Why You're an Ant in the Attention Economy with Ronen Tamari
https://www.youtube.com/watch?v=jjRoTZFaGxk
Transcript:
(00:00) a community's trust graph like imagine that a DA like currently how does a DA decide who it's following like what people it's following what accounts it's following but the practice of sort of having a sort of community setting the people they're following is already kind of the that's part of the collective attention governance like that is a very simple sort of aspect of it there's almost like these epistemic epistemic Hills where it's like this is where most of the belief is and there's people sort of like little ants going out seeking
(00:28) out new sources of information [Music] hello uh talking about attention and I had just asked you about your background in in general yeah so uh final year basically final month PhD student um in Hebrew University in Israel uh and my focus was and natural language processing MH and yeah somewhere in the middle of my PhD this was also related to co I kind of fell down rabbit holes of of how these um big Tech AI Industries are controlling the data they're controlling the compute they're controlling the algorithms uh centralized social media
(01:22) platforms and as I was kind of following the the sort of rabbit hole of of how this control manifests itself and how they exercise control I felt like I was getting LED on to this idea of attention and how attention is manifested in in sort of digital online environments like the digital traces of human attention because attention is a very human thing but it also has very specific sort of uh technici that's this a a word I learned recently technici of attention like how software reduces attention to code yeah
(01:53) so yeah so i' been I'm very interested in that because it feels like it's a very basic building block of like how we interact with information online uh yeah we can go into more detail I there's a lot to talk about and I'm also just in your sort of angle on attention too I'm just looking right now I just pulled up the common sense makers because I'm familiar with both Brad graph and um and Daniel from active inference Institute uh did you want to talk about sure that a little bit sure sure uh so so after I kind of fell down the
(02:30) rabbit hole of of looking at these centralized platforms and and getting the sense that things are really really messed up in sort of online information ecosystems I started getting interested in uh decentralization and kind of web 3 areas and um started connecting with people in the web Tre ecosystem I started working in a company that did tools for Dows like um D Dow stack Dow tooling company and yeah through this this I was exposed to some of the Brad and and Daniel are both kind of active in the these Lial
(03:07) spaces of of you know decentralized data ownership and uh that sort of thing Daniel was really instrumental in in in my personal Journey because yeah he he's an uh insect intelligent like he he does research on collective intelligence and ants and he introduced me to this idea of stigmergy which I think was like if I had to like three concepts the top three concepts that have changed the course of my PhD and probably even more than my PhD then I think stigar would be one of them yeah that's the uh paper that Drew
(03:41) my attention the most which was from users to sens makers on the pivot of pivotal role of s stigmar social annotation in the Quest for Collective sense making yeah that one so that's that's a paper with Daniel and and it was kind of through his um through him introducing me to the concept of stigmergy and then us trying to sort of of synthesize that with you know going from insect intelligence to what it means for social media and and online information ecosystems what's a sort of concise definition of stigmergy stigmergy is the idea it's a
(04:14) coordination mechanism and it's the idea of indirect large scale coordination it's mediated through environment modifications so basically like the textbook example is ant pheromone Trails so ants leave pheromone Trails other ants um see those Trails react to them and uh it's all very local sort of environment modifications but you get this Global emergent intelligent Collective behavior from those local interactions and stigar is that was it was coined by you know a researcher that was trying to understand how that uh how those those
(04:47) signals and other kinds of signals that insects leave drive this kind of emergent collective intelligence and even in those systems do it like we can sort of hack their behavior by modifying chemicals making ants think they're dead making them continue to carry themselves to the dead pile um right there's a spiral of death I think there's also this famous video that shows how they yeah it has failure modes of course it's not like some kind of fail safe mechanism it's just a very powerful example of like when it's functioning
(05:17) normally it leads to you know a lot of very interesting behavior and I think for me the sort of aha moment was like Platforms in a sense are the ground right the ants like the ground that the ants put the pheromones on and it just got me to think that the platforms are like messing up our our stigmergy like the signals are Amplified in ways that we don't understand there we don't even you know control really what happens with the feromon so it' be almost like if you took an ant colony and started messing around with the
(05:44) properties of the you know soil and stuff like you said you cause a lot of you know crazy failures and and it feels like we're kind of starting at that place right now we try to have to figure out the what is even healthy stigmergic infrastructure look like for for communications and it's evolved quite a bit since the beginning of the internet from you know intuition which might be something like a 10 star scale to just well we don't have the space on the page for this and most users are only you know doing uh
