Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Episode 52: The IIT controversy with Felipe De Brigard
117 Plays1 year ago
We discuss the recent controversy about Integrated Information Theory (IIT), a theory about the neural correlates of consciousness, with Felipe De Brigard, a philosophy and psychology professor at Duke University who signed a letter describing the theory as pseudoscience.
The letter: https://osf.io/preprints/psyarxiv/zsr78/
The adversarial collaboration: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0268577
Description of IIT: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1011465
Recommended
Transcript
Introduction to IIT Controversy
00:00:07
Speaker
Welcome to Cognation. I'm your host, Ralph Nelson. And I'm Joe Hardy. On this episode, we're discussing the recent kerfluffle over integrated information theory, which is a theory on consciousness that was first described by Giulio Tononi in 2004. Recently, a group of prominent scientists and philosophers signed a collective statement entitled the integrated information theory of consciousness as pseudoscience that will form the basis of our discussion today.
Felipe de Brigard Joins the Discussion
00:00:37
Speaker
Our guest is Felipe de Brigard, an associate professor of philosophy at the Institute for Brain Sciences at Duke University. Felipe studies the philosophy of neuroscience and is published extensively on cognition, memory, and consciousness. He also happens to be a signatory on the statement against integrated information theory, and he wrote a sub-stack post about his position that we'll link to on the show notes. So thanks a lot for joining us today, Felipe. Thank you very much for inviting me.
00:01:07
Speaker
Great. And before we get into the weeds talking about integrated information theory, what it is, and it can be a little complicated, maybe we could start out saying a little bit more about your background and your interest in consciousness and how you view the state of the field today.
00:01:24
Speaker
Right. So as you mentioned, I am a professor in the philosophy department, but I'm also a professor in the psychology and neuroscience department. And I also have a lab in the Center for Cognitive Neuroscience. So I am one weird philosopher that does imaging and uses a bunch of techniques from cognitive neuroscience and from cognitive science. I have been interested in straddling the two disciplines for as long as I can remember. I
00:01:52
Speaker
major in the Universidad Nacional de Colombia in Bogota. I majored in philosophy and I also did neuropsychology. Back then I thought that that was the only way of understanding how the mind and the brain would interact through neuropsychology and to a certain extent I still have a huge amount of my love goes to neuropsychology.
00:02:11
Speaker
And then I did a master's at Tufts University working under Dan Dennett, and it was impossible not to fall in love with consciousness when you're working under Dan Dennett. So I was very interested in that at the time.
00:02:25
Speaker
And then I did my PhD at the University of North Carolina in Chapel Hill, and I was doing sort of a parallel PhD, one in philosophy and one in psychology and neuroscience. My advisor in philosophy was Jesse Prince, who was at the time also working on consciousness and in particular the interaction between attention and consciousness.
00:02:44
Speaker
And so, and we're talking, you know, 2006, and back then it was, you know, a really hot topic. What was the relationship between attention and consciousness? I thought that I was going to probably do research on that. In fact, my first lab rotation was in Joe Hotfinger's lab, who's an attention researcher, and he was sort of interested too in attention and consciousness.
00:03:11
Speaker
But as I became, I guess, a better cognitive scientist, I became more and more frustrated with a lot of the terms of the debate in consciousness and attention literature. Because there were some issues that seemed to me
00:03:30
Speaker
terminologically in the bad sense, not in exactly what we mean by a term, but rather in exactly how we are using the term relative to a particular operationalization in an experiment, which are two very different things, right? So here is one example that I can give you that probably resonates with your interest in vision science.
Issues with Consciousness Paradigms
00:03:55
Speaker
So we kind of know that there is reentrant activity in V1, right? That you have the information coming, you know, through your eyes, optic nerve, probably what, 780 milliseconds after the stimulus onset, you get V1 activity, then it moves forward, and then we know that there is some sort of reentrant activity coming back.
00:04:16
Speaker
The temporal window of that activity is essential to understand the processing, right? But there were some of the members of the debate that were kind of not paying attention to that temporal distribution. So they would claim that attention was something that happens only when you have your eyes being directed toward the target.
00:04:39
Speaker
And then as a result, if you understand attention under those terms, then it looks at doing certain phenomena that you're conscious of, then attention is not necessary. But by contrast, you can have other people that define attention with some other, another experimental paradigm in which attention is deployed over a much longer period of time.
00:04:59
Speaker
And then you have now certain activity that is reentrant on V1 and then certain things that under one paradigm sounded like not attention and therefore, you know, not necessary for consciousness or whatever. And there are another paradigm in which attention is differently operationalized. It turns out that it is necessary.
00:05:16
Speaker
And that was a lot of frustration in that sense because it's not even that there is some clarity on how the term is used, it's that the terms become sort of piggybacking on particular opportunities that are highly dependent on particular measures that are highly dependent on one particular experimental design.
00:05:37
Speaker
and that makes the really, you know, it's lacking of what male calls, you know, criteria for construct criteria, right? So it means it was frustrating for me. So I started to gravitate more toward aspects of cognitive science that had
00:06:01
Speaker
I guess, better measures. I'm not saying ideal measures, but I'm saying measures that I could trust a little bit more. And then that's why a lot of the reason why I ended up doing cognitive science of memory.
00:06:14
Speaker
Yeah, that's really interesting. I mean, I guess just a little bit of just terminology. So Felipe mentioned V1. That's a part of the brain that neuroscientists refer to as visual area 1. So V1 is what we call the primary visual cortex as part of the occipital lobe. But yeah, I mean, I think that that's really
00:06:39
Speaker
a good point that just the kinds of ways that people operationalize terms and what they mean by them can really influence a debate when you start talking about, well, attention depends on this area of the brain or attention operates in this way or there's this relationship between consciousness and attention. How you feel about that, how you think about that really depends on what you mean by attention.
00:07:09
Speaker
and the other terms that you're using. So yeah, that resonates with me as well.
00:07:15
Speaker
And I do remember, at least in the first edition of the Vision Science textbook, under the definition of attention, it said, see consciousness. Under the definition of consciousness, it said, see awareness. And under the definition of awareness, it said, see attention. So there's this circularity and this changing of definition of all these terms all the time. It can certainly get confusing. So let's get on and maybe talk a little bit about
00:07:43
Speaker
What integrated information theory is now. How did you first run into this idea and what were you, what were your first takes on on this when you ran into it fully bank.
