Cognitive psychologist Axel Cleeremans explains his radical plasticity thesis: consciousness is something the brain learns, as higher-order systems come to redescribe what first-order systems are doing. He and Richard discuss implicit learning and connectionism, whether consciousness is graded or all-or-none, how his view relates to higher-order and global workspace theories, the social side of consciousness, and why he doubts a system like AlphaGo has any experience. The video of this one was lost, so it is audio only.
Listen: Apple Podcasts · Spotify · Original post
Chapters
- 0:00 Introductions: Axel Cleeremans and implicit learning
- 1:59 Connectionism and knowledge without access
- 6:26 Flow, automaticity and skill
- 8:35 Where consciousness fits: quality of representation
- 11:06 The brain learning about itself: radical plasticity
- 14:02 Is consciousness graded? Animals and infants
- 20:50 Redescription without concepts
- 39:07 Misrepresentation, colour constancy and the dress
- 46:11 Uniting higher-order and global workspace theories
- 51:26 The social side of consciousness
- 57:11 All-or-none or graded experience?
- 1:07:06 Partial awareness
- 1:08:32 AlphaGo, wanting and machine consciousness
- 1:15:07 The ASSC community
Where the big topics come up
- 11:06 Higher-Order Theories: “…perspective on conscious cognition basically leads naturally to endorse or embrace higher order perspectives on consciousness itself, right? Because whatever differences there may be involves these monitoring…”
- 26:34 Metacognition & Confidence: “…thing I could find that we can actually look at from an empirical point of view is confidence. So one of the things the high order system can do is compute confidence. You”
- 1:03 Unconscious Perception & Blindsight: “…question I was most interested in were all the differences between conscious and unconscious”
- 1:15:26 The Science of Consciousness: “…meeting ever. Because there's no other meeting I know of where philosophers and neuroscientists and psychologists can really share ideas in the same way. And that makes it a wonderful community.”
- 46:42 Global Workspace Theory: “theory either comes out to be like a higher order theory or a global workspace theory. Because what's doing the work of consciousness is the higher order part. And then the global workspace…”
- 1:09:58 AI & Machine Consciousness: “…something to you? That strikes me as a fundamental difference between systems like AlphaGo, for instance, even where they're endowed with”
- 17:18 Animal & Plant Minds: “about animal consciousness and I begin, you know, with, I don't know, monkeys. Who thinks a monkey is conscious? Everybody raises…”
Cite this episode
To cite a particular moment, add the timestamp (for example, “at 1:02:15”). Links to this page with ?t= and a number of seconds open the audio at that point.
APA
Brown, R. (Host). (2019). Consciousness Live! with Axel Cleeremans (Season 2, Episode 6) [Audio podcast episode]. In Consciousness Live! https://onemorebrown.com/consciousness-live/s2e6-axel-cleeremans/
Chicago
Brown, Richard, host. “Consciousness Live! with Axel Cleeremans (Season 2, Episode 6).” Consciousness Live!, 2019. Audio podcast, 1:16:34. https://onemorebrown.com/consciousness-live/s2e6-axel-cleeremans/.
Episode notes & links
Transcript
Transcript with speaker labels, made by machine from the episode audio. Expect a few slips, especially where people talk over each other or trail off. Click a timestamp to jump there in the audio.
Show / hide transcript
0:00 Richard Brown: Starting, oh, all right. So we are semi-live. Not exactly live today, but close to it using Skype call. So joining me today is a very special guest, Axel Cleeremans. So Axel, for those who don't know you, can you say a little bit about who you are and what you do?
0:19 Axel Cleeremans: Yeah, I guess I'm a professor of cognitive psychology, working at the Université libre de Bruxelles in Brussels, Belgium. I'm a research director with the National Fund for Scientific Research in Belgium, which means I teach but just a little bit on what I choose to, which is cool. I did my PhD at CMU in the early 90s with Jay McClelland on connectionist networks. So that's about it. And most of the stuff I do is around Consciousness, I worked a long while on implicit learning, the phenomenon by which you become sensitive to regularities in your environment without awareness of learning anything. And then consciousness happens with ASSC1 and other developments, Tucson. And yeah, from then on, the question I was most interested in were all the differences between conscious and unconscious information processing, of which implicit learning as an example. But then my whole field of view got a lot bigger after I got in touch with people like Bernie Baars and Dave Chalmers and other people who were already thinking about what consciousness itself is.
1:49 Richard Brown: So when you were first working on implicit learning, you weren't really thinking about the distinction between conscious and unconscious.
1:59 Axel Cleeremans: Yeah, you do have to think about it. And at the time, the epiphany for me was connectionist systems, because they were unlike traditional sort of box and arrow, Simon and Newell sort of traditional models of information processing. you know, starting out in the 60s, moving forward and so on and so forth, they depicted a view of cognition in which all of cognition was basically problem-solving. So from computing how you move your arm to solving a Tower of Hanoi problem, the vision was that this was all driven by computing possibilities in a problem space and working with basically hypothesis testing. That seemed so implausible to me and it was just impossible to reconcile this with the phenomena of implicit learning and unconscious cognition in general because it seemed to presume that it always entailed access to some stored information, right? So information processing always involves access according to those perspectives. We could talk about issues or solutions to that, like knowledge compilation and the idea that entire chains of explicit steps in a problem-solving episode can sort of get compiled through automaticity, for instance, and that's only an important issue. But the core of those sorts of models were always beginning with the idea that it's all driven by a processor accessing information that sits in a database, and that makes it possible for you to move forward with whatever you're trying to compute.
3:50 Axel Cleeremans: And connectionist systems offered a radically different perspective on this, which was not only, of course, obviously, that was the intent more connected to what we know about how the brain works, but also addressed this quandary that classical systems had at the time. This was all 30 years ago. accounting for the phenomena of unconscious cognition in a very simple natural way. Now we have these computing systems, first order or feed-forward or even recurrent connectionist networks that just carry out their computations and all the relevant knowledge is literally embedded in the connection weights or the synapses of that system and there's nothing that needs to access anything for these systems to compute effectively. And so I ended up working with Jay McClelland on such systems as models of implicit learning, because they offered this, you know, Very simple, straightforward solution. There's just no consciousness in these systems at all.
5:04 Richard Brown: Right.
5:06 Axel Cleeremans: Because there's no access, in a way. Because
5:08 Richard Brown: there's no access.
5:08 Axel Cleeremans: Yeah, yeah. There's nothing that accesses anything. The knowledge is just embedded in the very structures that drive the processing itself. Right. Yeah, exactly. And so that's what we're seeing, or that's the claim, at least in phenomena like implicit learning. you get all these changes accruing as a result of your practicing some task, from riding a bike to learning about the structure of natural language. And all of this seems to go without hypothesis testing, without access to whatever knowledge you're acquiring, or your performance is just improving as a result of repeated exposure to the problem space or the domain.
