Latest / Mahon McCann / Jelle Bruineberg - How SOCIAL MEDIA manipulates your MIND to change your BEHAVIOUR

Transcript

Jelle Bruineberg - How SOCIAL MEDIA manipulates your MIND to change your BEHAVIOUR · Mahon McCann

• 0:01
  SPEAKER_00: Thank you so much for joining me today, Jelle.
• 0:03
  SPEAKER_00: I've been, as I was saying beforehand, avidly reading one of your latest papers on adversarial inference and predictive minds in the attention economy.
• 0:14
  SPEAKER_00: And it's become, I mean, it's, I really think it's ahead of so much of the literature, which is why I really thought it was, it was a good paper to share with people because obviously there's a lot of talk about attention.
• 0:25
  SPEAKER_00: There's a lot of talk about the attention economy, what AI is going to do to that situation and just kind of ramp it up.
• 0:31
  SPEAKER_00: And so it's refreshing to see such a well thought out perspective on it.
• 0:36
  SPEAKER_00: And so I suppose to start off with people, what, how do you think about the attention economy?
• 0:42
  SPEAKER_00: You know, what is adversarial inference and maybe how does that differ from the normal presentation that we see in the literature?
• 0:51
  SPEAKER_01: Yeah, thanks Mahon.
• 0:52
  SPEAKER_01: Thanks a lot for having me on your podcast.
• 0:57
  SPEAKER_01: So maybe let's start with kind of the normal presentation of the attention economy as you find it in a lot of the different literature.
• 1:03
  SPEAKER_01: And basically I found that everyone points to a particular chapter by Herbert Simon, a cognitive scientist and economist, from already I think the end of the 60s, early 70s.
• 1:16
  SPEAKER_01: And this particular phrasing shows up in almost any kind of treatment of the attention economy.
• 1:22
  SPEAKER_01: Abundance of information will create a poverty of attention.
• 1:27
  SPEAKER_01: So the idea that attention and information are inversely proportional, the more information there is in the environment, the more difficult it is to selectively allocate those
• 1:42
  SPEAKER_01: Those information selectively kind of process that that information that is what according to that view attention is.
• 1:49
  SPEAKER_01: And so everybody kind of takes Simon kind of for granted like this is this is roughly kind of the situation that we're in there is too much information in the environment and it's very difficult to kind of pay attention to the relevant information in that.
• 2:03
  SPEAKER_01: Now, I think if you look under the hood of that kind of view of information, there's a couple of assumptions there that are based on very kind of prevalent and classical assumptions in cognitive science that are more recently have been put under pressure.
• 2:17
  SPEAKER_01: So the idea is that really kind of perception and action are kind of perhaps kind of
• 2:23
  SPEAKER_01: there's kind of serial process of like a parallel processing going on there at the same time but there is really this kind of core of cognition this kind of mind part of cognition and that's a parallel process and really attention is this funneling this bottleneck of how you get from all the information out in the environment to in the end kind of a mind kind of doing only one thing at the same time and there you see that one the
• 2:54
  SPEAKER_01: It's really a perceptual kind of problem, the attention economy.
• 2:58
  SPEAKER_01: It's really this problem of getting from all the information to the single mind.
• 3:02
  SPEAKER_01: There's not much to say there about action.
• 3:05
  SPEAKER_01: and there is kind of the assumption kind of built into that that if you have a poverty of information then supposedly we should live in attention rich kind of time so before some threshold before some time in in the recent history we should have all been able to completely find kind of pay attention to exactly those things that matter to us and if you look at kind of contemplative traditions from all over the world
• 3:29
  SPEAKER_01: over the last millennia, kind of like if there's one thing they all kind of converge on is the idea that kind of controlling your mind, controlling your attention is incredibly difficult to do.
• 3:39
  SPEAKER_01: And you'll find people complaining about that kind of throughout history.
• 3:47
  SPEAKER_01: So this idea that kind of controlling your attention is somehow the default case and only in cases of information abundance do we get problems with that.
• 3:58
  SPEAKER_01: Yeah that seemed to me very problematic to this and that seems to be an implication of this kind of standard way of viewing the attention economy.
• 4:07
  SPEAKER_01: So I've been kind of thinking about different ways of conceiving attention and conceiving the attention economy and basically focused around the question kind of like so what is it about the digital environments that make it more difficult to focus and to concentrate?
• 4:27
  SPEAKER_01: The amount of information in the environment is also kind of a bit of an unspecific kind of notion.
• 4:32
  SPEAKER_01: Can we get to something more specific and also to something where we are more kind of the active participants in this engagement with the environment rather than the kind of the passive recipients of information?
• 4:46
  SPEAKER_01: And so I've been looking at a couple of things and one is this idea that well there's a very prevalent idea of the mind as itself a kind of prediction machine and it's kind of interesting parallel between kind of algorithms as prediction machines and minds as prediction machines and if you frame the question in that way then it seems more like well if I'm a prediction machine that wants to
• 5:11
  SPEAKER_01: Gather, sample particular kind of information from the environment and start to manipulate the environment to receive the things I want.
• 5:18
  SPEAKER_01: And if these algorithms that underlie much of our digital technologies, much of the platforms that generate things like newsfeeds and so,
• 5:29
  SPEAKER_01: If they are built on similar principles and similar kind of goals that they have and those goals could be well spend as long as possible and have this user spend as long time as possible on social media.
• 5:43
  SPEAKER_01: How can I arrange their digital environment such that they keep there as long as possible?
• 5:48
  SPEAKER_01: Then there's this idea there's a kind of adversarial kind of relationship where I have my goals and I try to change the environment to kind of get those goals realized
• 5:57
  SPEAKER_01: This algorithm has its own goals and changes my newsfeed in order to get those realized and that's a kind of adversarial kind of relationship and there it seems that yeah if
• 6:10
  SPEAKER_01: These algorithms have so much information, not just about me and my behaviour, but also about the kind of interests and behaviours of people who are in systematic ways related to who I am, that that kind of tug-of-war seems to be lost by the individual.
• 6:24
  SPEAKER_01: And so that's what this idea of adversarial inference is supposed to capture, this kind of tug-of-war between predictive algorithms and predictive minds and how they play out.
• 6:36
  SPEAKER_00: Fantastic, thank you.
• 6:37
  SPEAKER_00: So that was a brilliant introduction to the whole thing.
• 6:39
  SPEAKER_00: And to go back a little bit, I suppose, the idea of this kind of information processing bottleneck when the broad Bentian kind of conception of attention, which you really, I had never said, I always knew there was something kind of that I felt was off about the Herbert Simon and the kind of
• 6:56
  SPEAKER_00: Williams takes that one up as well.