Skepticism Towards IIT's Complexity
00:07:55
Speaker
Right, so, um, So as I said, at that point, we're talking
00:08:03
Speaker
2006 perhaps, 2007. And the theory would have come about just in 2004. 2004, right. So, I mean, I was voraciously reading everything that I could put my hands on, like not only on attention, but like everything philosophy and everything philosophy of mind and so forth. So I led last year, we're like the classics, right? So I led Churchland and Denon and
00:08:28
Speaker
and even some stuff that was weird, like Sultan Torrey's theory, which no one has read, but I thought it was very interesting because the author being blind, I thought it was very intriguing to see how a blind person would come up with a theory of consciousness, given that a lot of the theory of consciousness that we have come from examples on vision. So I was just voraciously reading everything. The IT, it's short for information integration theory,
00:08:59
Speaker
was starting to sort of make the rounds a little bit but it was more my understanding or my feeling at least as a graduate student in philosophy at the time is that it was more of a sort of a big thing in a small group. So it was a new theory but only for for a very select group of specialists and part of the reason I think that it was felt as a very sort of specialized theory is because it was
00:09:29
Speaker
are shrouded with an enormous amount of mathematics. And so it looked technically really, really hard. And so only the most brilliant minds could get to really the core of this very difficult truth that they're so difficult that English doesn't serve the purpose in communicating that idea. You really needed to know this very complex mathematics.
00:09:59
Speaker
So, so that's sort of how I understood it, and I, you know, and I wasn't going to put the time into trying to understand the formulas for what it was supposed to be my hobby at that point, rather than my main area of research.
00:10:18
Speaker
Sure enough, at that point, I was less interested in going to conferences on attention. I didn't go to Tucson, didn't go to the Association for Dissentive Study of Consciousness and so forth. I was going to the Society for Philosophy and Psychology, and there were some issues on consciousness and attention, but it was not specifically on consciousness. But the sort of the rumor that I started to hear, and we're talking now perhaps, you know, 2010, 2011,
00:10:48
Speaker
was that people were starting to say, well, this is kind of mathematical mumbo-jumbo, right? It is kind of like this fancy, very technical stuff, but without a whole lot of content to it. But that was a rumor. It was just, you know, people would say this in murmurs. And I was like, well, you know, there's a lot of theories out there that's used
00:11:12
Speaker
weird metaphors to explain things. I mean, I don't think that multiple drafts is necessarily more objective than some mathematical formula. So I was willing to suspend a lot of judgment on whether or not what we have here is like a mathematical metaphor, if you want, something that is kind of like a model, but it's not exactly like a computational model, something in between.
00:11:40
Speaker
And I didn't know enough about it, to be quite honest, nor did I want to delve too, too much into it. Until 2014, when Aronson published that scanting sort of blog entry, where he basically shows using the math how you end up
00:12:00
Speaker
using the axioms and the mathematical formalisms of the informational integration theory, you end up having to say of certain systems that you would never say are conscious, that they are conscious. And it was, you know, that math was easier to follow. In fact, a lot of, it saved me quite a bit of time because at that point I sort of understood more what the math was. And it's also relative, it's just that I have problems with the notation
00:12:29
Speaker
even the notation used in the book and in the first papers. Like if you have a student that was saying the other day is like, this is probably not notation created by a mathematician. So, but anyway, so I thought that that blog post was kind of like the final, you know, counter argument and that people are going to see what IIT was for what it is, right? It's a theoretical, it's a mathematical formalism that just didn't work out. Period.
00:12:58
Speaker
So that's how I came and I remember, you know, spending some good time reading what was going on in 2014. And I'm just feeling at that point that this is just another theory. Cool. Maybe we learned something about it. It was an interesting perspective to see consciousness from. But I didn't think of it as being very promising at that point.
00:13:24
Speaker
Let me step in here to, and we can talk just a little bit about, so you've mentioned a couple aspects of the theory. First of all, that it's mathematically specified, or that there's some complex math that goes into it. Maybe I could read a quick description of it so that people get a high-level
Understanding Integrated Information Theory
00:13:43
Speaker
overview. And this is from Tanoni's website. So I'll just read a quick paragraph about this. Okay, so about integrated information theory.
00:13:53
Speaker
So consciousness corresponds to a system's capacity to integrate information. This is indicated by two key phenomenological properties of consciousness. Differentiation, the availability of a very large number of conscious experiences, and integration, the unity of each such experience. The theory claims that the quantity of consciousness is determined by the capacity to integrate information, which can be measured as the phi value of a complex of elements.
00:14:21
Speaker
fee, and this is the main term that gets used in IIT, fee is the amount of causally effective information that can be exchanged across the minimum information by partition of a complex. It's informational weakest link. Okay, and then there's a little bit more on fee there. But I think that gives the main idea behind IIT. And they also, so the latest version of IIT, it's gone through a couple different iterations. There's a paper that's
00:14:49
Speaker
Recently out, that describes IIT 4.0. It includes some other baggage that comes along with it. So I think, what is it, five postulates or five?
00:15:02
Speaker
five things that are sort of part of it. And those aren't necessary, those are sort of relating it to consciousness, not necessarily mathematical descriptions. Yeah, I think it's, it's worth also probably even taking one step further back to ask the question, like, what is a theory, like IIT, integrated information theory trying to solve? More specifically in this context, like, what, what are some of the problems that are trying to be addressed?
00:15:28
Speaker
by the theory. I think one of the things that seems to be trying to be addressed is what is the relationship between the brain and consciousness? So there's this idea that you've got a system in the brain instead of neurons and other biological systems that work together to
00:15:55
Speaker
create your conscious experience. That's like one way of viewing the problem, for example. And a theory of consciousness that relates that brain function to conscious awareness is a theory that
00:16:13
Speaker
one might be interested in developing that might lead you down a road that would ultimately become a thing like IAT. It's sort of in that world. Does that make sense? Am I saying something that? Yeah, I mean, IAT is similar in certain regards to certain other theories in the offing, but I think it is importantly different and kind of unique in a different sense, which is
00:16:40
Speaker
that it takes consciousness as the given starting point as opposed to the brain.