5:56 Richard Brown: Right. So that's a good evidence that it's an unconscious process, that there's no access. And when you're doing it, it's automatic.
6:04 Axel Cleeremans: Exactly. It doesn't matter. And in fact, there's even evidence, of course, that accessing the knowledge while you're learning the skill, and this is particularly true for athletics and stuff like that, interferes with performance because now you're involving these extra systems that actually look at whatever it is that you're learning.
6:26 Richard Brown: This is like from the psychology of flow and stuff like that, work on flow. Oh, flow
6:31 Axel Cleeremans: is a good example, but I think flow sits at the over extreme. Flow is opposed to automaticity. Flow, I think, is what happens when a skill you have learned has become so automatic and you need so little monitoring of that for the behavior to unfold effectively that you can actually sort of watch yourself doing it. And that's the experience of flow. You have this impression of detachment from your own performance. You become an observer of the actor that you are as well. Right. That's flow. That's flow, I think. Right. So floating above yourself with a first person camera, looking at yourself, doing it. And it feels so fluent and so. And I know you have this experience that you're you know, you were just watching yourself do it. And it's perfect. Right.
7:25 Richard Brown: Yeah, and so some people suggest that when you try to have that awareness of yourself performing the task, it actually interferes with you performing the task. I thought that's what you were kind of saying a second ago. Yeah,
7:38 Axel Cleeremans: I guess that's true. But only as you're learning the skill and the skill hasn't become intimated enough that you can actually do this monitoring without interfering, just as a pure observer. Right.
7:52 Richard Brown: And then sometimes, I mean, you know, just my limited experience as a musician, which, you know, not a great musician, but I've had some experience. There is a sense that when you're aware of what's happening as it's going on, you have a greater sense of control that you can kind of direct the automatic processes in a more focused and narrow way. So that's, and I think that's probably part of what you want to argue is the role that consciousness plays, right? It's sort of in those kind of control moments. when it's in between automatic and something that we're having to spend a lot of time trying to figure
8:29 Axel Cleeremans: out how to
8:30 Richard Brown: do. In between
8:31 Axel Cleeremans: implicit and automatic, yeah.
8:35 Richard Brown: Exactly, between implicit and automatic. So that's where you think consciousness has a special role to play. Is it fair to say that?
8:41 Axel Cleeremans: I think so, yeah. Yeah. Intermediate level representation, so to speak. Right. Whatever knowledge you have acquired is of sufficient high quality that there's actually something to detect there for your awareness. You can be sensitive. You're noticing that you have acquired these bits of skills, etc. But they're not yet strong enough that you can fully trust them to do their job on their own, so to speak, as would be the case in automaticity.
9:17 Richard Brown: Right. So, and this is something, sorry, go
9:20 Axel Cleeremans: ahead. Over learning a skill, I think we begin with weak sort of ill-defined representations that drive behavior. I'm thinking here of cases like maybe subliminal perception and things like that, things like that, right? And then we move along a skill continuum, which is also time, all the way to very strong representations characteristic of automatic behavior. Those are strong enough that you can trust them to do their job without conscious monitoring being required. That's the result of learning the skill to its fullest extent. Flow happens there. But I think automaticity is a state where Consciousness is not absent, like is the case for unconscious or implicit cognition. It's just that consciousness is optional. If you like, you can pay attention to whatever it is you're doing when you're driving or going down the ski slope or whatever, but it's not required anymore for control of the skill to be successful. Right.
10:31 Richard Brown: Yeah. I think that's an interesting idea. And really, I mean, so your your basic idea was when you have this first-order system, which is learning about the environment, then where the level of that consciousness is going to come in is where the system starts to learn about the properties of its own functioning, so to speak, the causal connections between the states.
10:51 Axel Cleeremans: It's exactly what it is. It
10:53 Richard Brown: comes to redescribe the things that's happening at that level. But your interesting twist on that is that it does this in an implicit learning sort of way. So it's just implicit learning at the next higher level up.
11:06 Axel Cleeremans: Yeah, exactly. So this perspective on conscious cognition basically leads naturally to endorse or embrace higher order perspectives on consciousness itself, right? Because whatever differences there may be involves these monitoring systems looking at whatever the first order system is doing as it is learning to carry out a particular task. And so that's why I find the high order thought theories or high order in general theories of consciousness most interesting. It's because they capture this intuition that consciousness involves these sort of observing systems that monitor the dynamics and the representations that develop in first order systems as they learn. And of course, you cannot do this unless there's something to look at that is sufficiently well-defined in the representational space, basically, right? That's why there's nothing to see for unconscious representations typical of, you know, subliminal perception and early, very early stages of learning a particular skill. There's just nothing to see. There's nothing for an inner awareness to grab onto at that stage. Then those first order representations get stronger as the skill develops. Now there is something for a high order monitoring system to become sensitive to in such ways that it can increase cognitive control, but also that would be the far out claim produce conscious experience.
12:52 Axel Cleeremans: That would be this inner awareness kicking in. and giving you an instance when there was none before. And then this continues all the way until the first order system is completely stable again because it has learned its skill. And then there's, again, in a way, nothing more to see for the observing systems, except it can. It can, right? So consciousness is optional in those cases. You can engage those monitoring systems if you like, but there's nothing that they do from a purely functional point of view anymore.
13:33 Richard Brown: Right, because they've already learned. They have already extracted the information.
13:37 Axel Cleeremans: Of course, yeah. You can apply exactly the same sorts of arguments about quality of representation, as I call it, that you apply to these first-order systems as they learn something to the high-order monitoring systems. That is, better representations about other representations can themselves become automatic in exactly the same way using the same principles.
14:02 Richard Brown: Right. I think it's a really interesting view and there's a lot I want to ask you about it. But I think the obvious thing that people probably ask you a lot is, If your view is that consciousness is something that the brain is having to learn how to do, then the obvious implication is that it's not, it doesn't come sort of automatically equipped with consciousness. So that when the brain first sort of starts to function, there's no consciousness. So that would entail like, you know, infant or at least neonates or something, you know, there'd be some level at which there's a brain, but it's doing something, but not these higher level functioning And is that something that you accept as a consequence of your view, or do you think there's a way out of
14:45 Axel Cleeremans: it? No, it's a hard bullet to bite. But, you know, first of all, it doesn't have to be all or none. On the contrary, I think it's very much graded. And so I think a lot, in the case of human infants, for instance, I think fetuses learn a lot already in the fetus. There's even evidence from social interactions amongst twins. in the fifties. So I don't think, you know, the birth is a good starting point here. I think this all begins way before. I think there's nothing that prevents a sufficiently mature brain to begin learning about its own functioning. That's what I call the inner loop in a way, the brain re-describing its own activity to itself. And there's evidence for that, even with metabolism, for instance. We know that the brain starts sending oxygenated blood, for instance, to regions that it anticipates will be necessary to carry out some particular task later on. So that's, you know, I mean, we are a subject in no way aware of these sorts of mechanisms. And so it's suggestive that these sorts of monitoring mechanisms starting from the brain, staying within the brain, observing itself or observing its own metabolic needs, for instance, as in place already. And, you know, there's no reason to believe that this begins with birth or anything like that.
