already found. But because everyone found that the correlation wasn't explanatory and what we said, they feel the lack of the explanation in the correlation, they find another way to restate the correlation with a filler term underlies is the basis for et cetera, et cetera. 

And but those words are not doing any extra conceptual work. They're just sounding a little bit more like an explanation going beyond the correlation. So there were a lot of things in there about what the neuroscience project really is. What are you up to, what are you trying to do when you find neuro correlates of things? And then the other point we made was that there's a big distinction to be made between understanding something and fixing something. So there were many things in that paper that we all wrote, many of which had been said before. 

But ultimately we thought the best neuroscience was the neuroscience which did the decomposition at the task level, came up with a model or a hypothesis and then went into the brain even to break the tie confirmation or maybe even update the algorithmic hypothesis because you had a little extra data in the form of neural data, but it was really flat evidence landscape, you have behavior, you have neural behavior. Those are two sources of information. Together you can construct a theory, but that's not the same as making a mechanistic claim. 

Just because you get a neural correlate of a behavior doesn't mean that you've given insight into the mechanism. Just because the neural data is neural data rather than limb data or behavioral feeding data doesn't automatically make the neural data more fundamental than the behavioral data. They're just two sources of evidence to construct a theory. You can use neural data to be mechanistic, but it isn't automatically mechanistic just because it happens to be neural data. So there were many points that were made in that paper and I'll be very honest with you at the time that it came out was in February in the winter when I had to get my stroke book into MIT Press in time for it to come out at Society of Neuroscience that fall. 

So my mind was entirely on getting the manuscript, the book manuscript to MIT. And I kind of had forgotten about this paper, not that we didn't work very hard on that, all five of us and then all help wrote loose with this paper. I mean Twitter, Facebook, email, none of us imagined that it would become this viral paper. And I don't mean that with false modesty, we just thought it would be interesting. We thought people would like it, but we didn't think that it was so of the moment that people all feeling this strange reductionist, methodological turn in neuroscience and along came this sort of counterblast and to this day, I mean it's a defining paper for us, but I stand by it. 

I think there's very little we would change in it. 

Michael Garfield (18m 36s): So a couple things there. One is that it seems kinda surprising to me that neuroscience, which almost anyone standing outside of it would assume is the place you go if you're not trying to think in a reductionist way. Like if you, you want to understand patterns and emerging, what is it about the history of the development of that field that led to this bias in the way that the science was being practiced? 

John Krakauer (19m 9s): Yeah, I mean it's a really good question. I think, you know, Tim Shallice in one of his books has sort of outlined sort of three parallel traditions in the 20th Century when it comes to brain science. One is to treat the brain in a biomedical sense, just like any other organ. We've understood how the kidney works and how the liver works and how the heart works and we have a mechanistic understanding and we understand how a nephron works, how a hepatocyte works. So I think there was very much this belief that you could treat the brain like an organ, like all the other organs in the body and just do pathophysiology, look under the microscope, look at cells. 

So there was that tradition. Then there was a sort of neuropsychological tradition, which was that you could learn very interesting things, especially about cognition and behavior from patients with lesions. So that was the neuropsychological tradition, 

Michael Garfield (20m 2s): Knock this out and see what happens. 

John Krakauer (20m 4s): And you know, patients with aphasia, patients with prefrontal damage. So in other words, there's an incredibly rich tradition of inferring the function of the brain through the effects of lesions. And that's neuropsychology. 

Michael Garfield (20m 17s): Same thing happened in genetics, right? You know the fat gene. 

John Krakauer (20m 21s): But, but I would say, you know, and I'm not in any way somebody who can talk about genetics is they both led to very interesting discoveries. The action potential, you know, Hodgkin and Huxley, it was very much understanding how a particular specialized cell did what it did. And then neuropsychology with this huge body of knowledge that we developed about the modularity of the brain and confirming how language might be structured because you could have syntactical errors and grammatical errors, et cetera. 

And then there was the information processing computer revolution going on at the same time. And what's fascinating, and you can look in the 1950s, all these traditions were running along in parallel, the biomedical model, the neuropsychological model, and the computer model. Now you could argue that what happened is that the organ model, the sort of reductionist model has gotten such a boost from the extraordinary methodology from photomicroscopy, optogenetics, transgenic animals. 