IIT: Metaphysical vs. Empirical
00:16:54
Speaker
Instead of thinking about how this part of the brain might relate to consciousness, let's start with what we know about consciousness.
00:17:05
Speaker
One way in which I like to think about it, and this is perhaps another sort of side way, I'll get back to IIT in just a second, but is that there was another moment in which I thought about IIT very carefully. And I talked about this in my entry on Substack, which was the first and only time which I have met Tononi, which was at a conference in December of 2018 at the Max Planck Institute.
00:17:29
Speaker
And Tomoni was one of the speakers there. And I was really interested in talking to him about IIT and about his theory because I was not willing yet to think that it was just mathematical mumbo-jumbo, that there might be something very interesting behind it.
00:17:46
Speaker
And I must confess that I was on the one hand disappointed, but on the other hand, I was quite satisfied with my conversation with him because what I ended up with the impression is that he is kind of like a rationalist philosopher.
00:18:01
Speaker
uh rather than an empiricist philosopher trying to talk about consciousness so i mean you can imagine him kind of like the card in the meditations going uh to like forget about the world forget about everything that we know how can we get back to something we're completely sure about
00:18:21
Speaker
our thinking. So for Descartes was, you know, that I think that it's undubitable that I doubt and doubt is an act of thinking. So it should be very clear that I think, right? Now, you know, I don't have to get into those details, but we know from the history of philosophy that that conclusion is probably not warranted.
00:18:40
Speaker
But I think that what to not only want to do was something similar to unconsciousness. When I reflect on my own conscious state, what do I get that is indubitable? And according to him, this axion,
00:18:55
Speaker
are like the basic building blocks of our experience, our conscious experience. And there's things that we need to give for granted and accept as self-evident, right? And it is that what becomes the explanation. In other words, that is what we want to explain. And his axioms, he has five axioms. One is like consciousness exists intrinsically. He also second is composition. Consciousness composed, structured,
00:19:24
Speaker
The third axiom is consciousness is supposed to be specific and there is a particular way in which consciousness is and he thinks that that is called information. There is another action which is integration and then there is finally another action which is explosion. At this point part of what happens and then he was explaining to me because we had a very nice dinner and he was sort of explaining to me and quite honestly what I was seeing was sort of a metaphysical view about consciousness.
00:19:53
Speaker
what he does, my understanding for what I've seen, what he does is that he tries to give a mathematical apparatus that according to him would sort of account for these axioms, right, that are self-evident and in so far as they're self-evident they're supposed to be so irrefutable about our conscious experience. So it's a very different way of going about like if you poke this thing in your brain then your consciousness is
00:20:21
Speaker
below certain threshold. And if you poke this, it's going to be above certain threshold. Or this part of the constant. It's very different. It is sort of a top-down approach, if you want, rather than the bottom-up approach. And that makes IIT unique. But at the same time, it makes it uniquely problematic.
00:20:42
Speaker
What makes it uniquely problematic is, first of all, those actions are, as many people have argued, far from being self-evident. Those actions are really difficult to understand in many ways, and others have argued that are profoundly ambiguous. So one way in which I like to think about this is you have this metaphysical view of a consciousness, and then there is a mathematical model
00:21:06
Speaker
that is supposed to relate to this metaphysical view of consciousness. And then there is the reality which is supposed to relate to this mathematical model, right? And the gaps between these three levels are ginormous. The amounts of degrees of freedom and post hoc possible explanations between those levels are ginormous, right? So this is where I started to think
00:21:29
Speaker
Look, IIT should be considered a metaphysical theory. There are metaphysical theories. Philosophers come up with metaphysical theories about consciousness all the time. So if it is to be a theory at all, it is in that sort of, that's the domain in which it should be argued with. And maybe if we clarified the theory and so forth, and it is not without metaphysical confusion and so forth, maybe we can see how that could potentially be translated into something that we could find in the world.
00:21:57
Speaker
But for me, that set it apart from other more scientific theories, if you want more empirical theories. So to me, the appeal of a theory like this is, well, it may not be right. At least it is an attempt at trying to make a determination about what consciousness is in any kind of system.
00:22:27
Speaker
It's a way to have a consciousness detector.
Critique of IIT in Biological Contexts
00:22:30
Speaker
So you could say, is this thing conscious? Is this thing conscious? And looking at the information and the integration of information, being able to say, make a definite answer. So I think maybe it eases people's mind about all that gray area in consciousness, right?
00:22:48
Speaker
that you can make a definite statement about what is or is not conscious. Right. But I mean, that would be amazing if you say like an absolute number, then you could see, you know, you can see whether a certain thing is conscious like. And you get a number, you get that number to go right along with it. You can say this, this thing is more conscious, this thing is less conscious, it seems so quantifiable.
00:23:13
Speaker
Yeah, yeah. And then maybe it will make you feel better about like treating your stove more harshly than your microwave or something like that. But I mean, that sounds in principle really nice. The problem is that in practice is, I mean, many have argued and I have explained in fewer words, why end up being intractable?
00:23:38
Speaker
I mean, it already becomes intractable in systems that are way easier than the brain.
00:23:45
Speaker
In addition to that, it makes really, I think, absurd oversimplifications of how units in the brain might work. One of the things that I remember pushing the money about and that I have yet no clarity over is what are supposed to be these units that are engaged in these causal interactions? Because if there are neurons, then we simply don't have the data.
00:24:08
Speaker
And I'll get to the neurons in a second. And if they aren't neurons, if they're like more rough measures, you know how difficult it is to determine real causality when you have something like electrodes. I mean, you don't know what is really the causal factor relative to the background factors and so forth. The evidence and the instruments aren't even there.
00:24:34
Speaker
And even suppose that there are neurons, why to think that neurons only do, like, are on and off. I mean, this is McCulloch and Pitts, and everyone should at this point in age agree that McCulloch and Pitts might have been brilliant. And it was lovely what they did, but neurons are not these thresholded things that only do one thing and go to bed when they are not triggering. That is just not how the brain works.