• 6:58
  SPEAKER_00: Davenport and Beck kind of work with the same thing and the idea that there was something missing there in a sense of this problem of action somehow being involved in it like just passive information processing.
• 7:12
  SPEAKER_00: But it's interesting that Broadbent was reacting to the kind of behaviorism of his time because there's this weird like link between the evolution of the attention literature and the evolution of information processing
• 7:23
  SPEAKER_00: of the information systems in general.
• 7:26
  SPEAKER_00: Because broadband, that perception and communication is a book where he uses the metaphor of the communication technologies of the time, the telephone, in order to describe the limited capacity view of attention that he came up with.
• 7:40
  SPEAKER_00: And I thought it was very interesting that you connected the hooks, the designers of the technologies, as being deeply behaviourist in the sense that they're trying to modify people's habits without much regard for agency, autonomy,
• 7:54
  SPEAKER_00: self determination or any of that and so in moving from the broad bending conception to selection for action theory which is what we'll talk about as well within predictive processing and is that do you think that's an attempt to get beyond the behaviorist thing to push kind of action back into the question as well yeah nice yeah yeah so so
• 8:16
  SPEAKER_01: Exactly the question how different views of attention are tied up with kind of pretty big theoretical... Metaphysics.
• 8:24
  SPEAKER_01: The metaphysical discussion... Massive meta... No, no, exactly.
• 8:28
  SPEAKER_00: Yeah, it's unbelievable.
• 8:28
  SPEAKER_01: That's super interesting to see how they relate to each other.
• 8:31
  SPEAKER_01: And I think also, like, Kahneman on his kind of quite early book on attention says something like, well, in times of the behaviorists, right, the very word attention was kind of forbidden because it was supposed to stand for this kind of internal process
• 8:44
  SPEAKER_01: basically start using attentional language when your kind of conditioning kind of story breaks down kind of like so okay well that breaks down then we have this inner inner resource we need we need to be able to talk about attention at some point but in a sense so my feeling was when reading this like the the the the hook model and uh i think like this this is um apologies i just switch off
• 9:13
  SPEAKER_01: So if you're reading Eyal's Hooked model, there's much to oppose against this because basically it's a guide on how to get people hooked on your project.
• 9:25
  SPEAKER_01: It's just plain out manipulation with very little concerns for ethical dimensions and so on there.
• 9:32
  SPEAKER_01: But it seems to me that
• 9:36
  SPEAKER_01: He had a better grip on the phenomenon than many of the people who write about attention, the attention economy in a more critical way.
• 9:44
  SPEAKER_01: So this idea, he introduces the idea of, well, sure, sometimes you have external triggers, so these are the things we know, notifications, etc.
• 9:55
  SPEAKER_01: Things in the environment that are supposed to capture our attention.
• 10:00
  SPEAKER_01: But this is like, well, yeah, if that's the idea of competition for attention, then that works only so far.
• 10:05
  SPEAKER_01: But in the end, you only perceive those kind of stimuli when you're already engaged with a particular device or a particular kind of technology.
• 10:14
  SPEAKER_01: So what is really kind of the next step is to not go for external triggers, but internal triggers, kind of these internal urges that you have whenever you have nothing to do or when you're doing something and you get a bit bored, suddenly there is this urge to
• 10:30
  SPEAKER_01: Check your emails, check your Facebook, check whatever.
• 10:35
  SPEAKER_01: And how to kind of manufacture those urges in the way that they connect up with the use of your product, that's what the IaaS model is about.
• 10:47
  SPEAKER_01: And that has indeed a much more kind of close connection between perception and action than these kind of classical kind of more information processing views of attention have.
• 10:59
  SPEAKER_01: So I thought there was in terms of the phenomenon much more to be gained in a more AR kind of way of thinking but at the same time of course kind of like that seems to give up on any kind of notions of agency or the fact that we could even try to control our attention or if we could even try to kind of
• 11:19
  SPEAKER_01: live up to the aspirations of things we call autonomy or so.
• 11:24
  SPEAKER_01: I was hoping that the predictive processing literature in relation to developments in what's called embodied cognition that also have typically worked with a much closer and more direct
• 11:35
  SPEAKER_01: connection between perception and action that those could kind of and capture the kind of the phenomena that Eyal is talking about and at the same time have these broader resources to to think about agency and autonomy and so on.
• 11:50
  SPEAKER_00: Yeah because you point out that split in the attention literature that Johnson and Fernandez look at between cause theories and effect theories and that the kind of split between that is like the voluntary top-down attentional control like where is that located and the kind of mechanism accounts like I suppose the broadbandian or the limited resources have a kind of homunculus smuggled into them that they don't really flesh out and so when you start to look into it it gets a bit
• 12:19
  SPEAKER_00: It's hard to defend.
• 12:21
  SPEAKER_00: Whereas I suppose predictive processing doesn't try and reduce it to one part of the brain.
• 12:26
  SPEAKER_00: It's, it's the entire, it's like a dynamical system.
• 12:30
  SPEAKER_00: Right.
• 12:30
  SPEAKER_00: Yeah.
• 12:31
  SPEAKER_00: Does that make sense?
• 12:32
  SPEAKER_01: Okay.
• 12:32
  SPEAKER_01: No, exactly.
• 12:32
  SPEAKER_01: So, so, so, um, so in the end, something like voluntary attention, of course, is something that we would want to explain kind of like, it is something that we, in some sense can, can, can, can do at least in cognitive sciences in the business of kind of finding a good explanation for this.
• 12:50
  SPEAKER_01: and many of these kind of older theories of attention and kind of like cognition as information processing in particular have this idea of kind of one area or one one kind of module or whatever that is kind of central central cognition that's why it's all coming together and that's really kind of the minded kind of thing and if that that if that thing makes a decision such as where to where to pay attention to where to direct your kind of sensory organs to then that's kind of top-down voluntary attention the problem is that that's kind of
• 13:20
  SPEAKER_01: scientifically there's there's not much evidence for such kind of place where it all kind of comes comes together and for the philosophically it's kind of kind of problematic problematic as well indeed that it seems to look like okay well now we have basically something with all the characteristics of a full-blown agent kind of located somewhere inside the brain to explain for agency or kind of voluntary attention that that that seems to be fairly circular um so
• 13:48
  SPEAKER_01: So that is a problem with these kind of course theories of attention.
• 13:52
  SPEAKER_01: They kind of reify this notion of attention is kind of at the disposal of this kind of central entity to kind of allocate wherever you see fit.
• 14:02
  SPEAKER_01: That's very much also in the spirit of kind of the way Simon writes about attention.