16:20 Axel Cleeremans: But the argument that consciousness is something that the brain learns leans towards a sort of fairly restrictive vision on who is conscious. But the truth is we don't know, right? There's nothing that would prevent such mechanisms from operating, in my view, in human fetuses. There's nothing that would prevent such mechanisms from operating in even elementary or simple organisms. Often I try to probe my students with this in the only class I get to teach. It's kind of funny because people have very different intuitions about this. Of course, you know, it ranges from banned psychism to people who think language is so essential to consciousness that no linguistic organisms are capable of consciousness. But amongst my students, I remember, you know, typically I probe their intuitions about animal consciousness and I begin, you know, with, I don't know, monkeys. Who thinks a monkey is conscious? Everybody raises his hands. Dogs, you lose a few people, you know, and then you go down to, I don't know, pigeons and rats and mice. And and here I did this exercise last fall, I think. And, you know, I ask who thinks a dog is conscious? Nobody. There's just one guy out of 40. And I was really surprised.
17:50 Axel Cleeremans: But nobody has a good answer to that question. Right. But the line, I don't know. My intuition is go fairly low. But yours may be very different. And I don't feel there's any sort of principled way of addressing this issue at this point, right? The speculations of the structure of the bee's brain and things like that and sort of rough analogies with human cortex and so on and so forth. But I don't know.
18:21 Richard Brown: Yeah. No, I think this is, you're exactly right. I mean, I think this is a methodological, my, my own intuitions are that, yeah, animals are conscious obviously. Um, but the question is, is a methodological one about whether you can, what it would count as evidence to catch, convince the skeptic. And that's what I think the challenge is. I think, I think one of the things that we kind of agree on, I'm not sure, so maybe you could tell me, but I think one of the things we agree on is that we assume that the brain starts off kind of, uh, non-consciously processing information, that the default assumption should be that information processing sort of started off without consciousness and then consciousness somehow gets built out of information processing structures and processes. As opposed to someone who might assume that, oh, but the biological structures themselves kind of bring consciousness
19:08 Axel Cleeremans: with them. Exactly. That's definitely my intuition. I don't see anything inherent or intrinsic to the activity of neurons. that should make that activity conscious. What is the principle? I cannot come up with a computational principle that would make it mandatory for neural activity to be associated with consciousness in any way. So I think some structures make it so that brains can be conscious for some organisms, But then I would appeal to these interactions between first order systems and high order systems. And so the learning question, you know, the slogan, the brain learns to, consciousness is something that the brain learns to do. Stems from a question, actually. Well, then, okay, how do we get there? Okay, we have this interesting mechanism that higher-order theories propose, which is interactions between first-order and higher-order systems. That's fine. That certainly captures my intuitions about it. But how do we get there? Are we born with this? Is this a sort of natural consequence of the maturation of sufficiently sophisticated brains, or is it something else? Is it a process that is more constructivist in nature? And the perspective I'm trying to develop is that latter idea, the idea that it is constructivist in nature, that the brain learns to re-describe its own activity as we develop.
20:50 Richard Brown: And then, so the re-description process is again, on your view, cast out in terms of an implicit learning mechanism so that the re-description is not a conceptual re-description, which is a little bit hard. I mean, the re-description sounds like it's got to be conceptual because you're describing it, but you're saying that it's actually, is it fair to say that the reason it's a description is because the The structures are isomorphic in some sense or the synaptic weight structures or the relationships between the inputs and outputs are similar. So what is it that makes it count as a re-description of the first
21:31 Axel Cleeremans: order state? Well, the re-description comes from this idea of the great developmental psychologist Annette Karmiloff-Smith and her theory of cognitive development. which was cast as a constructivist alternative to Piaget, a more sort of graded alternative to Piaget's theories. And a core concept in that theory, in her theory of development, is this concept of representational re-description. So the word re-description, I borrow from Karmiloff-Smith, who didn't have a vision of it that was driven by language. So she envisioned it more as a non-conceptual sort of process by which knowledge structures, the core idea there is that a knowledge structure can be embedded in a system, but it won't be an object of representation for the system unless it is redescribed in some way by an observing system or by a higher-order system, right? So that's the core idea. And her thinking about this was very much in line with Connectionist perspectives on it. With one network observing another network, basically. She never got to instantiate this or implement. We tried and we have some good examples of how you can build these simple systems. Back from 2010, I think we had a couple of papers, but it turns out to be a very complicated problem to try to operationalize this.
23:12 Axel Cleeremans: Today, perhaps generative adversarial networks are the best like-minded idea. But neither in her view, and certainly not in mine, should these re-descriptions be conceptual in any way. They do not involve language, they do not involve concepts. It can be very non-conceptual.
23:35 Richard Brown: Yeah, but that's what I'm kind of trying to figure out. It's like, what does that mean? So what does it mean for there to be a non-conceptual description of something? So is it just that, because I guess in the generative model idea, the idea is that the generative, so the discriminator tries to produce the same kind of activity that the first order systems are gonna be sending to it. So it tries to get or model that activity. So is that the idea that the higher order system has to somehow be in the right synaptic weight or is it the activity of the system that encodes the information? So I'm just trying to get clear about which part of it counts as the representation and how that representation kind of connects to the content of the first order state.
24:23 Axel Cleeremans: Yeah, well, so the difficulty there is to figure out what the high order monitoring system is actually doing, what it's trying to do. The broad version of this is that it's trying to characterize in some way the mental states or the states that the first order system can find itself in, right? It wants to learn something about the dynamics and the structure of those representations that exist in a first order system, right? So it wants to it needs to become acquainted with the structure, the geography of those states, basically. How many distinct states can the first order system, the observed system, confine itself in? And what do these states mean for the system as a whole? Now, the subject gets in there, of course, which is, I think, how meaning is achieved. But so my intuition here is that the higher order system's job is to characterize the first order representations in ways that are balanced as well. You know, is this something I want? Is this something I regret? Is this something, I mean, mental attitudes, basically. Now, to build a computational model that does that, it's very difficult to figure out what to train the high-order network on. What is the task
25:50 Richard Brown: of high-order?