(06:17) the extremes so let's just go to like to dislike to now they're actively removing the dislike and you know ants can't really design their system but we have that freedom but it does seem almost like it's more reactive uh on our end to Market forces at this point where there isn't a lot of Applied intelligence in the design of those signals that we're tracking and we've had like at least I think 5 years since the problems uh started to really bubble up and become readily apparent but there's a sort of desire to have them all be
(07:04) managed in the background passively rather than introducing some um some sort of active mechanism um did you want to talk about your just paper in general and what you sort of have found from the signals that we're tracking and uh maybe sort of the differences between like passive tracking of users versus uh active tracking which is sort of harder to get users to do but might be more valuable or sometimes even they might not even know their own self that well yeah um for sure you brought up a lot of really sort of interesting points
(07:44) that we can uh explore I think so talking about this kind of passive active distinction and yeah maybe you ask about the paper kind of more broadly so the paper is is is a really short paper it's basically just an extended abstract it's kind of call to attention we thought of it like it's trying to say hey this is a thing we really need to be looking at we're not really thinking about our relation to data in this way and this is kind of if I had to really summarize it I would say we have this kind of metaphor of you know data is the new oil
(08:16) M and we talk about how all these um social media companies and Platforms in general have managed to capture oil mine our oil and harvest it and extract it and convert it to Gold so there's this kind of pipeline of you know data kind of it's some kind of oil and they extract it and they turn it into gold and this paper is kind of trying to invite an an alternative interpretation where data is it's not oil that we're just kind of you know this sort of remains of like dead you know animals it's more like these are our markers that we like
(08:51) data is stigmergy it kind of invites us to be much more active participants like it invites us to think of each each each of us as like a kind of a neuron and a Collective brain and we're we're part of this collective intelligence network uh so it's a whole other relation to data it's like you're very like you know ideally right no one is going to do this in the beginning at least but you're very mindful of the traces you're leaving online what am I sharing with who am I sharing it how am I sharing it um and and what we talked about in the
(09:17) paper is is there's kind of two parts to this one is sort of our individual a mindset shift that like we become instead of like seeing ourselves as just passive you know Doom scrolling users were more active we coin the term sense makers so it's kind of like um we're trying to evoke sort of the hacker ethos where the hackers are like people that take ownership over the code you know they're producing and they create their own tools to you know create the digital spaces that they want to uh live in basically so yeah sensemakers are also
(09:48) creating their own information ecosystems and they have ownership over those tools and it's kind of trying to evoke that active sense of of of ownership over the data you're you're uh sharing and that's one part so that's the individual aspect but we also recognize that individuals are ultimately very limited in what they can do so even if you decide today I don't want to be user anymore I want to be a sense maker yeah you're set up for failure because the infrastructure is so lacking and it's so uh there's such
(10:16) powerful infrastructure raid against you that these platforms have you're very limited in what you can do by yourself and fight against Alm just yesterday I went through for probably the nth time uh just going through YouTube and saying don't recommend this don't recommend this don't recommend this but it just leaks in over time where you get into this passive deterministic cognition kind of mode where it's like I can't actively be controlling my attention all the time right and I and I I really sort of feel
(10:52) that attention is kind of it's kind of like breathing and that it can be both active and it can be automatic and most of the time it's falling into that automatic space and there are sort of boundaries on the extent to which you can control your tension so it really is important that we have these mechanisms of um what you I think you call in the paper or somewhere like the governance of attention itself um because that is defining so much of the public uh discourse is what is going to work itself through those filters and
(11:37) it seems after all this time either there's not a motivation to correct for the sensationalism that uh guides our attention or they haven't found a way that still manages to keep people on site like there's sort of a conflicting optimization plane there there where you have to have people on site and to do that you have to get people's attention and there might be some desires from some of the people working on those teams to align that attention with some better outcomes but if you aren't going to extract it then Tik
(12:17) tok's going to come along with their insanely powerful algorithm which I can't even leave that installed on my phone because it's it's so powerful and the model that they sort of have of me now on different platforms are sort of getting scarily accurate um which is it's just bringing up things that I'm almost actively thinking about at the same time it's becoming an extension of ourselves without any control but we should have some level of Rights over the direction and flow of our attention stream yes this is very
(13:03) much so there's yeah there's a lot of things happening because like you say the platforms have so much data over you that they can and you know so much data collectively that they've amassed from all the users that that they they track that they can build really good predictive models but in fact those models are kind of they're they're very dangerous in a way because they they're not models that are optimized for anything other than leaving you on the the site like they're not optimized for example for your long-term intellectual