• 14:05
  SPEAKER_01: And that seems to be, I mean, I guess both wrong and outdated.
• 14:11
  SPEAKER_01: Yeah.
• 14:13
  SPEAKER_01: And so then turning to predictive processing,
• 14:17
  SPEAKER_01: Indeed, I think that the if there's one kind of aim of predictive processing is to give a kind of completely distributed view of how these cognitive processing processes work and that there's not very much tied to specific models doing specific things but the kind of just basically something like error minimization is supposed to kind of to do all these kind of psychological jobs in particular kind of kind of ways and what I found interesting is that this kind of
• 14:44
  SPEAKER_01: kind of more early critics of this of the broadband view of attention like people like Alan Allport and so but there's sometimes point to this kind of predictive processing ideas as well saying okay these are actually allies in this in this idea of eradicating this homunculus from the brain and actually thinking yeah control is something that is distributed throughout the whole brain there's no one seat of the brain where this is where perception or attention gets controlled
• 15:11
  SPEAKER_01: But for someone like Alport, it's really kind of like, well, control is more about context sensitivity and seeing what context do you let your behavior be guided by.
• 15:21
  SPEAKER_01: And that could be distributed over different parts of the brain without having to assume this kind of inner homunculus.
• 15:31
  SPEAKER_00: Yeah I'm gonna throw some crazy stuff at you now in terms of that particular topic because it's something it's because it often gets linked to like executive function or somehow it's in the prefrontal cortex that there's this you know homunculus or whatever but in interesting I don't know if you're familiar with Russell Barkley's work on ADHD
• 15:50
  SPEAKER_00: So he basically looks at ADHD as a failure of executive functioning, a failure of self-regulation.
• 15:57
  SPEAKER_00: And he describes people with ADHD as having like a myopia in time, like they can't generate the future hypotheses in order to guide, to inhibit.
• 16:09
  SPEAKER_00: He looks at all self-control as inhibitory.
• 16:11
  SPEAKER_00: So you're inhibiting one set of
• 16:14
  SPEAKER_00: action policies, motor control routine or whatever, and allowing another one to take place.
• 16:19
  SPEAKER_00: And you can look at this, I think he looks at it in kind of a Vygotskyian way that this evolved from, like our Vygotsky looks at speech in terms of that we learn to speak externally and then we learn to suppress speech in order to think.
• 16:33
  SPEAKER_00: And he makes a similar argument about behaviour, that we learn to inhibit behaviours in order to generate possible other ones.
• 16:42
  SPEAKER_00: Yeah, this because again, I've heard the executive functioning thing is kind of some people say it's bankrupt.
• 16:48
  SPEAKER_00: Some people say there's nothing in it.
• 16:50
  SPEAKER_00: But I'm curious because he has great success in treating people with ADHD, which is like an inability to regulate attention.
• 16:58
  SPEAKER_00: That's I mean, it couldn't be any more like the problem of the attention economy.
• 17:02
  SPEAKER_00: And he treats them with writing, which is really interesting.
• 17:05
  SPEAKER_00: Like he gets them to externalize everything.
• 17:08
  SPEAKER_00: So write down everything in order to keep yourself accountable because they can't keep themselves accountable internally.
• 17:13
  SPEAKER_00: So you're bootstrapping, I suppose, their self-regulation with all these by changing their environment.
• 17:20
  SPEAKER_00: And so, yeah, I wonder what you think about that, the executive, the inhibitory executive functioning as a kind of locus of agency.
• 17:30
  SPEAKER_01: Yeah, no, it's good.
• 17:30
  SPEAKER_01: Yeah.
• 17:32
  SPEAKER_01: So I've
• 17:34
  SPEAKER_01: I think for a long time I've been just very skeptical about the very idea of executive control exactly for the reasons we just talked about right so that it seems to be like well it seems to require a kind of kind of a CEO kind of a central executive kind of like that that does the regulation and it can fail to do so but then you still kind of presuppose this kind of thing and then when it's going right then then there is this thing supposed to sit there
• 17:56
  SPEAKER_01: I think there's a different way of looking at it that is more to do with kind of balancing priorities, right?
• 18:03
  SPEAKER_01: And so I need like this view of a kind of, as you say, a kind of kind of myopia.
• 18:07
  SPEAKER_01: I kind of like myopia and time where you only focus on one thing and tend to forget kind of other goals that you also have, but that could kind of shift your behavior away from what you're doing now, but you kind of tend to downplay them, deprioritize them.
• 18:24
  SPEAKER_01: Yeah, you can explain that I think without having this homunculus and if that's what you call kind of executive control, that kind of balancing of different priorities over time in a kind of context-sensitive way, then that seems to be completely fine.
• 18:40
  SPEAKER_01: And I think if you have this view that some of the embodied cognition people have, that Ellen Alpert also has, that there's just not really this kind of homunculus kind of thing,
• 18:54
  SPEAKER_01: Then it becomes kind of tempting to see that something like executive control is indeed from the very beginning something that is scaffolded by the environment.
• 19:05
  SPEAKER_01: So indeed this Vygotsky line where you say, oh well, actually the voice of your parents in guiding your behaviour, kind of like that's the first kind of scaffold.
• 19:12
  SPEAKER_01: At some point that gets internalised and you kind of like almost simulate yourself with the voice of your parents.
• 19:18
  SPEAKER_01: And what we call executive control is exactly that, that kind of
• 19:22
  SPEAKER_01: Socially scaffolded structure that you grow into and that becomes part of the structure of your mind and the structure of your behaviour.
• 19:33
  SPEAKER_01: And then that makes complete sense that kind of like, yeah, we do have structures, we always had structures that scaffold our attention and our behaviour.
• 19:43
  SPEAKER_01: And
• 19:44
  SPEAKER_01: I think we'll probably get to this, but one way of conceiving of the current state of being is that we carry around devices that provide very little scaffolding.
• 19:54
  SPEAKER_01: They provide access to all sorts of information, all sorts of action, kind of like with an effortless swipe of a finger.
• 20:03
  SPEAKER_01: Which is, if you have this idea that executive control is something that is supported and scaffolded by the environment, then that's just a very bad
• 20:10
  SPEAKER_01: Bad idea, because what we require is actually a kind of set of good constraints to allow us to make this prioritization and having everything available at all time is just not such a good scaffold.
• 20:24
  SPEAKER_00: I spoke to Sebastian Watzel, I don't know if you know his work on attention.