25:53 Axel Cleeremans: So we tried different possibilities there. You can get unsupervised systems that just render a more abstract version of whatever the first-order system is doing. So in a sense, that's a re-description. But that work was disappointing for me because, okay, we get this sort of skeletonized, in a way, re-representation of what the first order network is doing, but then what is that high order of representation doing? It's not doing much. It's not doing much in the sense that Maybe it's right, but I cannot think of experiments that would sort of make it possible to sort of look at this empirically, right? So the only thing I could find that we can actually look at from an empirical point of view is confidence. So one of the things the high order system can do is compute confidence. You know, do I think the first order system network, the first order system or the first order network is correct in the decision it just took? And we ask that higher order network to express confidence then in the first order network's performance. And that works quite well in accounting for the relationships between confidence and first order performance in a number of tasks. And so the early computational work I referred to before, which was the work of my PhD student Antoine Pasquali, was exactly about that, was about looking at the extent to which we can simulate empirical data about confidence in a number of different experiments using this sort of computational architecture.
27:49 Richard Brown: Just really quickly, if you have a system that's outputting these confidence judgments, then isn't the, in a sense, the ultimate representation conceptual? It's like the, you know, 45% correct or chance of being reliable or something like that. So that seems like a…
28:11 Axel Cleeremans: I guess I would have to agree that it's conceptual. I'm not exactly sure what you as a philosopher mean as conceptual here. For me, it doesn't have to be connected with language. That's what I mean. It doesn't have to be connected. But obviously, if you are a linguistic animal, then it's going to help you tremendously to do these sorts of re-descriptions, which then, by virtue of them being linguistic, become conceptual. I
28:39 Richard Brown: see. I wasn't meaning concept and linguistic to be the same. I wasn't trying to… No, no. I was thinking of…
28:50 Axel Cleeremans: to make it more practical, embodied, and active in a way, right? Yeah.
28:56 Richard Brown: Right. So I guess, yeah, the word concept carries a lot of baggage with it, but if you just think of a concept as something that plays a role in the mental life of the animal, to help it to have thoughts with what I call intentional content, content that represents that things are a certain way, So that's what I would mean by conceptual representation, whereas a non-conceptual representation would be something that represents in a different kind of way, maybe like a pictorial way. Sometimes people put this by saying non-conceptual contents have every part of them represents a part of the subject, every part of a picture represents something, but not every part of a linguistic or a conceptual representation does. So if you just need something that simple, something that doesn't inflate it too much, then Even, I mean, so in connection, it's, I mean, I guess this is just more of a deeper philosophical philosophy of science points. So if you don't care about it, we can move on, but it just seems like it's connectionist models in general that there's a lot of emphasis placed on the fact that there's no concepts, you know, because it just, all the information is stored in the synaptic weight, but there's always an output and the output is always like, it's a dog or it's not a dog or it's a, it's a face or it's not a face.
30:15 Richard Brown: So, at that level there's a conceptual representation and if this thing's supposed to model what's going on in the brain then that state which is the output also has to be somehow represented in the brain not maybe in linguistic form exactly like that but something where the concepts of dogs and things like that are figuring in um yeah seeming i mean or do you think that that's the wrong way of thinking about how
30:41 Axel Cleeremans: No, what I think is exactly, yeah, it is connected with that because indeed in first order networks or first order systems, but in particular it's particularly clear in the case of connection systems or deep learning networks or whatever, they're all the same anyway, all the relevant knowledge is stored in the connections, right? All the relevant knowledge is stored in the connections. But then you have these patterns of activity, which are potentially data structures. And I wouldn't call those conceptual in any way, but they do exist. And so the idea is that those then become input to a higher order network, which runs in exactly the same way as the first order network. they become objects of representations. Just like the output of the system, but those intermediate states of computation along the chain, say the visual system or whatever, they themselves can then and only then become objects of representation for another observing system.
31:47 Richard Brown: Yeah, right. Which
31:48 Axel Cleeremans: does something with those patterns of activation, right? But otherwise, those patterns of activation are just part of the causal chain that leads to a particular output, to a particular decision, to a choice, and so on and so forth. Nothing has access to those unless there's a serving system. That seems to me both the core idea of high-order theories in general, and it's also the core idea of the theory developed by Karmiloff-Smith, but also Andy Clark and Karmiloff-Smith, they had this great paper called The Cognizer's Innards in the early 90s, I think, in which they spelled out the sort of consequences of Karmiloff-Smith's perspective on cognitive development in the wider sort of landscape of cognitive science. And so they argue the difference between you know, genuinely intelligent or intentional systems and simpler systems is precisely that joint where mental representations or patterns of activation can then become themselves objects of further re-description for the subject itself, for the organism itself, right?
33:13 Richard Brown: Right. I want to move on here in a second, but let me just try one more attempt at this. The higher-order system is an implicit learning system as well, so its knowledge is also stored in the synaptic weight of the higher-order system. But there's also patterns of activity that are going to be data-structured at that higher-order level as well. Which ones count as the representations of the lower-order state, the patterns or the activation weights in the synapses?
33:47 Axel Cleeremans: Yeah. Yeah, that's a good question. If you buy higher-order theories, I guess the patterns of activation in the higher-order system are what makes those contents available to awareness, right? It's there that it happens, basically, the interaction between those two systems. You minimally need two systems. I think that in truth you have hierarchies of such systems, right? Deep hierarchies of such re-describing systems. And so then I think of… after those strange loops, in a way, where he says, OK, you keep looping up the levels of hierarchy, and then you end up at the same place that you began. And obviously, you loop through the subject, which is perhaps the topmost high-order re-description of whatever the system we are actually is. So that's one point. And then there was a very difficult point that I could never figure my way out completely, which is that in a way, in a real way, in a strong way, you can argue that all the layers of a deep learning connectionist system, for instance, are re-descriptions of the earlier layers, right? They do count as re-descriptions. all the work that's been done on vision, for instance, you see these layers and layers in a 30 layers deep network.
35:29 Axel Cleeremans: With a lot of work, you can do imaging on those layers and figure out what the functions, what is it that they do represents. In a true way, you can say that all these layers are just re-descriptions of whatever the prior layers are doing. But this is not what I mean. Those don't count as meta-representational in any sense, right? Because they're embedded in the same causal chain that's involved in computing a first order decision. So you need the observing systems, the high order systems to be somehow at least partially causally independent from whatever the first order system is doing for that guy to play its meta-representational role.
36:26 Richard Brown: But it can't just be driven by the sensory, by the stimulus, the way that the first-order states. Well,
36:31 Axel Cleeremans: yeah. It also has to sort of sit outside the chain, right? Right. But then, since everything is recurrent, you know, everything is connected to everything. Exactly. And then, what do you reason about this? But this, I feel, is a question for physicists and people who have thought about causality more than we have in psychology because, Those are, yeah, I feel, very difficult issues, very difficult issues.