I mean you can play God with animal models. And so neuroscience, the predominant notion of neuroscience was we should be able to have identifiable circuits with identifiable connections, almost intuit the sequence of transmission from neuron age neuron B to neuron C and have a kind of intuitive understanding just like you'd get when you look at a diagram of the loop of Henle in the kidney, you go, ah, that's how urine is made. And you get concentrated urine, Oh that's how a ribosome works. 

You can see a video of a ribosome and an RNA swimming like a coral snake into the ribosome. You can do this sort of visualization at the level of the component parts and how they interact.t Sherrington with the stretch reflex, you see the one A coming off the muscle spindle, it's synapses, monosynaptic onto the motor neuron and you can go, oh, I can see exactly how if you bang on the knee why you'll get a reflex. So there was this sense that in a sufficiently reduced system you could have an isomorphic click between the neural diagram, perhaps the computational model and the behavior. 

Do you see what I mean? There was this, ah, we are gonna be able to sort of intuit visually and you know, there's a whole tradition in physics through Einstein about being able to visualize things. And so there was a feeling that we're on our way. You could do the same thing with the action potential. You know, Hodgkin and Huxley had a circuit diagram which you could intuitively understand. So there's this belief that this sharing Sherringtonian project of neurons, their connections one neuron and pinging upon another like a billiard ball and pinging upon another billiard ball with transfer of energy causality, that project seems to have had a 2.0 version with all the new techniques and animal models. 

And the idea is that that, and we call it in a paper that David Barak and I wrote quite recently, two views of the cognitive brain where we say that sensory motor Sharringtonian project could graduate to a higher level cognitive phenomena, that you'd still get that isomorphic satisfactory click between the connections, the flow of energy and intuition, a diagram and the higher order cognition. And it hasn't really worked in my view. And we said so in the behavior paper and the more recent paper. So that's where neuroscience has gotten is it wants to eek out as much as it can with the new methodologies. 

The Sharringtonian, the brain is like other organs type of project, but running along in parallel, which is gaining more traction is the computational information processing view merging with a population level view of the brain where you get these emergent computational phenomena at the level of the population that cannot be discerned from the individual elements. And neuroscience now is going in that direction. The question is, is when you do that, will this population level emergent computational approach substitute out the psychological terms that came from neuropsychology and even the computational theory of the brain, in other words? 

And what I wonder is, will we ever lose our psychological constructs? We'll certainly put them on better footing, but we're not going to get rid of them and have some neural story that we articulate in a sentence instead. And I think neuroscience is conflicted because it almost seems to want to be able to have a neural story for every behavior. Sometimes in the Sharringtonian framework, you can have a neural story for the behavior, the new population approaches. 

People are hoping that they will be the new neural story for higher level cognitive phenomena, but there's still a possibility that these new neuro objects constructed outta populations will still only play a confirmatory role for psychological constructs. So it's quite a complex story, but there's a moment now in neuroscience where you've got this ability to record from thousands and thousands of neurons simultaneously.  Do mathematics on these large number of neurons, construct objects out of these many, many neurons decide that these objects are actually what's used by the brain from the inside out? 

And this will get us new traction on cognition that will be as intuitive to us like a Feynman Diagram as the stretch reflex circuit was for simple sensory motor behaviors. I mean that's what I think some people hope, or it may be that they will never really be the terms of the explanation, they'll just be confirmatory and just to finish the framework we used with first level and second level explainers. So the notion of first level, second level explaining in the paper with David Barack, we say, look, what was the reason the window broke a ball? 



Engage with SFI

Through meetings, schools, community events, and media, we invite curiosity-driven individuals to share our disc...




That's the explanation. Okay, why did the ball break the window? Well the ball has a certain structure and an arm through it. Those are the explainers of the ball that broke the window. But they're not the explanation of why the window broke, their explanation of the ball itself. And so I think what happens in neuroscience is they want the second level explainers to graduate to being first level explainers of the phenomena rather than being the explainers of the first level explanation, which is then the explanation of the phenomena. 

And so what we're arguing is that we are not absolutely sure what the first level explainers are gonna be of higher level psychological phenomena, but you are not gonna get the same satisfaction outta the second level explainers that you do when it comes to something simple like the stretch reflex. So at a very interesting moment in neuroscience, which is what are the intermediate objects of explanation for higher level cognitive phenomena that don't get damned for being 19th century constructs, a kind of nephology can be grounded in population level neural data, but the form they'll take, we in the paper called them neurodynamic objects. 