• 20:28
  SPEAKER_00: He did have a very similar point that like attentional or social media big a big problem with them as well is that they're like every every kind of space mush together you know what I mean like they're not it's like a library you go there to get books but this is also like a pub and a nightclub and a social place and like
• 20:44
  SPEAKER_00: government place where people give speeches and everything's just squished into this one platform so it's very hard for your brain to say like this is for this or it's for that because it's for everything so yeah it kind of it sucks you in in that sense because you can't differentiate what it's actually to be used for but to return to the agency question i sorry i'm bothering you about this but i think you're the guy to talk to in terms of him i'm also speaking to adam saffron i don't know if you know adam's oh yeah yeah nick
• 21:09
  SPEAKER_00: because he wrote a paper on a similar type of bottom up agency in this way but he is kind of bringing homunculus back into it in a sense of that their models that we create are represent in order to
• 21:25
  SPEAKER_00: to INHIBIT behavioural patterns
• 21:44
  SPEAKER_00: and we can kind of choose between them and you see that in a lot of the wisdom traditions as well I mean Stoicism has one embodying the sage or contemplating the sage where you imagine what a sage would do like what would Jesus do and then you model their behavior you model your own behavior based on their behavior which affords you like new self transcendence and patterns of action so I'm thinking because I'm writing about agency so I'm trying to figure out you know how would you actually
• 22:10
  SPEAKER_00: normatively describe that.
• 22:14
  SPEAKER_00: Is any of that landing for you or do you think that's, you know, is that off the mark?
• 22:18
  SPEAKER_01: Yeah, yeah.
• 22:19
  SPEAKER_01: No, so on the self-modelling bit, so I'm not quite sure if the kind of, right, so we can sketch the Fygotsky route that says like, okay, we'll have this external, barren, kind of fierce, kind of scaffolding, you internalise that and that starts to become the structure for your own kind of, kind of,
• 22:41
  SPEAKER_01: right basically it's a bit like the sage root where you kind of like okay what what what would my parents say in this situation like it's already voices that you aggregate together exactly don't do these dangerous things or maybe maybe i didn't exactly do it or just refrain from it because of this um and this this and so i guess i guess somebody like adam sefron might say that yeah that's exactly what self-modeling is supposed to supposed to capture this kind of this kind of fagotskian kind of kind of kind of root
• 23:11
  SPEAKER_01: I'm not too convinced that that's the right way to go and that these two ways can completely collapse.
• 23:17
  SPEAKER_01: I think with this kind of self-modelling everything kind of starts to basically
• 23:23
  SPEAKER_01: the agent starts to have a completely internal kind of model of itself and how it would act in the environment internally that it can kind of run you haven't completely yeah you need like i should say you have a kind of homunculus back and a kind of internal representation of the whole world kind of back in your in your head i don't think that's necessary but perhaps and i think that's a different and an interesting at least a different an interesting different kind of route compared to the fagotian kind of kind of route that really kind of has this kind of connection with the outside world from the very beginning and i think also until the very
• 23:53
  SPEAKER_01: um until the very end um and about kind of yeah the normativity of of agency i mean yeah of course that that's that's that's that's that's that's a super big um uh topic um so what i find find super interesting is this is this kind of cultural idea that
• 24:15
  SPEAKER_01: attention is the control of attention is supposed to be both kind of normatively kind of ideal and also kind of descriptively feasible or so right and of course like if it's descriptively unfeasible it's kind of a bit difficult to make it a normative ideal but so so this this cultural idea that kind of what we should be able to control what we pay our attention uh to and i think evolutionarily there's good reasons to to suppose that that no we're not we
• 24:45
  SPEAKER_01: Why would you, right?
• 24:47
  SPEAKER_01: I mean, kind of like, if something happens in the environment, it's kind of good to kind of shift attention and pay attention.
• 24:51
  SPEAKER_01: Lower your attention reflex.
• 24:53
  SPEAKER_01: Exactly.
• 24:54
  SPEAKER_01: So we kind of, ecologically, we kind of set up to kind of be interrupted by all sorts of things, and kind of also to interrupt ourselves to check every once in a while, kind of like, okay, we're doing this one thing now, but is everything else still going so good?
• 25:06
  SPEAKER_01: Yeah, okay, then we can continue.
• 25:07
  SPEAKER_01: So it's so this kind of
• 25:09
  SPEAKER_01: this kind of self-interruption which is almost by definition not a kind of intentional action but something that just happens to you this kind of internal urges that happen to you all the time that seems to be ecologically completely kind of understandable and in a sense also kind of desirable
• 25:28
  SPEAKER_01: and so from that perspective kind of our our very capacity to just sit in a room and read and write is kind of that that's actually the old situation kind of like okay well what what kind of experiment what what kind of environmental scaffolding is required in order to actually pull those kind of things uh uh those kind of things up so i think i tend to kind of
• 25:50
  SPEAKER_01: Perhaps yeah, I sympathize a little bit more with kind of like the kind of ADHD kind of kind of line of thinking where kind of like oh perhaps this kind of like this state of distraction is actually a more natural state of being and the question is kind of like if we want to live in a society like this how can we kind of somehow constrain such minds and kind of scaffold such minds such that they can sit still and read a book for a couple of hours.
• 26:15
  SPEAKER_00: That's 100% because I kind of tend to think that the default is actually attentional deficit.
• 26:21
  SPEAKER_00: Like the default state is just, it's actually hard to control naturally.
• 26:26
  SPEAKER_00: I think that it's cultural evolution that has driven our ability to, and you see that with like what you mentioned before, like the Stoicism is a good example, Buddhism is all about agentic control and they're
• 26:38
  SPEAKER_00: there are paths of discipline and they're very traditional and there's practices and you know there is an entire almost way of life in order to be able to to do that um but i think in the modern secular world we've kind of been left with a certain level of inability to scaffold those things to create environments that actually scaffold that and james williams pointed out he called them cultural commitment devices i think in his work
• 27:03
  SPEAKER_00: and that the lack of that leads to the opportunity for predatory attentional technologies essentially because the attentional technology is scaffolding our attention but not in a way that's necessarily conducive to our well-being or human dignity or you know autonomy or any of these kind of our democratic discourse like any of these ideals and
• 27:26
  SPEAKER_00: And so I suppose what I'm interested in is how could we rebuild some of that infrastructure to actually support that?
• 27:33
  SPEAKER_00: Yeah, yeah, yeah.
• 27:35
  SPEAKER_00: So positive scaffolding versus I suppose what you'd mentioned hostile scaffolding.
• 27:39
  SPEAKER_01: Yeah, yeah.
• 27:40
  SPEAKER_01: No, exactly.