00:25:03
Speaker
over four different ways of interacting and communicating information. Neuronal gap junctures, RNA transport, many of which are non-synaptic. Neurons are doing all sorts of other things when they are not triggering this stuff. In addition to that, it is very likely that glial cells, in addition to neurons, are contributing to cognition in ways that are yet to be understood. So even if
00:25:32
Speaker
Even if we could measure each individual neuron and somehow derivate the causal input of one neuron relative to another one, which I think is already absurd to think. We're talking about millions of synapses at the same time. It is unclear that that is the right level of unity, the right for determining what the unity in the system is.
00:26:00
Speaker
it is not that it is difficult to understand how it could be implemented in a brain like ours, is that it is likely the wrong model for how it is to be implemented in a brain like ours. There's lots of views that would make it very attractive because there's a threshold, there is a number. It would be really nice if we can organize everything in the universe according to phi.
00:26:28
Speaker
practically, like in terms of when you really see how this could be implemented in biological system like ours, and sadly, this is the only biological system that we know is conscious. So if it's our best case, test case, then it just just breaks down. Well, yeah, I mean, I think what you're saying there is that the theory itself
00:26:52
Speaker
as a computational theory doesn't accord perfectly well with our understanding of how computation is carried out in the nervous system. And there's so much that we don't know about how computation is carried out in the nervous system. Like neurons are only one part of it, right? There's all these other types of cells that are important. We don't even know for sure that just the brain is
00:27:18
Speaker
is required. It could be that the other parts of the body are required as well. But notice that I tried to reframe myself for calling the mathematical model a computational model because I think they are refraining themselves from calling it a computational model as well. If it was a computational model,
00:27:36
Speaker
then there are ways in which it could be tested. So I like computational models of brain behavior interactions. I think I use an example in my sub-stack right up. I have seen lots of computational models where they try to fit dopamine as one specific factor.
00:27:57
Speaker
But notice how computation modeling, you have a particular behavior, say a choice behavior between A and B when A and B have different values, right? And you find that there is some kind of, I don't know, quadratic relation between the values and the rewards or something like that, right? And then you're trying to figure out what part of the brain
00:28:26
Speaker
could be sensitive to the parameters by means of which this behavior, you know, which this behavior could be fit, right? So you have a formula, you create a formula, this is a mathematical select. So we generate this particular mathematical formula that has these parameters. Is there anything in the brain that say functions as this parameter that varies quadratically in relation to a particular reward value?
00:28:56
Speaker
And then you go and probe and sometimes you find that, yeah, law and behold, actually, the level of dopamine in the destroyed atom fits perfectly well with this one parameter in my computational formula, right? So that's how the models are usually generated. I mean, this is a cartoon, but this is kind of like the idea. The idea is that you have some somewhat objective measure of some brain change that can be fit to a particular parameter in a computational model.
00:29:25
Speaker
I don't think that that's how IIT math works. If it was that the way in which it worked, it would be great. But I don't think that that's how it works. When I was arguing with people about this particular point, they told me, oh, well, you know, like you are an outsider, you don't know that there are all these
00:29:45
Speaker
these actual experiments that have shown that there is this measure of integration. So they mentioned this measure, I forgot, it's called the perturbation integration index. And I was like, oh man, like it looks as though this perturbation integration index is in fact a mathematical measure. And when you read the paper, it looks as though it's not different in time from like this experiment that I just told you about. But when you go and look at the math,
00:30:15
Speaker
We have that perturbation integration index. Well, if I understood correctly, it is a measure of compressibility of matrices. I mean, people do this when they work on the network neuroscience all the time.
00:30:31
Speaker
in which you have these massive amounts of data, so that you create a matrix of cross correlations between each specific voxel correlates with another voxel, and that gives you some correlation number. And then you want to have some dimensionality reduction measure. And there are plenty. Each lab uses their own and so forth. So it didn't seem that different to me.
00:30:57
Speaker
In no way this perturbation, of course, when you zap somewhere and your brain turns out that the clustering breaks down and it is no longer counted for by one latent variable or something like that. So that's where the papers, but they don't derive from a mathematical formula, right? There is a lot of different reasons, biological reasons as to why is it that
00:31:21
Speaker
that the brain in those situations organized in that particular way, none of those are directly derived from the mathematics, right? If the math of the information integration theory was such like the math of a drift diffusion model of dopamine energy, then I would see what they are doing and I would put it sort of on a par with that. But I just don't think that they are on a par. Even this attempt to gather the empirical measures aren't really
00:31:50
Speaker
that clearly mathematically related to the mathematics, the formalisms of the theory. Yeah, that sort of gets to this question about the adversarial collaboration that sort of sparked this whole conversation from the beginning, which was, as I understood it, there was a group of people from a couple of different theories of the neural correlates of consciousness from their perspective. That's what they were representing it as being.
00:32:20
Speaker
One being this global neuronal workspace theory and the other being this IIT theory. Proponents of each of these got together and they made some predictions about what should happen under a variety of different circumstances and perceptual experiments, essentially, which they relate to conscious experience in some way. So they're basically saying,
00:32:41
Speaker
IAT says that this area should be active at this moment with this time course. And then global neuronal workspace theory says these different areas, particularly the frontal lobe, should be active and involved in this way. And so they make different predictions, essentially.
IIT vs. Global Neuronal Workspace Theory
00:33:01
Speaker
And so I guess the question is,
00:33:03
Speaker
Do you think that that's true? Do you think that IIT in particular does make those predictions quantitatively? And if so, how does that relate to what you just said? Or if not, is that what you're saying? Yeah. So one way of thinking about this
00:33:25
Speaker
this so-called predictions, is using a framework which is slightly different. The framework that perhaps I would like to use is a more recent framework that Hakun Lao put together in his book on unconsciousness we trust, which is his latest book on unconsciousness. And he divides theories into sort of more
00:33:52
Speaker
global theories in which consciousness sort of happens at the interaction, like more like in the association criticism, something like that. But there is like this long lasting sort of connections versus more localist theories in which you have something about where the information is processed that that that matters. So it's a little bit more local, let's not broadcast it. And my
00:34:18
Speaker
feeling is that this sort of adversarial collaboration kind of falls in this dichotomy as well. So you get, of course, the blown neuron and workspace hypothesis that takes this sort of like cortical connections to be really essential for consciousness. Whereas the informational integration, what matters is the degree of integration and of exclusion in other parts of the brain.