(13:35) or emotional development or something like that um what is it what is an what is a model of sort of attention governance look like that does that like what ises a feed algorithm that sort of or or you know a platform that is trying to promote this kind of you know intellectual growth what does that look like it probably looks completely different than than what we have right now so they're they have a very certain kind of control that's very effective at the you know very short term and that's basically that's all they need because
(14:05) they're they're sort of exploiting our our you know our our worst weaknesses uh to to leave us on on the site and that's all it interests them so this is there's one of the papers we uh drawn heavily in in that paper is um I found a paper it's a book called Stand Out of our light and it's uh freedom and resistance in the attention economy it's by a Google former Google ethicist I think his name is James Williams and he has this uh mean that he calls from attention to intention and it's about like the kind of again this kind
(14:38) of passive to active mindset where these Technologies are purposefully sort of leaving us in this passive mindset they're they're entrenching us they're digging us deeper into the passive mindset because we're more predictable in the passive mindset as active users we're much less predictable much less controllable um they have much less that they can kind of even you know help us with it becomes something completely different like when the kind of tools we're looking at for example are like tools for thought those are something
(15:09) that are more close to sort of the active like sensemaker tools and not like a doom scrolling feed like Twitter so so they have very good reasons to want to leave us in this kind of passive uh domain so you've mentioned um Doom scrolling a couple of times and I wanted to talk a little bit about the role of dopamine and modulating attention um the sort of common way that people think about dopamine is like it's a reward that you get a little hit of when you get something good which is not really true it's it's more related to
(15:47) anticipation and um the modulation of attention over time and and goal oriented behavior and you know Asar at the center for uh Humane technology developed the infinite scroll and there's sort of like this always an anticipation that there's some other piece of novelty below the fold and so it's not like that we're getting a rush or a massive hit of it it's just messing up that entire Loop where the people who are most uh sort of addicted to social media actually have much lower levels of of from I mean from what I've
(16:30) I've read recently of uh of doine in their system but it's still that fundamental Loop that is modulating and controlling the behavior um like individually what do you have to say or what like thoughts can you you contribute to the the notion of it at a neurochemical level um or at like an individual level yeah so yeah the dopamine aspect is quite interesting and I actually I'm not an expert on on that um on that aspect uh it does I mean I feel like there's this kind of General I think reward you know reward is much more complex probably than
(17:16) dopamine and I I know there's like other signals I wish I could pull up the the names but there's all kinds of reward that we get and you know dopamine is just one of them and it's kind of again they're like hacking hacking a reward system to go after something that is going to sort of reliably keep us engaged but not necessarily good for us just like you know sugar is going to keep you like yeah make some food sugary it'll keep the person like it'll keep a kid happy what does a design where our our flows of attention are reoriented towards the
(17:46) self look like where we still have access to community where we still have access to information where we still have access to novelty because novelty is good learning things is good uh it's just that there's sort of a reduction down to the most base common denominator in a lot of cases and also the emergence of just hyper targeted um content per individual which is just like it's just scratching the perfect thing that you can't fight so in your mind what is a better design yeah so I think one sort of overarching principle
(18:27) is diversity so it's it's not going to be one feed it's a lot of different ways to interact with information and people have a lot of different styles too that you know some people prefer like a visual layout maybe they want like to see a graph and they kind of would explore a graph uh some people it's actually easier yeah maybe maybe some people it is actually easier to have the feed but um but having the ability to sort of shift in between different kinds of of modes of interacting with information I think that is really important to get
(18:57) you out of the sort of rut of Doom scrolling I there's a nice tweet by I think visa on Twitter do you know he's got yeah Visa I'm not sure the last name but talking about sort of this metaphor we used to have the metaphor of surfing the web it was like a very active thing like we used to like explore between different websites and follow the hyperlinks and um it was much more sort of this kind of Journey uh as opposed to this kind of very passive Doom scrolling and I think bringing back some of that energy of like letting the
(19:26) user do more stuff like so you might like have a you're on this on-ramp like the highway of you know the feed is like a highway of information but like it invites you to leave it actually and and go off and explore some rabbit hole and and like do some deep breing as well because the feed is ultimately just for Snippets of information but it would send you out to like um to off ramps to like you know here's like this enchanted forest of a lot of like deeper content and you like read stuff and yeah and and and right now I think there's a lot of