• 27:41
  SPEAKER_01: And so, so I think I think the first step is to kind of the
• 27:48
  SPEAKER_01: to realize the precariousness of agency or so right so so i think a lot of a lot of the cultural debate has been kind of okay we just are more or less rational agents we can make decisions and so if you give people a different set of options yeah they might make make different kind of decisions but they wouldn't really undermine that kind of their very decision making or so in the first place and from a kind of more scaffolded kind of
• 28:10
  SPEAKER_01: perspective where you say like well it's it's we are only kind of those those sort of agents in virtue of this environmental scaffolding then then the whole situation looks a bit more precarious kind of like okay actually if you start systematically changing that environment and making it a hostile environment then then this might undermine indeed our very capacity to to make make to do things in the first place to kind of make it make actions in the first place and i think that is
• 28:34
  SPEAKER_01: Indeed like James Williams points it out as well kind of in the in I think yes it's tripartite structure in his book when the third one is really kind of like at some point you just you don't even know what's what's what's what's right anymore you can't even kind of conceptualize decisions anymore and so for me in terms of a positive answer kind of like I think that's where I depart a little bit from from at least what James Williams is saying in the in the book and he's
• 29:03
  SPEAKER_01: Some places he's kind of advocating for saying okay well if we only would reduce the amount of information in our environment then we would kind of regain this kind of kind of kind of things and I think from from from my perspective it wouldn't just be about although kind of putting barriers between you and information that might might be a good start but it's not just about kind of reducing information it is indeed about finding this particular kind of kind of kind of scaffold
• 29:26
  SPEAKER_01: um and one step there is is indeed i think that that resonates a bit what you were talking about with uh with Sebastian Watzel as well is is realizing how much spaces are scaffolding our our decision making so i sometimes indeed like make the things like if if i'm having two rooms uh at home a kind of
• 29:52
  SPEAKER_01: work kind of place, kind of library kind of thing, and a kind of play kind of place, reading the news, kind of newspapers all around kind of things.
• 29:59
  SPEAKER_01: At every moment I could kind of get up from my desk while writing and go to the other room, but it would be a kind of decision, it would be leaving the place, and so it's possible, it's not kind of impossible, it's not blocking me or so in one way.
• 30:17
  SPEAKER_01: But the fact that they are one room away kind of puts them at a little bit of kind of a phenomenological kind of distance or so like a little bit kind of less holistic kind of performances.
• 30:29
  SPEAKER_01: And of course what these technologies kind of offer kind of devices like we have now is that the switching between these different rooms kind of me reading a newspaper versus me writing a paper happen completely seamlessly and so having
• 30:46
  SPEAKER_01: Having some understanding and perhaps also respect for the way different places scaffold my decision making and just put particular options kind of for the time being like a bit further away.
• 30:59
  SPEAKER_01: How the structural environment does that, I think that will be a really important thing to bring into the kind of design of our technologies.
• 31:06
  SPEAKER_01: And so I mean, very practically, one thing I've been trying to do, it's been having different
• 31:12
  SPEAKER_01: users on my computer kind of like and saying like oh I want to have these kind of programs and these kind of things kind of tied to one user and then sure I can switch to kind of like to this other user where where I can can actually read the news and check Facebook and so and go go back and forth between them but that provides enough of a constraint such that you don't have this mindless kind of shifting you don't give in to this kind of internal internal triggers
• 31:38
  SPEAKER_01: All these things are of course completely difficult to do because all the infrastructure is built to have all these options available all of the time, to do exactly the opposite.
• 31:49
  SPEAKER_01: But I think making that switch, understanding that our agency is really helped by constraints rather than by more choices, that will be the first step.
• 32:00
  SPEAKER_00: Yeah that's I mean and it's such an interesting even for me studying I mean I did a literature review there on like all the ethical issues of social media and looked at like over 2000 papers and like read in depth like 200 and still I take out my phone more than I would like.
• 32:16
  SPEAKER_00: There's no connection between the like propositional understanding and the procedural like playing out of the behavior loops.
• 32:24
  SPEAKER_00: I still have to separate myself from it physically and like use like a lockbox or like have time where it's in another room and which kind of shows like what a level this is working on and I wonder there was something that you didn't it wasn't mentioned in the paper but I wondered what you thought about in terms of active inference and affect
• 32:42
  SPEAKER_00: like on what level the positive emotion because like dopamine the in the literature it's often related to a lot of dopamines driving a lot of these feedback loops like the infinite scroll the kind of anticipatory reward um and does that or does that come into your thinking is that something you've you've considered in it i was curious about it yeah so i i haven't really
• 33:07
  SPEAKER_01: I thought much about it specifically in terms of like the neurotransmitters and so right so so I know that both kind of reward and novelty are associated with with with the dopamine and I think this kind of general active inference framework kind of gives a nice kind of uh kind of way of integrating that and having it make sense that kind of one neurotransmitter kind of combines both of these both of these roles um and I guess in terms of
• 33:34
  SPEAKER_01: Affects, there's a couple of things to disentangle there, right?
• 33:38
  SPEAKER_01: So one thing is this notion of, in the paper called attention as salience, as a kind of, and that's almost like, how much does a particular action possibility stand out to you, kind of phenomenologically?
• 33:54
  SPEAKER_01: And so it could be a craving to urge to kind of check something on your phone or so,
• 34:02
  SPEAKER_01: I would say like those kind of things that you could kind of model as being a prediction error or so but they are very much effective right so they're kind of an in the moment kind of quite targeted kind of disachievement or kind of phenomenology sometimes people call this kind of a lack of optimal grip but kind of like you're being drawn towards a better optimum or so and it's very hard to kind of resist that pull.
• 34:29
  SPEAKER_01: So these are kind of like in the moment kind of effects that drive a lot of our decision making, our action selection.
• 34:39
  SPEAKER_01: And in that sense, I see kind of predictive processing very much tied up with this kind of more effective kind of notions.
• 34:46
  SPEAKER_01: There's the other question, of course, about how things like moods and this kind of more prolonged, more kind of stable kind of things kind of influence decision making.
• 34:55
  SPEAKER_01: And there, I think there's also this kind of
• 34:59
  SPEAKER_01: back and forth between if you're in a negative mood, you're more likely to kind of start scrolling and keep scrolling.
• 35:08
  SPEAKER_01: And the fact that you keep scrolling actually also kind of reinforces that mood.
• 35:12
  SPEAKER_01: So there you also have a kind of effective back and forth between these kind of scrolling behaviours and mood.
• 35:18
  SPEAKER_01: But these are probably a bit more complicated and less directly kind of targeted like the salience ones.
• 35:25
  SPEAKER_00: I'd heard Friston describe positive affect as entropy reduction.