They would be both functional objects and they would have a neural flavor. But how ontologically real those are versus just epistemologically useful for us as scientists is still very much in debate. But you should know that we're at a very interesting moment in neuroscience, which is what is form will it take to explain things like thought and planning and motivation. And there's two camps, one thinks it's gonna be just be Sharringtonian, and again, and the way we can be Sharringtonian is studying a fly, it's an identifiable circuit, it's an honorary stretch reflex and extrapolate or go the population level route, come up with a new object and have explanations of a new kind that give respectability to psychology because at least you've created some kind of neurally derived object. 

So it's a fascinating moment and then you can decide whether you think that's a victory for neuropsychology information processing, computational theory of mind, or for the original organ-based neuroscientific project. It's a very interesting juncture. 

Michael Garfield (28m 24s): What you're talking about reminds me a lot of what my friend and and mentor Richard Doyle at Penn State University, he wrote a trilogy of books on rhetorical transformations in the life sciences and on beyond living. He talks about how the 20th Century was sort of dominated by this conceit, you know, look at something like sea elegans this is 959 cells. That's perfect. That makes it so that we can actually follow every one of those cells to talk about like spimes, which is like, you know, the digital twin, you know like a software object that reconstructs its entire history maintains the sort of like blockchain of its embryogenesis and so on. 

But the joke is of course that that path undermines itself through its own success, right? You end up revealing that it's insufficient. And so all of the stuff that you're talking about seems kind of of a piece with the conversation I had with Simon DeDeo on this show or he was talking about the different explanatory heuristics that people use. 

John Krakauer (29m 29s): I think that work is fascinating that he's done and Simon is always so thoughtful about these issues. Exactly. In other words, I think there is a dovetailing in terms of what are the component, what are the weightings on what counts as a good explanation and what will that look like when it comes to explanations of cognition and behavior And the idea that we're gonna have the same weightings and the same component objects that you explain with when it comes to a sea elegans or a knee-jerk or even eye movements versus thoughts. 

And it does doesn't seem right. I mean I can have a medical student go to the whiteboard and explain why the patient had a hyperreflexia, you know, a very brisk knee-jerk or why they have a difficulty with their eye movements. They can draw, you can literally draw circuit diagrams of eye movements in the brain stem, knee jerks in the spinal cord. But when they, the patient can't tell you the name of something, they have anomia, they just can't say the word that you show them your thumb and they just can't say it and they go draw the diagram for the trouble they're having thinking of the word thumb. 

There's no diagram to draw. So why could they draw something for the knee-jerk? They could draw something for eye movements. Here's another behavior and one doesn't even know how one would begin to draw something. Where if you looked at the drawing you go, Ah, yeah, they're not gonna be able to say thumb. It's like you're almost in a different explanatory regime where whatever the explanation ultimately ends up being, whatever the compressed form of the sentence you can utter to go, Aha, I get it. It's not gonna look like a drawing of an RNA in a ribosome or a segmental reflex circuit. 

It's gonna take some other form. Do you see? It's not obvious why phenomena demand different flavors of explanatory object to compress and explain them. 

Michael Garfield (31m 19s): If I've taken it upon myself to hammer on a theme through this entire show as a way of transducing the essence of this place. It strikes me that a lot of people come into look at this institution and have this idea that it's people seeking the one ring to rule them all. Yes. And in fact what you're saying and what Simon says when he is talking about how we have to have both parsimony and consilience in explanations and how people bias towards either approach keep each other in check, is that there is something fundamentally plural. 

John Krakauer (31m 59s): And the important thing about this exactly right, that like David Bohm, there is no privileged foundational level. There is no one ring to rule them all. I do think in a way that article in 2017 was because one began to think that neuroscience thought that the ring to rule them all was the circuit. But the really deep point I think is it's not just pluralism for us as humans dividing the world into disciplines, each with its own vocabularies and conceptual frameworks and explanatory objects is I think as I said before, the nervous system from the inside develops different objects for control as you go up the neuro axis. 

So I think that there's pluralism in how the nervous system sees the control problem and the representations it uses and that is the ontological truth of pluralism. And then it has a kind of mapping onto the epistemological pluralism we have, and it may well be the reason why we have psychology and neuroscience and we have psychiatry and we have neurology, is that those are as valid and separable as economics and philosophy are, or economics and sociology. 