(19:56) there's a lot of value that we can have in in making that kind of making those parts the deeper content interesting so I can give you a few examples of of things I'm aware of um one is uh like more synchronous interactions so for example you're on a web page and you have the ability to sort of see that your friends were on a web page or say friends friends or some community of people that you're like you know you respect or are interested in and they like have an invitation there like a standing invitation for someone
(20:24) to uh you know join like a live session like to talk about some you know if it's a blog post or scientific paper or something like that you could also you could almost imagine like a Tinder over like web pages like I saw this paper I swiped right to have like to indicate that I'd be interested in like doing a you know doing a live session over it and then someone someone else you know goes and swipes it right and then it connects you um obviously you need a little more filtering to you know get the noise out but that kind of sort of
(20:51) drawing us into deeper connection uh that's kind of the example I'm trying to bring here so you know every web page can be this kind of invitation for some kind of deeper connection with another human being and this this sort of each web page is basically like a I like to think of it as like a shelling point you know the that people use like a meeting place like we're meeting each other on these web pages and that's why I'm so interested in the idea of like stigar is these Trails we're leaving because right
(21:14) now we're sort of all navigating the information uned we like if you're on a web page you have no idea if your friends were there what they thought of it if you know anyone else was there you really don't know any of that information and by bringing that information to the surface and and letting it um actually be you know leveraged and used for for connection then it's a whole different experience because like yeah all your all your surfing all your all your actions you know all your browsing actions uh can
(21:39) sort of contribute to connecting you to people and finding you know more interesting information so yeah that's kind of one example I mean there's there's others that we can talk about but um what is sort of an ideal signal uh or set of signals that are sign Stig iic signals that we can optimize in or towards um you know like the word like uh in its semantics is is very Bland and you know there might be some people say well we're we're optimizing towards you know well-being perhaps or something like that and you could have
(22:23) self-reported well-being but even then there's sort of this time pref preference towards uh short term when you know real transformation often takes longer periods of time and there are trows that we have to go through often to get to a more stable sort of uh plane uh of well-being um and so there can be just as many problems from directly optimizing self-reported well-being so what in your mind are better signals that we can provide to an algorithm to help direct our attention in a healthier way yeah um I think there's a few
(23:07) there's two parts of this question one part is the sort of what are the signals we leaving like and you you kind of raised this question about like is light signal enough is that or is that too sort of reducing too much into you know one sort of uh one dimension so that's like one aspect what are the signals relieving and the second aspect is how are we leveraging those signals with AI or other algorithms to to you know guide us through information so those are the two parts I would say and then I think for each part there's you know we can
(23:34) talk about it but for example for the like part yeah I think totally I mean there's a lot of room here for Innovations and sort of expanding the range of of expression right like just like Facebook they decided instead of a like now you can do like five emojis over a post right so they give you more more latitude for expression but again this is not anything that they designed for like improving your epistemic environment they just you know they they figured it's going to help keep people on the site so I think actually recently
(24:01) someone introduced me to an idea from the less wrong I think they have something called epistemic statuses it's like it's emojis basically related to like epistemic status so you like this was insightful this was too complicated uh this made me laugh or whatever all these different things so it's basically yeah inviting people to sort of Express their their more full range of emotion and with the knowledge that that data is actually going to be used later and and not in ways that I think trying to get people more into the loop of like
(24:31) how this data is actually being used like I don't want it to be like today where I don't really have such a good idea of how the like is used when I make the like on something I don't really understand how it's contributing to the sort of algorithm I don't have any way to influence it but you know imagine like you could you're kind of partner with you know you're part of the you're creating data like so you're saying this made me laugh this made me cry this was insightful and later then you have also the ability to kind of configure the
(24:55) algorithm and tell it you know whatever I'm in the mood for something that make me think right now or I'm in the mood for something that make me laugh and just by that ination it kind of shifts you in this mindset of like okay I have to set an intention now I'm not just like Doom scrolling but I'm actually like I have to be more active I have to think what do I want to see right now and that already kind of puts you in the more sensemaker space and less in the user space well there's an interesting sort of transformation um from I think models