• 35:31
  SPEAKER_00: In the cybernetic literature, it's moving towards a valued goal.
• 35:36
  SPEAKER_00: Positive emotion is tracking your movement towards a valued goal.
• 35:39
  SPEAKER_00: And Friston argued that that was because of entropy reduction fundamentally.
• 35:43
  SPEAKER_00: It's like free energy reduction.
• 35:46
  SPEAKER_00: And yeah, so I wonder, I mean, cause that goal structure is kind of running.
• 35:51
  SPEAKER_00: Cause again, what you talked about, like the algorithm is playing out a goal, essentially, that's trying to keep you in that kind of feedback loop while you're potentially searching for epistemic rewards.
• 36:01
  SPEAKER_00: And that's kind of where this, the conflict is playing out in social media because it's the curation algorithm.
• 36:07
  SPEAKER_00: So I think that kind of leads maybe to the relevance question a little bit, but I don't, I think you want to, did you want to say something on the entropy reduction question?
• 36:15
  SPEAKER_01: I think, yeah, I think two things.
• 36:19
  SPEAKER_01: So, yeah, so one, I guess some people working on the free energy principle and predictive processing and this kind of whole set of ideas talk about the difference between the low road through kind of predictive processing and the high road and kind of one, the high road gives kind of very principled kind of account of why everything cognitive is exactly the way it is based on some certain set of principles.
• 36:44
  SPEAKER_01: uh and the low road is more like okay well let's let let's assume this kind of set of ideas about prediction error minimization and and and see how it can be a lens to kind of uh to make sense of particular kind of kind of phenomena such as kind of attention and digital technologies and so uh and so in in the past i've been writing a little bit more on the kind of high road high road approach to kind of the energy principle but i think this is very much kind of kind of
• 37:09
  SPEAKER_01: kind of low road approach and I'm interested in this question about how if there is a way to kind of like find a direct connection between positive effect and particular very abstract dynamics in mathematical structures of entropy reduction.
• 37:26
  SPEAKER_01: That's an interesting open question but I think much of the story here can be bought without buying into those more high roads ideas.
• 37:36
  SPEAKER_01: I think what one other thing just to briefly pick up on is this is this.
• 37:41
  SPEAKER_01: So one half of the paper is about kind of adversarial inference in which there's kind of a clearly kind of there are clearly baddies kind of like like kind of algorithm set up to lure us into doing things that are not aligned with our personal level beliefs and goals and kind of aspirations and so forth.
• 37:58
  SPEAKER_01: But there's
• 38:00
  SPEAKER_01: I think there's a different story as well, which is that if you take something like an email inbox, right, which is a kind of structure that it's not adversarial, it's built adversarial at all, right?
• 38:11
  SPEAKER_01: I mean, my email provider has no kind of monetary gain to be made if I get more in more emails or spend more time on Outlook or whatever.
• 38:22
  SPEAKER_01: But it's still the kind of structure in which
• 38:25
  SPEAKER_01: messages information comes in with with very different kind of rewards some of them are just like oh forgot to unsubscribe from this newsletter kind of like been right away some of them might be kind of life-changing kind of decisions and they come in continuous throughout the whole day kind of um and and you need to check actually to figure out what's what what's what
• 38:45
  SPEAKER_01: And so that's not a kind of adversarial structure, but it's still one that is kind of set up to develop checking habits towards.
• 38:52
  SPEAKER_01: Because you're still kind of like, well, there might be something interesting in here.
• 38:55
  SPEAKER_01: And well, yeah, I did indeed check five minutes ago, but in the meantime, there's no rule that says nothing interesting has come in.
• 39:03
  SPEAKER_01: So that structure of just having continuous kind of variable reward, kind of potential variable reward kind of coming in, that sets up the conditions for developing things like checking habits.
• 39:15
  SPEAKER_01: I think at least some of the adversities that we that we encounter in digital environments are not even kind of like results of big bad companies kind of like and the business model underlying kind of attention economy but just this mismatch between kind of the kind of kind of agents that we are and and and that we think we are and and that we actually are and so it might make more sense for example in terms of scaffolding you know your environment to to constrain
• 39:43
  SPEAKER_01: Ideally also the amount of emails coming in but just even the temporality if you say well they only come in on every couple of times a day or every whole hour or so then it wouldn't make sense to kind of keep checking every every five minutes or ten minutes or whatever but you would kind of probably kind of track those more regular kind of kind of kind of intervals.
• 40:04
  SPEAKER_01: um so so yeah so so even without the kind of the the adversarial kind of big big big structures there is there is still this kind of mismatch going on
• 40:13
  SPEAKER_00: Yeah, and that's such a, I've heard that done in email inboxes of kind of reducing it to like one hour a day of when all your emails come in at the same time.
• 40:21
  SPEAKER_00: So you don't spend all your time checking.
• 40:24
  SPEAKER_00: And I think, is it, is it a problem with design?
• 40:27
  SPEAKER_00: Cause I kind of tend to think, I don't know, it's probably a bit of a generalization, but there is a very behaviorist kind of bend to design and to a lot of the tech space as well.
• 40:36
  SPEAKER_00: Even in the AI developers, a lot of them seem to have this very kind of determinist model of human beings.
• 40:42
  SPEAKER_00: That's not.
• 40:43
  SPEAKER_00: particularly respectful of people's agency or self-determination.
• 40:47
  SPEAKER_00: And it's kind of spawned another industry of people that are trying to build tools to afford people those kinds of self-regulatory things.
• 40:53
  SPEAKER_00: But I'm kind of interested in like, what is the move from, so like, if you could teach designers, like, Hey, there's, you've got this behaviorist model of people.
• 41:04
  SPEAKER_00: Here's the predictive processing model of people, you know, what would the difference be there in terms of how they think about the people that they're designing for?
• 41:14
  SPEAKER_01: Nice, that's a super good question.
• 41:18
  SPEAKER_01: So let me think for a moment.
• 41:22
  SPEAKER_01: So I think what I encounter in kind of the design literature are exactly the things that we basically started this conversation with, kind of like either people are extremely behaviouristic and it's kind of like how can we nudge people kind of
• 41:39
  SPEAKER_01: guide people, tell people, somehow convince, not even convince but just kind of like make people do things that are good for them basically and that's completely kind of behavioristic kind of like in terms of kind of conditioning or whatever kind of like set up the rewards in such a way that they do those things that they do exercise kind of that they do make a particular kind of amount of steps per day kind of those kind of things and the alternative is a pretty strong cognitivism I think where
• 42:09
  SPEAKER_01: Basically, so Tim Wu in his book The Attention Merchant has this example of the invention of the remote control and this was really a kind of device that was meant kind of like okay there's now commercials on television and multiple channels and so we invent the remote control such that people can
• 42:31
  SPEAKER_01: really decide what they want to watch with as little effort as possible.