37:02 Richard Brown: Yeah, that's, yes, I mean, that is related to the question that we were talking about a second ago about what makes it count as a re-description, because if you have a liberal enough definition, then every stage is a re-description. And then you have the question of, well, which stage is the one that matters, or why aren't these matter representations in the way the other ones are? And it sounds like your answer is that the first orders, the cause and effect relationships that exist between the stimulus and the first order states, these re-descriptions are a part of that process. So they should be counted as first order and not matter representational. Is that the basic idea? Yeah, because I guess generally the distinction is made in terms of the kind of content. So if the content is, The content of the state is more directed towards the external object than it's a first order content. If it's directed at something internal, then it's a higher order content. So a lot of people might want to count these things as higher order, but you're saying that you wouldn't want to do that.
38:06 Axel Cleeremans: No, not really. No, no. I'm just working based on intuitions here.
38:15 Richard Brown: Well, we all are.
38:18 Axel Cleeremans: Because there are better causal chain, right? That leads to yeah, from a stimulus to response or something like that, right? And you'd want the observing system to have at least some of the features of the observing of the observers of those first order systems, right? A bit like Like what we do when we look at a connectionist network and try to figure out what it has learned, right? I'm obviously completely independent from whatever that network is doing. And so I have a very sort of a remote vantage point on whatever it's doing. And maybe you don't need to go all the way like that with these higher order observing systems, but they do need to be partially independent.
39:07 Richard Brown: I mean, I think I think it's a good point. I think traditionally, you know, this is something even like going back to the work of David Rosenthal, who's a, you know, higher thought, there's a well-known defender and so forth. His early work, he used to say that the first order states caused the higher order thoughts which represent them. And then he backed off that. And I think part of it was for these kind of reasons. So what he ends up saying is that there are lots of causes in the brain. Some of them may be first order states, but there's also a lot of higher order causes as well. You know, other psychological states that may be bumping around that aren't necessarily driven by the stimulus or these first order states. So that would be enough independence, I think, right? Everything is connected to everything. So there's not going to be full independence, but you need some kind of influence that isn't just from those first order states, which, which obviously, I mean, so I was getting a little bit distracted here, but since this is coming up, This leads us to what sometimes people think is a really dramatic objection to the high order theory in general, which is this mismatch problem, that there could be a somehow mismatch between the states at the first order level and the high order level.
40:19 Richard Brown: And if you allow some kind of causal leeway there, then it seems like you're going to end up with some version of this problem, that there could be a state which in some sense represents The first order state as being other than it is if there's a personal representation of seeing red, it could be represent using green or. Or something like that, so do you think that your view allows for this kind of misrepresentation or do you think that it somehow some part of your view
40:46 Axel Cleeremans: prohibits that definitely it allows for that definitely definitely knows for that there can be various. disorders in the higher-order system. There can be noise in the first-order system. So yeah, sure, misrepresentation. But I don't see this as a big problem. In fact, if you take, you know, Anil Seth had this slogan saying, you know, we're hallucinating all the time. What we call reality is just when our collective hallucinations agree. I think that's not entirely wrong. It's too strong, probably, but yeah, I think we do often misrepresent things.
41:29 Richard Brown: But I mean specifically misrepresentation between the higher order and first order. So I was meant to be talking about specifically misrepresentation between the higher order and first order states. So maybe the first-order system represents that there's something green in the environment and the higher-order system represents that you're in a state that represents something red. So the question is what is the
41:56 Axel Cleeremans: experience of the person? That's a bit radical. That's a bit radical. This is what is happening with color constancy, for instance, right? constancy, you see white. Input is radically different given the lightning in the environment.
42:18 Richard Brown: But we
42:20 Axel Cleeremans: all take it to be white, right? We all take it to be white. It's only when you have these illusions, the famous dress is a good example of that, where, you know, okay, fine. Then you see things in a way that is different from what you know they are. And the dress example was interesting because there were unusually strong individual differences in the way people perceived the dress so that you could see the dress in a completely different way than I see the dress. So what is this? It seems to me to be the case of identical input and misrepresentation at a high order level, right? Right.
43:03 Richard Brown: So I mean, that's an interesting case, but going back to the color constancy real fast. So do you really? So is it? I mean, I'm not an expert in this field, actually, but is it generally accepted that you that you consciously experience the color as white? Because it would have been it would have seemed to me that that's not obvious. I mean, so if I'm looking at a white wall, there's a shadow on it. Then this part of the wall that is in the shadow, I will judge it that it's white. And I will perceive that it's like, I guess, in a sense that there's a judgment about it being white, but I don't think I consciously experienced it as white. It seems to me that it looks gray. I would, if I had to paint it, I would pick a kind of grayish paint and paint that part gray, even though I, even though it still looks white to me in a sense, I think that makes sense to say it looks white, but I don't think you experienced it as white. Do you, do you think you, you do?
43:57 Axel Cleeremans: No, you're probably right that this is graded and that this is more about judgment and about experience itself. But in some sense, as soon as you ask someone to think about their experiences, obviously they're judging them. A lot of this, of course, is happening unconsciously. But I think there's also good evidence that this is at least flexible in terms of misrepresentation. I certainly think that you wouldn't perceive that gray in the same way as you would perceive actual gray. There may be cases where this is indeed the case and that those optical illusions where Indeed, you know, that is actually the case. I mean, the Adelson illusion is a terrific example of that, with these two identically great squares, but one is in the shadow, et cetera, et cetera. You experience one to be lighter than the other, but in truth, they're the exact same color. Exactly, yeah. So I think, yeah, there's no real judgment involved there, this genuine misrepresentation resulting from evolutionary mechanisms meant and perhaps learned mechanisms meant to correct your experience so that you keep getting it straight in terms of the color of things despite the different illumination conditions, right?
45:33 Richard Brown: On your view then it is going to be the case that when there is this kind of mismatch, the experience tracks the higher order state as opposed to what's going on.
45:44 Axel Cleeremans: Oh, I would say, I mean, you would have to say, to be fair, that it's all about interactions between these two systems, right? So it depends on qualities of the sensory input and whatever the first-order system is computing and on properties of the high-order system. But I would say experience follows the high-order re-description of the first-order states.
46:11 Richard Brown: Yeah. Because I guess part of me why I'm asking this question is because a lot of times you want to present this stuff as a way of joining all the theories of consciousness together so that we can all co-exist guys. Yeah. Workspace, higher order.
46:26 Axel Cleeremans: Not all of them, but coding.