00:34:46
Speaker
So presumably what you're going to get is these hot areas, I think, that they color or are more local. So to a certain extent, the predictions that they were making were going to be fall in either one area or the other. Now, the problem is that, first of all, those are not unique predictions, it seems to be, for each specific theory.
00:35:16
Speaker
higher-order theories of consciousness can be, you know, sympathetic to some of the globalist predictions that they were, even though higher-order theories are not similar in important regards to the global neuron of workspace hypothesis. There is a theory that I really like out there by Michael Glaziano. And for instance, he's the theory of the attention schema.
00:35:45
Speaker
And there is a sense in which I can see how the attention schema could, one could think, because the attention schema is that you have sort of a degenerate and attentional model of the situation in which you are in when you are engaging in some sort of stimulus responsive. So it's going to be less, and I think that the animal says in these terms, it's going to be less of what is the neural correlate of consciousness.
00:36:08
Speaker
at all, but what is the neural correlate of a conscious person at the time? So if you have a conscious person that is visual and stuff, I can see how that theory, Graciano's theory, could make you predict that you're going to have something like a localist response.
00:36:25
Speaker
So in other words, the point is that even if you get super clean results with these so-called predictions, they're still underdetermined by theories in the sense that there are other theories that could account for them perfectly well. So that's one thing. And then the second concern with this sort of so-called predictions is that they are very vague. Mind you, I didn't know much about the adversarial collaboration, the details of that adversarial collaboration,
00:36:54
Speaker
until after, basically, when I was asked to sign the letter and I read some, partly I read it because I'm friends with Lucy Ameloni and I respect her an enormous amount, which is one of the PIs of this collaboration. And I share some of the documents with me. And part of what I was thinking is, well, these predictions aren't really risky.
00:37:21
Speaker
in a sense. And my sense is what is going to happen is that the data that is going to come out of this experiment is not going to be super clear.
00:37:30
Speaker
And that's exactly what happens. It's not, you know, when you have predictions that are relatively vague, then the data has to be interpreted. And that's a problem, it seems to me. When you have room for interpretation, each one of the theories is going to interpret it as, well, you either didn't do X or X happened because we didn't think that this other thing could have happened. And that's, it seems to me that that's a lot of how this discussion has evolved.
00:38:00
Speaker
So the results of the experiment are not as clean cut as people seem to have made them out to be too. Yeah, yeah. Because they I think they make clear in the paper itself that here I'll read a section here. These results confirm some predictions of IIT and GNWT that is global neuronal workspace theory.
00:38:24
Speaker
Well, substantially challenging both theories. So they are aware that this isn't a clean win for one or the other or, you know, maybe as people. Right, right. I think that that happens. And then that explains to how, you know, first of all, if this
00:38:44
Speaker
sort of unclear what the purpose of the Adversarial Collaboration was, if not to have something like an experiment that both teams fully agree that it was going to either support or, you know, turn down the theory. But part of my concern, you know, after this, Paul Brouhaha, I have learned more and more about the process. And
00:39:09
Speaker
And it turns out that the adversarial collaboration doesn't seem to have been proposed as a single experiment or maybe like four experiments that we're going to support one theory and another. But it was supposed to be something else, like some sort of exercising conversation or maybe increasing evidentiary support for one theory versus the other as opposed to be like the one experiment that, right?
00:39:39
Speaker
But I think both sides, both ways of thinking about the other side of collaboration, either as a single experiment that is knocked down or as more a conversation sort of thing, are pretty sort of problematic. On the one hand, you know, science just doesn't work by a single reputation by experiment, right?
00:40:03
Speaker
Even the best examples that we have of allegedly single experiments that have refuted a theory are very different from what happened here. So I'm thinking about Eddington's and Dyson's experiments of the curvature of light that supported a prediction of the general theory of relativity. So for the artists that sat there, just basically the idea is that they
00:40:32
Speaker
if light really did have mass, then it should be curved when there is an eclipse in light of the sun. So the general theory of relativity have a particular prediction of what the angle of that curvature was going to be, whereas the Newtonian had a different prediction of how the light was supposed to go.
00:40:51
Speaker
that and then they did the three measures but this is actually quite complicated because it turns out that some of those measures didn't coincide with others and there is a lot of sociology involved in why Einstein ended up being right. But the important point for this one is that the predictions unlike what happened with this adversarial collaboration
00:41:10
Speaker
were supremely risky, right? They predicted that to the degree of angle, and then if there was some movement in the angle, which was what, in other words, if what gets recorded is not exactly what it was mathematically predicted, then there is some theory of error as to why it might be possible that the curvature of the lens was such that it diverted the light or something to that effect.
00:41:33
Speaker
This is not at all what we have in the case of consciousness. In the case of this predictions, they were neither risky
00:41:41
Speaker
Nor was there an acceptable, because we don't have it, an acceptable theory of error. I can always say things like, well, I would have found more integration had I recorded from single cells rather than recording from a population of cells. Or maybe if I recorded it for intracellular space rather than for extracellular space. Or maybe if I had used a different TR in my fMRI.
00:42:03
Speaker
Like, you know, all these things are being arguing. They also fit a lot of models to the to the data and the model fitting was all based on different algorithms that are evolving also in the space. Exactly. So there is that. Yeah. So that was not what happened with that adversarial collaboration. So what happened, it seems now I have come to learn later, is that it was not supposed to be like a unique experiment and so forth, but it was supposed to be more like sort of maybe
00:42:32
Speaker
and even like an issue of evidentiary
00:42:35
Speaker
Wait, like given these experiments, which one was going to get, but of course, but not, not a hold up one hand as the winner. That's right. That's right. But of course you're dealing with people and you're dealing with, with scientists that had made a reputation and a scientific career of holding to opposing allegedly opposing theories of consciousness. Do you think that they really going to sit together and say like, Oh, let's come to agreement now. I mean,
00:43:04
Speaker
Do you think that at some point, if the evidence came in one direction, that Tononi is going to say, like, OK, hold the presses. No more issues of fire are going to be published. Or that Dehane was going to say, like, that's it. I am going to write a book. I said, like, how I was wrong about my book on consciousness. So no, that's not going to happen. We do that precisely because there is so much room for post hoc constructions.