(25:23) built on discret values to something more less quantitative and more qualitative um from the emergence of these large language models in theory you could write a textual description of what you expect to be filtered out of your stream and that can be applied at every single step but even with those large language models if you dive into the reinforcement learning with human feedback the signal that they're optimizing towards is something like quality or Rel relevance and there's a cont there's a context in which the
(26:03) people are evaluating what is quality or or or relevant and part of it is that they're being paid to do that um and they have sort of the larger motivation so they might not be inherently honest and the second part of that is really I think the time preference I think our our tension spans are getting shorter and shorter almost sort of flowing into this sort of Singularity and if we don't find a way to stretch that attention out further uh we're almost like it's almost like we're losing our agency to self-direct um and
(26:43) so oh goad do you have a thought in your mind I'd love to hear it uh no yeah yeah you finished that I'll just give you an example of an app that that is kind of trying to lengthen the attention span but but we can get back to it what what app is that um this is an app that some colleagues of mine are working on it's called gather and the basic idea is just to create sort of a campfire around some social media or not around social media around a a media object like it could be a blog post or you know any any any
(27:14) media on the internet um and what does that mean it just sort of invitation to gather around some kind of uh you know some some kind of media object and then have a discussion about it that is sort of structured they have sort of a way for it's basically I mean a video chat room that has a little more some affordances for like people can Mark sort of uh moments in the chat that were insightful and there's some like structured prompts to get people talking and there's an AI that helps like summarize what happened
(27:42) in the end so you're got to get this kind of summary artifact that that you can go out with but but the whole thing I mean is just kind of I see it as an invitation to lengthen your attention span these things are actually very pleasurable like I you know you spend an hour talking with people and if you curate the right set of people I mean it's really interesting and you're going to get a lot of insight and and it and it's not like yeah there's so many boring you know Zoom calls we've all been on but if you do it right then it's
(28:06) actually really interesting and and yeah that's kind of just a practice of widening our attention I totally agree with you I think that that like super important and and there's just like so many ways that we need to explore to to start doing that because you know every person has their own preference and and there'll be a lot of again diversity here there is like a also a disconnect between how we use the platforms and what we ECT out of their behavior the curation of any algorithm takes a lot of work and often
(28:38) is like on Facebook it'll be like going down a long list of checkboxes uh for categories specifically and I don't find it to be a particularly useful approach um I've been posting uh in the last few days um um quite a bit about forethought which is part of this cogniz framework and it's sort of a pro protocol for sense making and it's built around the concept of belief staking um where individually staked claims or beliefs are tied to long-term um reputation or amplification of your voice um meaning that there would be a distinction online between a
(29:37) passive post which is that you don't particularly expect this to affect you in any way and a staked post and a staked post you know will be tied to your reputation you know will affect particularly in the context of what you're saying in the future whether you say that again for example if you're um a nutritional blogger and you tend to post all of these things which later many scientific studies come out and contradict that if you're continuing to try and post in that context I don't think it makes a lot of
(30:18) sense for the algorithms to continue to amplify that content just because a lot of people at that moment liked it and so there has to be a longer term time preference in the algorithms themselves Beyond just what people are doing or thinking right now even in a lot of the conflict that's going on right now the primary thing that I notice is just even waiting 12 hours there will be better information and by that time it's already spread through the internet and there's no real mechanism to go back to each of to the people who might have
(30:58) received that misinformation and grab their attention again and be like by the way that thing that you just posted to everything you should be a little you should go back and check or correct that um because it often it's not g to always be 12 hours you know sometimes it's going to take three months sometimes it's going to take three years sometimes it's going to take a decade before we know the error of where we placed our attention and sometimes myths about various truths get so baked into the collective Zeitgeist that it takes just a really
(31:40) really really long time to break through that for like example the the dopamine thing which is where people think that you know either dopamine or serotonin are just sort of a reward drug that you just sort of get a hit of um whenever something good happens and there's like a lot of different and wives tales that sort of fit our cognition in a way that requires us to have to think less um I'm really enamored with Daniel's work with active inference and the the notion of the brain doesn't want to use a lot of energy and if you have a hierarchy of
(32:23) beliefs and some piece of information comes along that violates some base Foundation of that there would be a there would have to be a lot of rewiring to get that flow back into that like a chugging engine a chugging optimal engine that is a well like a person who is happy and and good and so it's very easy I think for whatever passive mechanism that exists in the brain to reject um something even if it's very logical even if it's it's very clearly presented because you can't just plug a piece of information into it and just like put it