46:28 Richard Brown: Yeah, not all of it. Exactly. There's some unnamables we won't mention, but a lot of them. And while I like the spirit of, you know, let's try to see how far we can get with the different pieces of the different theories. But ultimately, it seems to me like the theory either comes out to be like a higher order theory or a global workspace theory. Because what's doing the work of consciousness is the higher order part. And then the global workspace stuff you have a role for it there, but it doesn't really play a role in the generated a conscious
46:59 Axel Cleeremans: experience. I agree with that. Yeah, so if I had to choose a camp, it would be high order definitely. So this is yeah that attempt is a sort of. An attempt at unifying ideas that I've sort of developed myself, but that are connected to so many other things now. that I think the way of making it work together, but at the core, it definitely remains a higher order perspective on consciousness. Yeah. If I had to choose a camera, higher order, yeah.
47:30 Richard Brown: Yeah, I mean, you're welcome in that camp. It's not really about choosing camps, but I was really, I'm really curious about whether there can be a kind of melding of these theories of consciousness. And it just, it just seems to me that there, in a real way, can't. That theory is either going to ultimately at the end say, look, some kind of inner awareness is what's responsible or important for consciousness, or some kind of global availability is. And the two may go together and play off each other in a lot of cases, but one is gonna be the one that the theory of consciousness really hangs its hat on. Yeah, I
48:05 Axel Cleeremans: agree with that. About the global workspace, I think, well, if you think of the brain as a big re-description system, where you have these hierarchies, of systems that we describe other systems, right? In the way we talked about earlier. Then the sort of global workspace architecture emerges naturally out of that. you know, you connect all the central bits. I've had slides like this, you know, almost joking about it. But if you connect the sort of higher order, the higher level, higher order systems together, then you end up with a functional architecture, which is exactly like the global workspace. And that can play out its functions of making content globally available to a variety of consuming systems, right? But if you think of it that way, the core business is representational re-description by high order systems, by hierarchies of high order systems. And then the functional aspects of the global workspace just fall out of it.
49:22 Richard Brown: Would it be fair to kind of say that what you're saying is that when the high order system re-describes the content, it puts it into a format that makes it available compatible with the other?
49:35 Axel Cleeremans: Yeah, sure. Probably, yeah. You would want it to be that way in order to achieve some sort of multisensory integration. Right. You would want it to be that way in order to withstand global availability, which is obviously a functional advantage. So you would want it to be that way. Exactly how that happens, I'm not so sure. That's where the connection with language and the social world, I think, is crucial. Through the social world, I think these higher level descriptions of ourselves ultimately are shaped by what we know about the mental states of other people, which we deduce from the interactions that we have with them. So I think there's a joint sort of human higher order architecture that we're sharing. simply by virtue of interacting with each other and trying to figure out the consequences of the actions that we direct towards other people, which requires us to make inferences using triple quotes, because I don't like that word if it is non-conceptual or non-abstract, but inferences about what the unobservable states of those other people may be. And I think those inferences shape our own high order systems, right? So there's that and then there's the common format. Yeah, language has to play a major role in that.
51:03 Axel Cleeremans: This is not to say that only linguistic organisms are conscious, but that we have as a species a big advantage through language in making not only the sharing of experience possible, but also the re-descriptions of whatever we're doing for ourselves, right?
51:22 Richard Brown: Right, yeah. And I think the social component is really important, too. Well,
51:26 Axel Cleeremans: I was
51:28 Richard Brown: going to say, I think that's a really important point that you're making about the social component, that a lot of theories leave out that component. So it's good that you are calling attention to that. Well,
51:41 Axel Cleeremans: most theories leave out the subject And I was always, I was really struck by this quote by Frege, you know, who says, mental states do not rough about the world independently without an error. The mental experiences presuppose the experience of the subject. He didn't say it that way, but the mental world presupposes the person whose mental world it is. And I mean, yes, obviously. When people talk about the differences between conscious and unconscious information processing in sort of isolated chunk of the brain, I feel that we're missing some fundamental aspect of it, is that whatever the differences may be, there are differences only to the person who's you know, whose mental states it is. And if we're trying to figure out how that works, we cannot sort of skew the subject and pretend it doesn't exist, right?
52:44 Richard Brown: Exactly. Oh, I don't know if you can see this, but I have yours.
52:46 Axel Cleeremans: Yeah, I know. I was going to say, only I can figure out that this is Joe's t-shirt.
52:52 Richard Brown: Exactly. I'm not putting
52:56 Axel Cleeremans: mine on, but…
52:58 Richard Brown: You said copyright. No self, no fear.
53:00 Axel Cleeremans: That's it in a few minutes. He said this in Kraków and it struck me and I used it in my own slides later on because, yeah, OK,
53:08 Richard Brown: fine. Exactly. So
53:14 Axel Cleeremans: the social embeddedness is the second aspect that's missing. Some theories, of course, have pushed this. Graziano and also Wolpert in a different way. Peter
53:28 Richard Brown: Carruthers has had places talk about this. Yeah. Do you know Peter Carruthers? Yeah,
53:35 Axel Cleeremans: yeah. Carruthers, of course, are links between theory of mind and introspection. Yeah, exactly. Or self-awareness. So I think those links are very strong.
53:47 Richard Brown: So one of the questions I had about your views on this was, I mean, obviously human-human interaction is very important, but do you think a human-animal, non-human-animal interaction play any role in this at all?
53:58 Axel Cleeremans: Absolutely. There may be interesting differences. Nobody knows anything about this, I think. I mean, some animals are obviously social, other animals are much more solitary. What about ants? You know, they are fundamentally social animals. Does that play a role in whatever experience they may or may not have? I don't know. We're completely speculating here, but I tend to think you need a combination of criteria if you have to decide who's conscious. And I would say, you know, you need a brain that's sufficiently complex and I mean sufficiently totally undetermined because who knows? You need a system that's capable of learning about itself over people and the world. So learning is an ingredient in that. And I think you also need emotion in a social environment from which models of yourself can be built, basically. So yeah, I think it plays a big role. Of course, it makes outlandish predictions. So a totally isolated organism that has no interaction with anything else but inanimate objects wouldn't be conscious. Who knows? Maybe that's true. I would think. If I were to go out on a limb, I would say an organism like that maybe wouldn't have much of conscious experience at all. But who knows?
55:28 Axel Cleeremans: And there's no way of testing these ideas in either case. There's interesting examples of three-year-old children that certainly suggest that, but those are mostly anecdotal. But they are certainly suggestive that your conscious experience of the world and of yourself is profoundly shaped by the organisms or agents that you interact with. There was a relatively recent case, I think, of a funeral child found in a chicken pen in South Korea. And he had been raised there from three, I think, till when he was found, 13 or something like that, with minimal interaction with human agents who were feeding him like, you know, like they would feed the chicken. When he was found, he was non-linguistic, obviously, but he was also behaving as a chicken, which I found, you know, that was stunning. He would get into a room and, you know, jump on a table and stand like a chicken, make chicken noises and so on and so forth. So
56:40 Richard Brown: there's no
56:42 Axel Cleeremans: way of doing experiments about this, but there are a few compelling anecdotes like this which suggest that at the very least or the experience you have of the world is shaped by the agents that you find yourself embedded with. And if those agents or human being or other human beings, then some of their phenomenology sort of leaks out onto yours, right? So some of their behavior leaks out into yours.