00:43:31
Speaker
that there's going to be endless ways in which people can explain the pattern of results. Yeah, I mean, my, you know, I guess the question becomes a little bit then, you know, in terms of the back to the controversy for a second. I think where people got
00:43:49
Speaker
excited about what was going on both from the positive and the negative side was when they saw headlines that basically said there was one headline I think that said like neuroscience zero philosophy one or something like that right yeah yeah and uh basically saying you know uh this was like a major event in in the the field of consciousness research and I guess maybe that that was maybe more the way that it was presented
00:44:19
Speaker
there that I feel like people were objecting to than necessarily the theory itself, which I think you've presented a lot of problems with, but at the same time, I think all of our theories in neuroscience have similar problems in other ways, right? So especially when we're trying to relate brain function to conscious experience, there's really no actual theory of the hard problem.
The Hard Problem of Consciousness
00:44:47
Speaker
at all, including any of these that even, like, to my mind, even purport to speak to the hard problem of how you actually transfer something from a material existence to a conscious existence. Yeah, no, for sure. I mean, ultimately, many of these theories, even the information integration theory, are sort of description of the phenomenon is normally correlated with consciousness. But that doesn't mean that it solves the hard problem of consciousness.
00:45:16
Speaker
Even if you think, suppose that the information integrity theory is true, that somehow there are these units, whatever they are, and that it just so happens that when they lead to a certain level of phi, although for what I understand, and I think that I talk a little bit about this in the sub-stack, it turns out that there are multiple solutions to certain structures, and then there is a little bit of cherry-peeking what's the phi. I call that phi hacking.
00:45:47
Speaker
And assuming that that is not the case, assuming that there is a unique solution to each cause of the structure and that somehow all the ones that have a certain level of phi, whatever, are conscious and that for what miraculous reason the human brain happens to be one that you get like phi and maybe the Martian that is really nice to me on Christmas also happens to have a really high phi.
00:46:14
Speaker
but my stove doesn't. For some miraculous reason, I think that there's still a question as to why is it that when information integration is like this, why is it that it feels the way in which it does? For me, even information integration theory is not a solution to the hard problem. There is a sense in which
00:46:42
Speaker
I don't know what's the right word for that, but I'm not a mysterious in the sense that I don't think that the hard problem of consciousness cannot be solved by science. I am more like, and I don't know what the right term is, but I just don't think that we would notice if we have solved it, right? My sense is that the solution to the problem of consciousness is going to be something of the order of, you know, we,
00:47:10
Speaker
describe functionally everything that happens in in our brain and then someone comes with a theory like I don't know maybe they they they uh the graciana's schema attention and schema theory and then you say like well when you have a schema that is like this then it just is a brute fact of how things are wired and it feels the way in which it does just like when you have uh rocks with sediments that
00:47:39
Speaker
It just happened to be that the sediments have this particular form, but there is no explanation as to why is it that there is a crack right here. It's just, sediments are like that, right? So I feel, but we wouldn't know, right? Because I think that we always feel that there is a gap. There is actually some work here and saying that the gap is more like a psychological. So even if I come up with a theory that might be right,
00:48:03
Speaker
I'm still going to feel unsatisfied in my heart problem of consciousness. That just always seems to be the case, no matter what theory you could possibly come up with. Someone can always say, but it doesn't tell me why I feel pain, or it doesn't tell me why I see red. I agree 100%. When we were talking with Bernie Barres, he was describing the global workspace theory.
00:48:32
Speaker
some really interesting evidence that goes into his thinking behind that. But at the same time, to me, it doesn't even come into contact with the question of like, OK, so if you have to have this global workspace in order to process all of this complicated information that we do as people and maybe other animals do as well, why does that have to lead to the
00:48:57
Speaker
that I would have an experience of that. That doesn't even touch it. Could you have something that's descriptively interesting without specifying anything causal about this? I think that many of these theories are descriptively very interesting. I mean, my full disclosure, up until 10 years ago, I was a fully convert into global neural network space hypothesis.
00:49:28
Speaker
bits of scientific work or philosophical work or whatever that I did on consciousness was to try to relate the global neuronal workspace and hypotheses to the experience of conscious recollection in episodic memory retrieval. Right. And it was so that was that was sort of my last attempt. I really I really liked the neural workspace hypothesis linked with some things that I have read about a pension and and broadcasting of information and so forth.
00:49:56
Speaker
Mind you, all of those are metaphors, right? All those terms. Broadcasting, what do you broadcast? It's like when people talk about attention and I say like, you know, attention of resources. What are the resources? Is it metabolic resources or is some kind of, you know, currency brain uses that I don't know of? So anyway, so I mean, that's why I read them, right? Because they're descriptively very interesting. Look, in my, in my view, I am, I am a very
00:50:25
Speaker
austere neuroimager or austere cognitive neuroscientists. I believe that the progress in our disciplines, and it happens for memory, but it probably happens for consciousness as well, is very clumsy and very confined to relatively small findings, right? So I am an experimentalist and I love experiments that are very well, sort of, we're limited and very tight.
00:50:55
Speaker
and very focused and as controlled as possible where you have as much control over the independent variable as you can. That's why I like psychophysics so much, perhaps, because you feel that you really have, it's tiny, like the advancements are tiny and they understand how the specific observations are going to fit with a larger theory. I think we're still far from it.
00:51:19
Speaker
But we advance by smaller observations. We are like, in a sense, we are like the astronomers that recorded very specific observations. Whether there is going to be a Kepler or a Copernicus or something in the future that is going to come up with a larger theory, I don't know. I think we are in the stage in which we should be concerned that our observations are solid.
00:51:48
Speaker
Yeah, there's certainly a replicate replicability problem in neuroscience and psychology for sure. So that that's, that's a very, a point very well taken. Yeah.