We don't believe that those disciplines anytime sooner gonna collapse onto the more basic one. So it may just be that we are beginning to discover through a kind of disciplinary crowdsourcing that our disciplines in their divisions are actually picking up on true ontological differences along the neuro axis. I think that that's what's going on. I think pluralism is a deep metaphysical truth. I don't think it's just our limitations as humans having to divide the world up into these disciplines. 

But an alien species could just see them all as a kind of particle physics. I just don't think that's true. 

Michael Garfield (33m 48s): So there's a couple things there. One is, you know, David Kinney's work on understanding physics, chemistry, biology, social science and so on as different granularities. And then kind of akin to that, you've got Jessica Flack talking about collective computation and slow and fast variables, decoupling, and you get emergent top down causation and collective computation and so on. But I think that you're also saying something even beyond that in as much as you know, my graduate advisor, Sean Esbjörn-Hargens talked about ontological pluralism that in his work in exploring 200 different disciplines of ecology, you know, how can we integrate, how can we reconcile all of these? 

Then you end up talking about domains of enacted knowledge. 

John Krakauer (34m 39s): Yes. The first place I heard this example I'm giving was from Carl Craver, you know the phosphor of science who's very deep about these things. He talks about causal contrast, which is if you ask, and I've used this example on podcasts before, why did Socrates die. If you said he died because the Athenian authorities condemned him to death because he was corrupting Athenian youths, that would be the correct answer. And then you could go all the way down and somebody could say, no, no, no, he died because he took hemlock with a particular poison which acts on these particular receptors and leads to respiratory arrest and that's why he died. 

Now which one of those is correct? They're both equally correct. Then in between you could say Socrates died because he chose to take the hemlock over harmless tea or he refused to let himself go off into exile. But they're all true. And knowing the mechanism of hemlock doesn't mean that you never need to talk about the fact that the Athenian authority is condemned into death. Do you see? In other words, they're all equally true. So you need to decide as a scientist what is your causal contrast. And it turns out that where we, in our paper in 2017, and then David and I in the paper more recently get annoyed is where people choose the wrong causal contrast or they choose one and then they pretend to answer it with another set of materials that belong better with another causal contrast. 


So they get a sort of mismatch in the level at which they work, but the type of question that they actually prefer, there's a vertical set of layers of questions. You pick one of them, you do very good horizontal work at whatever level you've chosen, but then you have the hubris to believe that your horizontal work can jump vertical levels. You're greedy, you want to believe that your particular horizontal level of work is the ring that rules them all. 

And so it can begin to encroach upon the other types of answer. 

Michael Garfield (36m 38s): Is that just sort of the socioeconomic product of the way that science is produced? 

John Krakauer (36m 46s): I think that's absolutely right. The causes of this kind of confusion, a myriad, you know, it's like coon talks about it's the way that you agree to talk about the science. Everyone wants to believe that what they're doing has the most generalizability and it's not just particular and cute. It's very interesting. I was reading the interview with the Nobel Laureate who was here, the ribosome. I read the interview with him and he said, you know, you have to avoid working on cute things. You want them to be important and to generalize. 

And I think there's a desire to believe that the work you do has the most generalizability. 

Michael Garfield (37m 22s): Rama Krishna. 

John Krakauer (37m 23s): Yes. So he talks about that. So I think no one wants to believe that what they're doing in his words is cute. They want to believe it's important. And one way to prove what you're doing is important is that you jump up the vertical and you're not just horizontal. These things are very complex. In other words, people don't get much philosophy of science. You know, whatever area you are in, you don't really look over at the other disciplines that have gone through this before. Whether it's ecology or physics or economics. Everyone seems to have to reinvent the mistakes. 

And I think neuroscience is in a particular moment where it's methodologically so powerful then it's okay for it to be a bit philosophically dimwitted for the time being. 

Michael Garfield (38m 2s): There's another thing that comes up for me in that and it was about how there's a way of looking at that paper and saying, this is asking about the selection of hypotheses, like how do you decide what questions to ask? But you co-authored this piece in Neuron and Cell Press in 2021 on the learning salon and you've been doing these, I guess weekly, 

John Krakauer (38m 27s): Well they were weekly and then they went to biweekly and then we took some summer hiatuses, but we're trying to do it once every two to three weeks. 

Michael Garfield (38m 35s): Okay. So yeah, but somewhat regular. And the revival of the salon, which for whatever reason was once very prominent and it fell off. And it's something that I think you and I see eye to eye on the importance of this modality and I would just love to talk a little bit about this particular program and it pleased me immensely to see a paper like this in a peer reviewed journal where you're basically saying like we're having these conversations that flatten academic prestige hierarchies and invite a much more egalitarian participation. 