57:11 Richard Brown: Very interesting. Yeah, wow, that's really fascinating. I can't keep you here forever today, but I wanted to ask you, I've come up a couple of times in the background, but I didn't get the chance to ask about it, but this idea of the gradation of consciousness. So I know you think that it's important that we somehow reconcile the competing evidence that we have for consciousness being a kind of dichotomous phenomenon on or off versus also there's some suggestion that there's these gradations of consciousness. So I wonder if you could just say what you think is important about that distinction Well,
57:51 Axel Cleeremans: I'm not so sure it wins out anymore. I think it's, again, one of these very hard questions that I – yeah, I've been thinking about it for quite a bit, and I don't know. Well, first of all, you have these phenomena where continuous changes in one dimension result in dichotomous changes in another dimension. A good example is water. You apply continuous changes to the temperature of a body of water, and then at some point you have a nonlinear transition where it freezes, basically. So in a way, this freezing is a phenomenon that involves both continuous changes and nonlinear changes. That is why the question is complicated. The reason why I think consciousness is graded is really stems from work with connectionist networks for which it is a big feature. The fact that information processing involves these graded changes, right? So it's not the case that a piece of knowledge is there or it isn't there. There's a sense in which A concept even can be sort of represented halfway because the notion of representation is radically different in such systems than they were and than they are in classical systems, right? If you think of cognition as involving propositions, then okay, you're stuck in an all or none world, right?
59:24 Axel Cleeremans: A proposition is there or it isn't, it's true or it's false. If you think of it in terms of connectionist systems, then it's a completely spatial world where representations are points in a multidimensional space. Greaterness is a very natural consequence of this style of cognitive representation. where you have all these intermediate states that you can at least represent in some way that is meaningful when connected with other half-baked states, basically, right? So it opens up this completely different way of thinking about what a representation is. And so the ideas I have about consciousness being graded are sort of derived from that because if you associate phenomenology with activation of higher order systems, and those systems are graded systems in virtue of them being neural systems that use these principles of spatial representation, then at least it opens up the possibility for experience to be graded as well. From an empirical point of view, there's also very interesting differences between Connecting all or none reports versus collecting graded reports, and I've done quite a bit of work on that at some point Where you know you can compare Reports such as I have seen it versus I haven't or reports collected using measures like paths the perceptual and awareness skill developed by Morten Overgaard, and you get different patterns of results.
1:01:09 Axel Cleeremans: You do one or the other, right? Or if you use more binary measures like post-decision wagering, for instance. We've run experiments where people are exposed to simple mass-shaped and asked to decide which one was presented, and then they have to ask to express a confidence about this. about the decision, they say it's square, how confident are they that it was a square, you get radically different renditions of this using pass versus all-or-none judgment or using post-decision wagering, for instance. All sorts of biases, for instance, apply in the case of post-decision wagering that do not apply in the case of pass. Likewise for blindsight. Blindsight is also a good example where, in fact, if you use graded measures to question blindsight people about what they see in their blind field, even though they claim they don't see anything, but you ask them to express confidence or to rate their experience, then it turns out that there's much less of an association than you think is the case if you use binary reporting methods. So the binary reporting methods often have led to the idea that it is all or none perceptual states. But in fact, if you use appropriate measures, or at least more sensitive measures, I'm not really sure they're more appropriate, then you do capture this gradedness to a larger extent.
1:02:50 Richard Brown: So when you talk about the gradedness, I wonder, so the kind of stuff that you're talking about is you ask someone how On a scale of one to four or one to five, how visible was the stimulus? One being clearly visible, four being… That is
1:03:03 Axel Cleeremans: the percent scale.
1:03:05 Richard Brown: Yeah. OK. So if someone marks an intermediate score on that of a three or something, then you're suggesting that we infer that they had a kind of partial experience, but not a full experience? Or is it that because The way that I would find it natural to describe that is that they had a full experience with a degraded content. So the experience is all or none, but the content can kind of go from very vivid to sort of degraded.
1:03:37 Axel Cleeremans: And I told you that my thinking was changing about this a little bit, and I'm almost convinced that this may be the case. Okay.
1:03:46 Richard Brown: So what's the reservation? I think that's interesting. Why, what's the thing that's holding you back from being fully convinced?
1:03:53 Axel Cleeremans: Oh, I think it's just ideology. I think I would like things to be all graded, but maybe there's no good argument. But that is the tricky bit with this question. Is consciousness graded or all or none? And the tricky bit is that you can make, you can make binary stuff out of graded stuff. You can make binary stuff out of binary stuff and the reverse is true as well. You can make graded stuff out of binary stuff and you can make graded stuff out of more graded stuff. And so it becomes a little bit difficult when all these possibilities exist to figure out what is the case and what is the sort of decisive empirical strategy that you could deploy to actually answer that question. And so far I would say, well, there's a little bit of arguments for both. But in particular, you know, global workspace has always defended the idea that consciousness is this very strong all or none phenomenon. And I think that if you look at the empirical data that is appealed to, then you can sort of empirically sort of try to attack that claim a little bit at least. But whether it is the case that Awareness is always on and on, and then you have gradedness in the contents, or whether the accessing mechanism or the high order mechanism can itself be graded.
1:05:18 Axel Cleeremans: That is a more difficult question. I still tend to think that because the higher-order systems are structured in the same way or appeal to the same sorts of mechanisms than the first-order system they're observing, there is also at least potential for gradedness there. And I think expertise can sharpen your perception in ways that go beyond simply better perception of the first order contents, right? So you may appeal to attention there, you may appeal to the precision of your priors or other germane concepts, but I think there is some gradedness that's possible there. I remember flying with a pilot over North America as we reached, well, your location, the coast, you know, He was just right next to me looking out the window and he was pointing out all these landing strips that he was… And he
1:06:13 Richard Brown: was like, I don't see them.
1:06:15 Axel Cleeremans: I don't see anything at all, right? But I can imagine myself in a situation where I learned to pay attention to the distinctive features of landing strips as seen from the air and then ending up with Yeah, some gradedness in my experience of perceiving these little landing strips. Given the fact that perceptually, first order wise, all the data is there, right? So there has to be a difference in high order. I'm seeing these landing strips just as he does, right? And I may find myself in situations where I thought I saw a landing strip, but I'm not sure and things like that. And so does that characterize my experience itself? Well, I don't know. That's a big that's a big interesting issue.