Skepticism on Unified Brain Theories
00:51:59
Speaker
And I don't know, like, and this might sound kind of counterintuitive to certain years. But
00:52:06
Speaker
I don't know that we need a theory. I am very skeptical of theories of everything in the brain. The brain is profoundly weird in the sense that it evolved for all sorts of different kinds of biological, to solve all sorts of different kinds of biological and evolutionary challenges, and it's very clunky on top of another.
00:52:32
Speaker
that everything happens because every part of the brain is trying to reduce error, not sure. I mean, we know that the neural anatomy is not such that there is always an independent challenge of error signal. Well, and as you say that too, you can see the, I mean, you see the appeal to simplicity and there's almost nothing simpler than the basic idea of integrated information theory compared to the actual complexity of the brain. Yeah.
00:52:59
Speaker
Yeah, it reminds me, there is a paper, I forgot what the title is, by Shannon, the Shannon information. He wrote a paper warning people about overextending the notion of information theory and using information theory to stuff that wasn't initially applied for.
00:53:18
Speaker
And I think that when you find theories that work, that are beautiful for specific areas, they appeal to try to get them to apply for everything else. It's very difficult to not fall in love with that appeal. And some of them appear to be very powerful. Right, that pure explanatory power just feels, yeah. It looks that way. But then what seems at some point as explanatory power and breadth of explanation,
00:53:46
Speaker
instead of becoming accommodationism, right? People try to feed it and feed it and feed it. And then you end up with epicycles like Copernicus, right? Every trend to accommodate this theory by creating tiny little epicycles because the beauty of the theory has to be somehow preserved. And I think that we are, as neuroscientists, are no different in that regard than to other scientists.
00:54:10
Speaker
Yeah, well, I mean, I guess one question would be, what would be a good conscious, what would be a good theory of consciousness as it relates to brain? And then, you know, we could even get into like, whether we want to have that conversation, or whether we want to, like, talk about whether that's even the right conversation, but
00:54:28
Speaker
Maybe I'll just ask it as a question. Do you think there is a good... Yeah, I mean, that's the right question. I agree with you. That is an excellent question and it's a very difficult one. And it is a question that I have thought, I mean, there is a sense in which I perhaps left the field of consciousness research in part because I just didn't know what a good theory of consciousness could be. I think I had an idea and today I am completely convinced that there are good
00:54:53
Speaker
experiments that tell us stuff about the how the brain processes conscious information, right? I am and I am
00:55:03
Speaker
very much in awe with a lot of work that has been coming from some contemporary and typically younger cognitive neuroscientists, like Megan Peters' work, Ordegaard and Jorge Morales and Northwestern, Steve Fleming, Hackett and Lau. I really admire their work. Their work is more in the psychophysics, but with the component of consciousness.
00:55:29
Speaker
This clumsy way of going about the science that I think is more promising than finding like overarching theories with extraordinarily limited data.
00:55:42
Speaker
And how would the right theory of consciousness go? I don't know, because I have... So here's the contrast category. The contrast category is with what I think are good theories of, say, memory or, say, you know, perception or audition, how they pass.
00:56:00
Speaker
Part of what happens is that we have behavior there, right? I mean, do you have no idea how much I love my recognition rates and hits and false alarms? Because I have that behavior to prove. I work sometimes on probabilistic models of recognition memory, and I can show in my model that if you modify the prior, which I have full control over,
00:56:26
Speaker
then you're going to have more or less false alarms in a recognition test. I love that stuff, right? Because you feel, and then you can also have maybe parts of the brain that are sensitive to the priors and that are going to explain some of the variance in my behavior later on. So if we could have something like that for consciousness, then I think we will be golden.
00:56:49
Speaker
But it is hard because we don't necessarily have behavior because everything gets modeled by introspective reports and so forth. So that's what gets it tricky. If we sort out how the terms of the theory relate to specific measures and there is sort of agreement and clarity on how they go, then we might start talking about how the theory would look like.
00:57:19
Speaker
But I don't think that we have clarity on that. If you were to make some wild speculations, 50 years, do you think people will be more happy about experimentation and empirical methods to figure out consciousness?
History and Sociology in Consciousness Research
00:57:39
Speaker
Or do you think it might be more of a shift in conceptual space that might change thinking on consciousness?
00:57:51
Speaker
Hmm. Well, uh, that's a good question. So I, on the one hand, I think that the science of consciousness is as old as any other science of the mind, right? It just comes in different, in different ways. Here's an example, something that I absolutely adore because it's a, it's just very clear. So do you have heard about this notion of aphantasia, right?
00:58:18
Speaker
the idea that there are people that have difficulty, if not complete impossibility and capacity to imagine and visualize certain things in their eyes and they call it a temptation. And I remember seeing about, reading about this in some recent articles saying like a new phenomenon discover about the mind. Not new at all.
00:58:44
Speaker
Galton had talked about a phenomena that was similar. In fact, some of the first proofs that the Gaussian distribution and sensory perception were, I mean, the distribution of the sensations were kind of Gaussian and formal Gaussian distribution, but because there were people that were at the tail of the distribution in their capacity to visually visit. And we're talking about 1886 or something like that. Psychology was a baby.
00:59:12
Speaker
like Donald Hebb, for instance, had work on visualization, much of which spoke to aphantasia, right? So there are certain phenomena in the science of the mind that have revitalized under a different name and then they
00:59:29
Speaker
You know, it's placated somehow and then it resurrects and consciousness is the same. Consciousness has been going on like that since like forever. It just comes with different... Yeah, we think of it as sort of being revived in the 90s or so, maybe because people were just more willing to talk about it or sort of directly address it.
00:59:50
Speaker
Maybe it's a possibility. I always think that those things are, you know, there's a lot of sociology that goes accompanied with that, right? I mean, my own experiment, I do philosophy of memory, which was basically a non-existent field in the early 2000s, because everyone was talking about perception, right? And now a lot of people talk about
01:00:14
Speaker
about memory and probably was good to decay later on again uh... there is a lot of sociological reasons as to why that happens do all you need is
01:00:23
Speaker
two philosophy professors in an admissions committee or in a hiring committee that are willing to take the risk of getting someone else, give them a desk to write and then they end up publishing these papers and then they move a different, or all you need is some person who didn't think that X was a topic that was worth mentioning, you just need that person retiring or stop being the editor of a journal for that field to flourish.