• 42:34
  SPEAKER_01: Kind of explicitly kind of not even just branded but also kind of designed as a kind of autonomy enhancing device.
• 42:41
  SPEAKER_01: And of course yeah Tim Wu's kind of analysis there is kind of like it became so easy kind of like to use this remote control that people don't end up watching exactly what they want but they just end up kind of channel hopping and channel surfing the whole time.
• 42:57
  SPEAKER_01: So I think the main lesson there is that
• 43:00
  SPEAKER_01: People are opportunistic, right?
• 43:02
  SPEAKER_01: That people don't come to situations with kind of fixed goals and then look for means to realize those goals.
• 43:09
  SPEAKER_01: But what you want and what you think is relevant is depending on what's possible.
• 43:14
  SPEAKER_01: And that requires exploration of the environment.
• 43:16
  SPEAKER_01: And if you build environments in which there is an infinite amount of things possible, then what do people start doing?
• 43:23
  SPEAKER_01: They start to
• 43:25
  SPEAKER_01: scroll those environments to figure out what it actually is relevant and kind of like basically never stop never stop doing that um so understanding that people are opportunistic and therefore it's kind of good to constrain their opportunities such they can actually do those things that that they actually want to do i think that that that's the design click that that is to be taken and that is this kind of middle road between
• 43:50
  SPEAKER_01: There's in a sense kind of allowing, going with the behaviours and saying like, yeah, much of our behaviour is this kind of direct connection between perception and action.
• 43:58
  SPEAKER_01: That's what the cognitivists are wrong and the behaviours have a better story.
• 44:01
  SPEAKER_01: But that doesn't mean that we should kind of give up on things like agency and so, but find ways to scaffold agencies into our environment such that this kind of behaviouristic creatures can
• 44:15
  SPEAKER_01: I have these kind of loftier kind of goals and kind of achieve some degree of autonomy.
• 44:21
  SPEAKER_01: And so there, I think, I hope that a bit of predictive processing, a bit of embodied cognition, a fuse of relational autonomy, so all kind of can contribute a little bit to kind of have a story of agency and autonomy that's both feasible and aspirational.
• 44:37
  SPEAKER_01: So that's where I think both the behaviours and the cognitive base kind of have a less good story to tell.
• 44:45
  SPEAKER_00: Hmm, I think that really is the next, where I see if there was to be another iteration or another draft of a lot of these things, and that was a better one rather than a worse one, which is also possible, that that would be what needs to be considered more.
• 45:00
  SPEAKER_00: I mean, some of the literature, there's interesting metaphors in terms of like,
• 45:04
  SPEAKER_00: Social media is like the car before we had a lot of safety structures on highways, before we had seatbelts, before we had speed limits.
• 45:13
  SPEAKER_00: And so people are just racing around in the information landscape, you know, ruining themselves and it's having these big negative effects.
• 45:20
  SPEAKER_00: But really what's necessary is a rebuilding of the infrastructure, which is unfortunate because a lot of the regulatory environment
• 45:26
  SPEAKER_00: isn't really doing that it's more focused on the content regulation at the end to say like you know what what's good content what's bad content but without realizing that the infrastructure is really involved in what content is promoted what content people post and so it does seem to be a real choice point for that and what do you think in terms of AI do you think about what effect that's gonna have
• 45:52
  SPEAKER_00: I mean, I don't want to say like the apocalypse version, but the, cause AI, I tend to think of social media as like the first public facing encounter with AI, like Tristan Harris's argument.
• 46:03
  SPEAKER_00: This was like the first time people met large scale AI in the public sphere.
• 46:09
  SPEAKER_00: And I've seen how that's gone in pretty, in good detail.
• 46:13
  SPEAKER_00: And so.
• 46:14
  SPEAKER_00: Do you think the future of AI will be similar stuff?
• 46:17
  SPEAKER_00: Like are we just going to enslave ourselves to like a fake behaviorist AI god that can just, you know, make us, you know, give us infinite distractions?
• 46:28
  SPEAKER_00: Or do you think, you know, it's going to be something we haven't thought of?
• 46:33
  SPEAKER_01: Right.
• 46:34
  SPEAKER_01: Yeah, yeah, yeah, yeah.
• 46:36
  SPEAKER_01: Oh, good.
• 46:37
  SPEAKER_01: So when I read the literature, I find
• 46:41
  SPEAKER_01: I'm a bit surprised by the amount of fatalism that you kind of see.
• 46:45
  SPEAKER_01: People, someone like Harari in Homo Deus or so, right?
• 46:49
  SPEAKER_01: They just say, well, yeah, look, we are algorithms in essence and AI is an algorithm in essence.
• 46:57
  SPEAKER_01: And look, there's absolutely no reason to believe that we will kind of be better algorithms than AI.
• 47:02
  SPEAKER_01: So we're doomed, basically.
• 47:04
  SPEAKER_01: We're kind of like, we're this little stepping stone and kind of like AI will take over.
• 47:08
  SPEAKER_01: And I think I have the feeling that that's a kind of general kind of sentiment, kind of like, oh, we really stand no chance.
• 47:14
  SPEAKER_01: But I don't see why.
• 47:18
  SPEAKER_01: Why not?
• 47:18
  SPEAKER_01: I think I think many of these things that are taken for granted, this kind of like automatisms are really kind of decisions made by made by people.
• 47:29
  SPEAKER_01: And it's an open question how we want to kind of
• 47:35
  SPEAKER_01: what kind of role we want these AIs to play in our society.
• 47:39
  SPEAKER_01: So I don't see any reason for fatalism.
• 47:43
  SPEAKER_01: I just hope that the discussion, and on regulation and so, is based on the right kind of descriptive premises.
• 47:51
  SPEAKER_00: That's a key question.
• 47:57
  SPEAKER_01: As you say right so things like fake news and so kind of like take up a lot of kind of the regulatory attention and so and how can we regulate content or so and I think what I've tried to think about here is more like so how do basically kind of like how do these technologies as a whole kind of change our field of affordances a kind of realm of kind of possibilities such that
• 48:27
  SPEAKER_01: some things just keep standing out as relevant that are actually not really relevant and other things that are actually relevant kind of kind of it's very difficult to kind of perceive them as being the right thing to do and there there's more a question like how do this kind of kind of technology provide a set of particular choice architecture that that allows for doing many things at the same time kind of all complex kind of collapse together which makes it just cognitively just very hard to kind of keep your mind on what's what's what's relevant uh and that's
• 48:57
  SPEAKER_01: I'm not sure, I think there's a good normative debate to be had.