1:07:06 Richard Brown: Yeah, that's interesting. And this is where you're the kind of view you're trying to defend here is different than like Sid Kouider's partial awareness hypothesis, right? Because he wants to suggest that you have the full awareness, but just different levels of processing, whereas you're saying, no, there could be kind of a real partial awareness, not just a fully awareness of a partial content.
1:07:29 Axel Cleeremans: Yeah, that's a very good point. I'm a big fan of partial awareness. I think he nailed it with that paper. I've often asked why he wouldn't continue to develop it. And he said, oh, I've lost interest in partial awareness. He's moved on to other things. But I think that that was really a great paper aimed at capturing how difficult these questions may be and that offering a sort of solution that both preserves this idea that awareness is an all or none phenomenon, but yet that you can get greatness out of it if you assume that it's organized layer by layer in a sort of complex hierarchy. I thought that was a very elegant solution to the sorts of issues that we're discussing here. And I would be, yeah, as I said, I've become a little bit more agnostic about it now, but I would, yeah, be ready to sort of buy it. I just wish he would do a couple more experiments about it.
1:08:32 Richard Brown: Okay, well, maybe you can convince him. Got him down. Okay, so I know your time is running short here. But I'm wondering, is there anything we haven't talked about yet that you would like to have come up?
1:08:55 Axel Cleeremans: This was a great conversation. I'm not really sure there's any other point that feels essential. I think, yeah, maybe one point is that in these ideas, there's also the central idea that we care about our experiences in ways that nonconscious organisms do not. And I think the caring, and by this I mean it matters to us. Experiences have meanings. I'm often struck also by how little these sorts of considerations enter in discussions about consciousness. And this is maybe where that sort of line of thinking connects with the work of Joe LeDoux and his interest in consciousness. coming from emotion, basically. And so I feel this is truly important as well. Experiences have meaning for us. And you may even say that experiences are what drives us to do anything at all. You know, why would you do anything at all if the doing wasn't doing something to you? Why would you do anything at all if the doing wasn't doing something to you? That strikes me as a fundamental difference between systems like AlphaGo, for instance, even where they're endowed with high order monitoring systems of the kinds we've discussed, and actual living organisms, basically.
1:10:28 Richard Brown: So you don't think AlphaGo would have the same kind of… I mean, not the same kind obviously, but any kind of consciousness experience, even with the right kinds of higher-order monitoring systems?
1:10:39 Axel Cleeremans: Yeah, I don't think that would be sufficient, definitely. What's missing is of course the question we're all trying to answer. But I think this is where these discussions about consciousness connect with fundamental issues of life and death in biological systems. I think, you know, there's nothing that would prevent us, I think, from building conscious artifacts, but they would have an experience that is fundamentally different from the experience that biological organisms have. You know, they can reset themselves, they're infinitely duplicable, they're immortal,
1:11:20 Richard Brown: But mostly
1:11:23 Axel Cleeremans: what strikes me is that no artificial system wants anything for its own reasons. That had this great quote as well. He often has great quotes, but he said in a recent talk, how do we go from doing things for reasons to having reasons for doing things. That absolutely nails it, the difference between the sorts of systems that we are and any AI system at this point. Yeah, it's tightly connected with those ideas of Clark and Karmiloff-Smith about representational re-description. It's by re-describing our own states that we can have reasons of our own because we know what matters to us. So that is what is doing the job here, I think. Yeah, I think that's right. All these issues about caring about your own states and doing things to find yourself in particular states and so on and so forth. I feel this is an absolutely central aspect of discussions about consciousness that is almost never sort of discussed.
1:12:41 Richard Brown: I mean it's kind of interesting because you know Ned Block, his famous, his now famous critique of higher-order theories is partially based on this idea that consciousness matters and he thinks that we can't account for that, but actually it's the opposite. It's hard to explain how something could matter if you weren't in some sense aware of it being happened to.
1:13:02 Axel Cleeremans: That's, yeah, that's exactly what it is, absolutely, absolutely. I think the critique, I mean you discussed this in your own recent paper, right, this over-enterprising critique of high order theories is just misleading, it's just, or mistaken, it's not It doesn't need to be that way. Perhaps in some of Rosenthal's writings, it was that way.
1:13:29 Richard Brown: It depends on a very uncharitable reading of his writing. You could get that out of his writing, I think. But it certainly is not something he's ever defended. I mean, the problem is, you know, you want to say things kind of quick and zippy in your introductory stuff, and then most people, that's all they read. They come up with the quick and zippy takeaway, but they don't spend the time to really try to figure out the details. I think that's part of the problem. But yeah, so I think that this debate, I mean, it's interesting because some philosophers argue, you know, consciousness doesn't matter at all. There's nothing important about it. There's a philosopher, Geoff Lee, who was a student of Block, actually, who makes this argument. A couple of other people thought it's really hard to make that argument, I think, because
1:14:07 Axel Cleeremans: it's really hard. It's all that matters in a way. Right. It's all that matters.
1:14:13 Richard Brown: It's what we experience,
1:14:16 Axel Cleeremans: right? It's all that matters. There's nothing else. It's what we experience. So that has to be at the core of everything else.
1:14:25 Richard Brown: That's absolutely right. Yeah, that's absolutely right. Well, thanks again for joining me. I don't want to be sensitive to the time constraints here, so I want to make sure that you have time to get off the computer. But I want to thank you very much for joining me. I really enjoyed this discussion. And all your hard work behind the scenes promoting the ASSC mentoring young scientists and just being a kind of person that we can, you know, as a person coming up in this area, you've always, your name is one that we hear a lot and that you're recognized as a very friendly and welcoming figure in the field. So I want to thank you for that as well. Really great.
1:15:07 Axel Cleeremans: Thanks, I appreciate it. ASSC is, you know, it's just a wonderful community. It really changed my life to get connected with this budding society when it started. And it's just a great group of people. Zoltan Dienes, a good friend of mine, and a well-known psychologist himself, has always said that it's the best meeting ever. Because there's no other meeting I know of where philosophers and neuroscientists and psychologists can really share ideas in the same way. And that makes it a wonderful community.
1:15:44 Richard Brown: Exactly. And especially important is that it's an environment where we can first cut the bullshit. And the scientists in the… I mean, I go to Tucson as well, but Tucson is just a different kind of select.
1:16:01 Axel Cleeremans: Yeah, I've been once. I had lots of fun.
1:16:05 Richard Brown: Exactly. It's a burning man of consciousness at this point. It's a burning
1:16:09 Axel Cleeremans: man of consciousness. Exactly.
1:16:12 Richard Brown: But that's what… That's why it's important to have places like
1:16:16 Axel Cleeremans: this.
1:16:20 Richard Brown: Yes, exactly. No, no, no. No burning that for me. Okay. Well, thanks again for joining me. Hold on on the line for a second and I'll say bye after we stop recording.