01:00:49
Speaker
we should not diminish the role that sociological aspects play in the evolution of the sciences. That's an interesting, that's a really interesting point. And I do think, yeah, the history of scientific research on consciousness has been going on for a long time. It's maybe not recognized by that name as much, but certainly aspects of consciousness for a lot longer than we think of as sort of the current revival.
01:01:17
Speaker
Exactly, exactly. And we have, despite that we study memory, we have really bad memories for reading all research. And I just, I, when I teach my, I teach a class in philosophy and neuroscience and I love reading older stuff and showing students to read older stuff. I mean, one of my favorite papers from 1966 or so by Posner talking about prototypicality effects on recognition memory.
01:01:48
Speaker
This guy, like Posner, this invention researcher, shows different dots that approach a prototype, but he doesn't show the prototype. It turns out that your errors are approaching the prototype. So it's the first sort of systematic showing of prototypical effects in reclamation memory. Then we have in the 1970s,
01:02:09
Speaker
For further recognition and that even in former artificial intelligence. Then there was beautiful work by Putin butcher and colleagues, showing how they they form based like a basic model are able to predict those kinds of errors and people rediscover them and then of course
01:02:27
Speaker
eight years ago or so. Oh, finally, a basic explanation of memory errors. Like, we've known this since the 1950s. A typical answer in vision is Helmholtz must have done it in the 1860s, right? Right. Exactly. Exactly.
01:02:44
Speaker
So there is a lot of, we forget a lot, I mean there is a sense in which is kind of historical replication and it is nice to see it that it gets pops out again and again and again and again without necessarily the recognition of who did it earlier but that is a similar effect that's going to happen.
01:03:02
Speaker
To get back and nail you down for something that we can hold you accountable for in 50 years, do you think it's going to be a conceptual shift that might inform people?
01:03:17
Speaker
the wave of experimentation and sort of this general kind of approach might bear fruit. Maybe not IIT but you know. Or is it tour number three that we're on this epic cycles and ourselves and we're just going to be right back where we started in 50 years. You know I've
01:03:37
Speaker
I don't know. I think I go back and forth in this because, of course, I'm a cognitive scientist and neuroscientist, and I love what I do as a scientist, but I'm also a philosopher whose family is skeptical of the sciences, kind of like a schizophrenic life of living this life. I think that there is going to be certain phenomena
01:04:00
Speaker
that relate to behavior that is going to be replicable and pretty solid, particularly in areas like psychophysics and particularly in certain memory phenomena. I mean, we have results like that.
01:04:17
Speaker
One thing that I detest is when people say that the school of introspectionism was all debunked. You know, you remember this from the history of psychology classes that tells you introspectionism, when Tickler and all those guys were doing experiments in introspection, then it turns out that different labs, and back then there were very few labs, so things didn't replicate. I mean, that was the first replicate that happened, right? Things didn't replicate and different measures and stuff.
01:04:45
Speaker
Yeah, but there are some stuff that we got out of there. So, for instance, pain scales were developed back then, and they are an extraordinarily reliable method for tree-aging people in emergency rooms. Right? Typically, things that are going to kill you hurt more than things that are not going to kill you. Right? So, sure.
01:05:05
Speaker
There are lots of, some of them that are going to go underrated, but overall as a very rough estimate of whether I should see you right now or if you can wait 20 minutes, they are relatively decent. And where were they based on? Where they were based on intersection, right? They were based on these things that we tend to disparage. So I have faith in results.
Memory and Forgiveness in Colombia
01:05:27
Speaker
I have faith in specific findings. How exactly they are going to fit
01:05:33
Speaker
in a theory of our cognitive ontology, I don't know in part because it is, I mean, we have been chasing this since forever, right? That first, our cognitive ontology is not that really different from Aristotle's and everyone, and we think that we have made some progress. But where we have definitely made progress is in results, in results that speak to specific questions about the mind,
01:06:01
Speaker
even if they don't give us an answer to the overarching question of how is the structure that that is so i don't know that that's kind of like a wishy-washy answer but you can but i i go with
01:06:15
Speaker
I need to proportion my degree of confidence to the evidence. Sounds like a vote for empiricism, though. Yeah. That sounds good. I like that. So maybe that's a good place to wrap it up. We always like to ask a question at the end, which is, what are you excited about in terms of your own work or the work of others that you're following that's coming up in the next period of time? Just what are you excited about in terms of your research? In terms of my research,
01:06:44
Speaker
Well, I have a line of research now exploring how memory and forgiveness interact. So I have a large research project currently exploring
01:06:57
Speaker
how the peace process in Colombia, my home country, has affected the memories of the victims of political violence. So I do a bunch of research on the autobiographical memories of victims as a function, and then I explore how the degree to which they have forgiven their perpetrators have modified their effective reactions toward the autobiographical memories of those wrongdoings.
01:07:24
Speaker
And also how it is possible that we can, how could it be possible that we can employ emotional reappraisal strategies for autobiographical memories to help people live together again in community and reconcile. Because, you know, because Colombia is now undergoing this big process in which victims and perpetrators of this violent acts are back to living together. So that has been a big thing that I'm working on these days.
01:07:53
Speaker
Sounds exciting. In terms of research that others have been developing, I love the research and medical mission, to be honest.
01:08:06
Speaker
So I didn't know much about it. And my student Kevin O'Neill actually started working on the metacognition of causal reasoning and causal judgment, which is something that I never really would have thought of unless, until he started working on this topic in my lab. And that got me really interested in metacognition. And so the work of Steve Fleming, the work of Megan Peters,
01:08:35
Speaker
the work of Jorge Morales in Northwestern, that work excited me a lot. I really like thinking about that.
01:08:45
Speaker
Well, great. Yeah. Well, Felipe, thank you so much for being on the show. That was great. And we'd love to have you back on to talk about your own research, especially the research that you're talking about, autobiographical memory and forgiveness sounds fascinating. We'd love to explore that more with you. So yeah, thank you so much for being on the show and really appreciate it. Yeah, of course. I'll be happy to come anytime.