And you say specifically in quoting Justine Colada about the function of a salon, Colada says “the ideal salon participant was a person who was uniquely interesting and offered fresh ideas that were well communicated and advanced the conversation he or she possessed an innate love of learning, exhibited a reflective intelligence, firmly held principled opinions, but also demonstrated the utmost sensitivity and thoughtfulness towards others.” To emulate the spirit and salons of the enlightenment was your goal. This is something where you've got people that are seeking out different explanations, operating different disciplines, having a preference for different levels of cute, but they're all coming together in service of generating interesting questions is what I got. 

Talk about this a little bit more. 

John Krakauer (39m 58s): The formation of the salon was serendipitous, you know, I should say the salons of the 18th and 19th centuries and before were led by women mostly. And I would like to say that my current co-host, Ida Momennejad, who was the first author on that piece, very much, I should say this again, 2.0, The idea of a salon led by a woman who is not just expert in their own field of science as either is, I mean she's a little bit unusual because she's a neuroscientist and you know now works at Microsoft and so she's a computational neuroscientist, machine learning, she has that perspective, but she's also very interested in philosophy. 

She has a masters in philosophy, she's from Iran, she's multilingual. And so she very much represented that ethos. And then to your point, we both thought it was so annoying that at scientific conferences you never have any time for questions. I mean you can see it in general that you get a talk and then there's almost no time for questions. So no dialogue ever happens. People are together at scientific conferences, but in the moment of the meeting there's no back and forth. It's just like a talk, 45 minutes, an hour, five minutes for questions. 

I mean it's so lopsided. So we said, what if we turn that whole thing on its head, give a talk for 10, 15 minutes and we have two hours of questions cuz everyone wants to have a discussion. And that organ of argument, which doesn't sound like it's ad hominem, is just with it, it's atrophied like an appendix. So we wanted to turn that all around. So what do we wanna turn around? We wanted to turn around that, far more time for questions and then we wanted to make sure that everyone felt courageous enough to ask difficult questions. And you could be an undergraduate, you could be a URM, you could be calling in from India or Africa and you had just as much time as anybody else and there was no such thing as a stupid question. 

If you didn't understand or you took or had a different take on it, we have the time, go for it. Go on a riff. Also, you had people asking questions from vastly different disciplines. In other words in general, you go to your specialist conference, you give your talk and you get a few specialist questions for five or 10 minutes. Here you give your science talk and an anthropologist is gonna ask you a question for 15 minutes. So in other words, it was so different in terms of the ratio of talk to questions, the people who could ask questions in terms of their level of seniority and what discipline they came from. 

So everything was in the mix. And then finally, I mean again, Josh and Ida very much wanted and so did I to make the who you saw be a little bit different. So instead of it being mainly white men, we were gonna actually deliberately alter that. So you were mainly seeing women or people of color as much as possible out of proportion to what you're usually used to seeing. Now we engineered that. You could argue that it was ultra woke.

But I think it was really very important, we all did, that you saw something different in every facet, different times, different people asking the questions, different types of people asking the questions. And basically what that led to was a feeling that you were joining the rebellion that there was this parallel galaxy made up of the learning salon, Paul Middlebrooks podcast Brain Inspired, eLife The Journal, bio archives, science, Twitter, all these leveling mechanisms in science that were beginning to team up against Elsevier journals and tenure and the top universities and the hierarchies and treat your professor with respect and that world. 

So we thought that we were part of this rebellion, this countercultural scientific universe that is trying to make more people get involved, be more about team science, be more about multidisciplinarity, be more essayistic, conversational, philosophical, no sacrifice, rigor, but just mix it up a bit. So in other words, I think there's a lot going on with these alternative venues for scientific discourse. 

Michael Garfield (44m 11s): This is a key thing here. There's another link back to the first piece of our discussion.Iin this, you and your colleagues talk about the second person perspective. This is the first time I have ever seen the second person perspective explicitly mentioned in a peer reviewed like science paper I think ever. You reference the Islamic and Persian philosophy, going back to at least the 10th century of the knowledge by presence. 

You actually start the piece with a, a quote from Hafez. And so this is, you know, the link that I see here is in deciding to lead from observation of behavior again there's like interactions and situatedness and context dependency and a recognition as you put it here “The endeavor of thinking together, sense of participating in a collective mind.” I'd like to hear you talk a little bit about this because this, maybe I'm trying to bite off more than we can chew, but it strikes me that one of the things you know is kind of defining characteristic of complex system science. 