(33:07) on top like a Christmas tree like there will inherently need to do some rewiring and you'll have to think over the a lot of different things you'll have to re-evaluate and restructure and that would just take too much effort and so there also is this process of sort of synaptic pruning for most people as we age um which is trying to whittle down those complexities to get sort of a baked in model that is going to serve us for until we're like 35 not until we're like 75 and you really have to actively challenge your mind with a certain type
(33:49) of information flow to keep to keep the plasticity up yeah no I mean I me yeah you raised like yeah like five different points uh that we could dive into but yeah just this last one I think of the idea of like this cognitive effort and like meeting people sort of where they at where they're at cognitively I mean you would think that this would be something that would be of like Supreme interest and importance to like funding agencies or science but it's like there's very little I feel like actual research being done here one
(34:26) of the reasons is because the data that you would need kind of you know you would need like a lot of social media data and YouTube like YouTube could be a treasure Trove if you had access to all the trajectories people are making through the platform um over time like that would be a treasure TR right but we don't have access to that so so and there's so there's probably like it's just a if it's a it's a blue ocean there's really not much information I think about like how what do those cognitive trajectories
(34:51) look like what does it look like you know the the intellectual growth of someone along with their you know emotional growth and all those things Um Zack Stein is another guy who talks about this a lot like I don't know if you're familiar with with Zach but he's got a lot of really interesting sort of research and thinking about like the future of Education he talks about these kind of trajectories through information and YouTube and other platforms um so I totally agree with you and I kind of feel like you know part of what we've
(35:18) been talking about earlier we talked about these epistemic statuses and the different kind of reactions and expanding that range so I think part of what that is should be like tailored and designed to do is is really help us sort of record our trajectories through content online and and you know if we're all if we're all like if it was you know s of a normative practice that people would kind of do their thinking out the open and we'd all be sharing this kind of data I'm sure you could do really interesting stuff um in terms of like
(35:46) tailoring people like for example you know I'm I'm at this misinformation story um but I'm coming at it from I don't know very whatever uh uh you know right-wing perspective so the the next content that you know I maybe would want to see is not the same as someone who is coming at the story from you know the other side of the political Spectrum so being able to have those kind of trajectories based on other people I think would you know allow you to to address some of these issues um and I wanted to say something
(36:14) about the earlier part that you talked about the um this idea of kind of you called it sort of the staking of reputation mhm and I think this is also another really interesting topic that kind of ties into um another this is kind of more my recent research related to scientific publishing and uh a pra like there's a practice now in in scientific publishing called Nano publishing and the idea is to instead of producing you know like a paper like a large PDF as your research output you can actually start thinking
(36:49) of much smaller units of research basically you know a tweet size assertion can also be a unit of scientific knowledge you know that's call Nano Publications and you can think of these Nano Publications in a sense they're almost like it's just like a tweet but it has a little more structure and it also comes with like the the staking of reputation that that science brings with it so science has already kind of been exploring this way of like you know epistemic communication sort of right it's already been doing that kind
(37:15) of thinking about like there is some issue of reputation baked into science there's an idea of retraction so you mentioned how you know when something becomes disproved then there's methods of sort of R rolling back that research so that's the idea of the retraction in science M so you know when you talked about the nutritionist that maybe published stuff and then that turned out to be false and how would people get notified so science kind of has a mechanism for doing it again it's like way way too old school and it needs to
(37:41) be you know revamped for like the fast Bas of current social media but you could almost imagine that you know if we were sort of adopting some of these practices you might get like let's say you know you took some uh you tagged some piece of misinformation you said this was really insightful for me and that would kind of automatically kind of trigger this kind of watch for retractions and if there was a retraction on that piece then you kind of get a notification and it wouldn't matter if it's you know a day after or a month
(38:05) later or a year later um and I totally agree with you that this is the kind of infrastructure we need to be thinking about like people interact with information uh and currently on social media it's say kind of ephemeral stream you're never going to see that tweet again but in a new system it should be something that's not a funeral it's it's you know if you reacted to it very strongly then we should take that into consideration and account for later you can also link that back into sort of the trajectories of thought or trajectories