• 49:01
  SPEAKER_01: One source of hope, I think, is that a lot of the companies that build more like platforms, so something like Apple that builds the operating systems which we use on phones and computers,
• 49:19
  SPEAKER_01: They are not necessarily contenders in the attention economy, right?
• 49:22
  SPEAKER_01: So it's what matters for Apple is whether I buy a computer or not, not whether I spend as much time as I can on that particular kind of device.
• 49:32
  SPEAKER_01: So I think they have a bit more leeway to kind of actually reorienting these devices using the right kind of descriptive framework saying, okay, well, maybe we would want to make it possible kind of to
• 49:44
  SPEAKER_01: Navigate different spaces and different rooms on the digital environment such that you can put some of these distracting things a little bit more at the background without having to block them kind of all around.
• 49:54
  SPEAKER_01: So to think together a little bit more about how to build agency enhancing digital technologies and there I think neither AI or the attention economy necessarily stands in the way of that project but just kind of being a bit more observant about kind of human psychology in a sense.
• 50:14
  SPEAKER_00: But that's, I think that's actually what is standing in the way is actually our understanding of ourselves.
• 50:18
  SPEAKER_00: Like, it's a weird paradox that really the obstacle here isn't, like, if the technologists had a better way, I think they would actually do it.
• 50:26
  SPEAKER_00: I mean, if you had technologies that would promote well-being and people like using them, I mean, surely that's better press down the line.
• 50:36
  SPEAKER_00: yeah it is i think it's the descriptive limitations and which is why i thought your paper was so striking because it feels like it's on that edge like there i've been reading tons of stuff man but it feels like that just that that sense of the connection between agency and autonomy which requires this big picture metaphysics as well which is the problem with a lot of the attention literature is is that i you run into like three massive philosophical problems like the hard problem of consciousness
• 51:02
  SPEAKER_00: free will and determinism and the problem of perception so like you have to solve have a say something that solves all three of them in a sense to actually you know deal with it um yeah yeah so in a sense i started thinking about attention kind of like five five six years ago so i think and thinking reading like like james williams's book and so i think like oh there there is
• 51:27
  SPEAKER_01: obviously there's something to say about this from a kind of predictive processing and body cognition kind of perspective and i think that that has a better grip on these kind of issues than than what i see now but but yeah to to make it all come together and these kind of hurdles and kind of like all the different moves available in the attention literature and lots of people who just don't believe in attention in the first place or something and kind of like how to how to play that field and how to bring it all together such that you can kind of um yeah i mean it's
• 51:56
  SPEAKER_01: basically like so you have a slightly better understanding of the kind of descriptive playing field that can support a normative debate kind of like it's been quite a roller coaster kind of trying to
• 52:06
  SPEAKER_01: to change your BEHAVIOUR
• 52:37
  SPEAKER_00: philosophy anyway.
• 53:07
  SPEAKER_00: What do you think a unified theory of attention would require?
• 53:10
  SPEAKER_00: Or do you think you've presented one in... Are you trying?
• 53:14
  SPEAKER_01: No, no, no, no.
• 53:17
  SPEAKER_01: No, no, I definitely don't want to claim to have presented one.
• 53:23
  SPEAKER_01: So, I mean, I think one of my initial, like in my PhD work, one of the reasons to
• 53:32
  SPEAKER_01: get interested that kind of feels at least to me as a master's student it felt unprecedented and like okay this is a kind of revolutionary kind of kind of approach um i'm i'm i'm i'm i'm not sure that in the end kind of kind of kind of this version of predictive processing will be will be dead but i think the aspiration of
• 53:58
  SPEAKER_01: providing theories of cognition that are kind of like even if they're pretty pretty minimal but at least kind of like throughout having systematic kind of explanations of how perception, action, attention, memory, prior experience, how that all hangs together or so right so so much of cognitive science is kind of like I build a theory of memory of episodic memory or so and there's a little nice nice kind of model of it
• 54:22
  SPEAKER_01: What kind of like, oh, how does it hang together with the rest of the kind of architecture?
• 54:26
  SPEAKER_01: Oh, yeah, there's here perception and there's there there's action again, somewhere in between the memory, right?
• 54:30
  SPEAKER_01: And so the idea that you build cognitive theories in which all these things already come together, and that might be over simplistic.
• 54:37
  SPEAKER_01: And I think I think like the kind of predictive processing work at the moment is overly simplistic.
• 54:42
  SPEAKER_01: But still, it has this kind of kind of unified kind of aspiration.
• 54:47
  SPEAKER_01: and so I think one thing if anything uh uh I think that that's that's one thing I learned both from kind of reading Sebastian Watzel's work but also kind of thinking myself about about attention like if anything a theory kind of a theory of everything called attention is just a theory of everything cognitive kind of like attention just kind of a theory of a human
• 55:12
  SPEAKER_01: Exactly, exactly.
• 55:13
  SPEAKER_01: Attention permeates absolutely everything.
• 55:17
  SPEAKER_01: And it's just, it's the selectivity, it's the relevance, it's foregrounding, backgrounding and any kind of cognitive process involves those kind of, those kind of things.
• 55:29
  SPEAKER_01: And so, so it's futile to kind of carve off a little part of what we call the human mind.
• 55:35
  SPEAKER_01: Okay, let's call this a few attention and let's, let's see how it works.
• 55:39
  SPEAKER_01: it is really kind of it needs to explain how it how it hangs together with perception action embodied interaction with the environment memory etc only then i think can we can we can we aspire to uh can we yeah can we call something a unified theory of attention
• 55:54
  SPEAKER_00: Yes, I think that's what we need to rebuild the internet.
• 55:58
  SPEAKER_00: So it could be a while.
• 56:03
  SPEAKER_01: But I think the design spirit here is to do like a little back and forth and a little bit of a better one than the current.
• 56:10
  SPEAKER_00: Yeah, exactly.
• 56:11
  SPEAKER_01: Like a kind of theoretical modifications and then having design modifications and see how they kind of co-inspire each other.
• 56:17
  SPEAKER_01: We don't need to first figure out the philosophy before we can start building better environments.
• 56:22
  SPEAKER_01: That would be a bad situation.
• 56:24
  SPEAKER_00: we'd be in big trouble then we can do it incrementally okay well that's a hopeful vision i appreciate that yeah i appreciate your fantastic work as well i'm sure hopefully we'll speak again in the future um because i i found so much value in it um it's really good to see um super nice to be here yeah thanks