This actually gives us a leg into what I hope we, we have time to discuss in like a third act here. But you know Brian Arthur when I had him on the show and we were talking about how you can look at economics through the lens of algebra or you can look it through the lens of algorithms, through nouns or verbs and you disclose different worlds in that way. And that algebra, seeing it as a collection of objects or in here I'm calling the shot talking about how learning is fundamentally about the future this other paper that you co-authored, collection of nouns doesn't give you endogenous novelty production. Yeah, it doesn't create an opening through which something transcendent can erupt. This seems to me to be a recognition of something that is irreducible and inescapable about knowledge and about learning and discovery. And yet we've somehow neglected this. 

John Krakauer (46m 21s): I really think that you're zeroing on something very deep. I mean first of all, the quotes you're mentioning and focusing on this was very much came from Ida Momennejad. And we had been noticing, and we mentioned in the piece that at a certain point when you get to sort of hour two precisely what you're saying is something collective begins to happen. It's almost like a knowledge sensation of some kind that everyone is swimming in knowledge being generated by everyone. 

Sometimes it even got a little bit drunken cuz the people calling in from Europe, like Tim Barons for example, you know, it was late, they had a drink. We would often have cocktails while we were doing this. So it was almost like a scientific psychedelic, it was almost Jungian. It suddenly hit that strange resonance where even between us as the interlocutors together, we were more than each of us alone. And so you began to be in this space and you know, people have talked a lot about this and David and I have been talking about this recently about science really having been about bands of friends talking all the time. 

So that's where science really came from, it was small bands of friends sharing each other's data. And that happens to this day. And you know, people within a particular domain of science, you tend to know what's coming from your colleagues. So in a way, what you're discussing and what we tried to capture in that article about the learning salon is that in some microcosmic, mystical way, it was recreating the conditions under which science best progresses as a band of slightly drunken friends late at night trying to hash things out. It's almost a demonstration of how science itself should be done, even though this was not a lab. 

Do you see what I'm saying? And other words, I think it, it captures a dynamic that we should celebrate more than we do. And I feel like one of the ways that you get that dynamic to happen is because you have to have diversity. In other words, true diversity, having people from many different stages of their careers, having people of color and women makes that mystical thing to get it spinning more likely. It's almost as though we've missed out on the way that science could feel. 

So I know this sounds a little bit mushy, but I actually do think that something collective and new does happen if you promote it and believe in it. 

Michael Garfield (48m 45s): I don't know if you've read Tyson Yunkaporta, his book Sand Talk, but he bemoans now that he wrote that book before finding SFI and now he's a friend of mine, he's a friend of Jim Rutz. He's much more well acquainted with what's going on here. But I think everything that he wrote in Sand Talk holds true. And you know, he makes this case that basically like he's the director of the Indigenous Knowledge Systems Lab at Deacon University in Melbourne, and he says the indigenous have no conceit about you talk about closed systems, like show me one, their angle on complex systems is one that is practical, it's rooted in experience and direct observation and it hasn't been abstracted and you're not trying to do the physical. First let's start with a spherical cow. 


Michael Garfield (49m 35s): Right? So I think that maybe that's part of it. It's my experience that two dudes get together and fly off into the abstract and imperion and this system's knowledge that's grounded in praxis is key. And that, you know, Tyson was talking about how the indigenous Swedes, who are, you know, blonde haired and blue eyed, and he is like, but these people were blacker than I am because they are still rooted in it. They're still connected to the land and they're suffering from colonization and so on. They've still got the thing. 

John Krakauer (50m 5s): It is trying to concentrate the thing and try and bring back the rigor, the joy and the mystery in the scientific enterprise by realizing that it is about friends and teams and about different areas of expertise. Very SFI ish, but you know, it's still not the norm. In other words, as I said, you have to have these mushrooms that have sprung up in the forest. And I like to think that the learning salon is one. I've got and SFI is one, a much bigger mushroom, but there's something to it. And I think that we need to fight for a different way to do science. 

Michael Garfield (50m 40s): Thank you for listening. Complexity produced by the Santa Fe Institute, a non-profit hub for complex system science located in the high desert of New Mexico. For more information, including transcripts, research links, and educational resources, or to support our science and communication efforts, visit