(38:32) of research where it's like if something goes out there into the world and bakes into sort of the priors of like the these number of researchers and contributes their work they're building upon this you know first principle um and then making a publish publishing about that and you can sort of see how that information creates flows of reconstructions in different people's brains because ultimately what happens is there is this there's this a wiring of the physical structures of our brains that we don't have that much
(39:14) control over um and yet it affects so much of how our attention is being guided if you want to break an addiction you know it takes an intense level of application of will to create a new structure that will guide the flow of your attention in um in a healthier way um would you say that is like information that you're lacking or sometimes it's not information sometimes it's actually like embodied practice that you need or maybe some It's a combination of both it's like how the information within yourself is stru
(39:56) structured and oriented to each other I mean two people can believe like 10 separate facts equally and have that same cognitive structure produce a very different belief about as the same article um because they're not always necessarily relating in the same way each each AR like architecture of mind uh is very very different and we don't really have a lot of good tools for how we construct that um so before we started recording you you you asking a little bit about my background and I I think I could sort of
(40:40) Zoom back to sort of the individual level and talk a little bit about my experience with attention um because when I was five you know a teacher said okay this this student is having trouble paying attention and they had read um or heard from somebody about the concept of attention deficit and so for the better part of 30 years on and off I've been on some sort of stimulant um but it's not like there is ever a lack of attention our attention is always somewhere it's whether or not we the placement of that attention
(41:24) aligns with our identity and aligns with the identity of the community so I was never for example um asked about my social deficits I was primarily asked about my attention in school whether I was able to focus on you know math and reading and a huge aspect that went by the wayside was my capacity to create friendships to create connections which is such an important um part of being a human so I Advanced um in my grades really quickly and I was able to um learn very quickly but they didn't really understand that it was not a
(42:18) cognitive failing on my behalf like they thought I couldn't read which was I I was just not interested in the things that they were presenting to me as I grew and as I I've gotten older around 25 um I had sort of well with the add I I had a manifestation of like hyperfocus right which is very helpful at times you know if I need like medication or was it independent of I I if I am interested in something if I want to do it there has always been an intense level of attention it's when I feel a a need to place attention
(43:06) in a particular direction that I just don't want to that it becomes very challenging um and so hyperfocus can be very positive it has it's very connected with flow for me and and the effortless type of cognition and then around 25 I developed OCD which is the opposite type of hyperfocus which you'll have a thought which for whatever reason makes you feel uncomfortable or doesn't agree with your identity in some reason and then you just can't stop thinking about it and it's almost gravitational and and it's it's almost
(43:47) like magnetism right where my Consciousness my awareness will be drawn physically to thoughts and then they get into an orbit and once I'm in that orbit around that thought all of the cognition happens in relationship to that that root and if I don't have the momentum to escape that trajectory it's going to start to reshape the fundamental architecture so my entire life has been in relationship to attention and when I look out to a lot of my friends now in this world who are Suddenly at in their mid-30s getting prescribed Aderall for the first
(44:39) time in their life I try to explain to them you're not you don't we probably don't have ADD it's just that the information architecture that we have has deconstructed our capacity to self-regulate and there are expectations about where you are to place your attention that don't necessarily align with your identity so I just said a lot so I'll just let you sort of say whatever you think in relation to that yeah I think it's fascinating um I I have done a little bit of reading about add uh and something that I remember really caught
(45:20) my attention was the sort of importance of environments like if you put a you know take a kid with ADD out of the classroom and let's say I don't know you know take them to some like I don't know be out in nature playing with animals or something like that it's like a completely different you know they could be completely different and uh you wouldn't see any attention promps at all maybe completely normal um so it feels almost like there's and you you said this piece about you felt the struggle to sort of you felt the struggle to sort of
(45:53) force your attention where it didn't want to be mhm and it feels like yeah that feels really like an important piece because you know you get the sense in this kind of industrial era we're in we're in this kind of uh you know culture is the the ratchet of like creating machines out of humans right it's like the the ratchet is turning we get the the you know ever since the industrial era where you know we had the manual labors on the production lines and more and more we're kind of being coupled more tightly you
(46:22) know the gig workers in the Amazon warehouses or the the drivers or right we're kind of becoming part of a machine and the Machine is kind of dictating our attention regime and that's why as kids you know they start early with that the the sort of indoctrination that as kids you're expected you know read this book write this thing answer this uh very sort of rigid kind of regimes of attention and it seems like you know some people maybe are good at that and it almost feels like this's kind probably a cultural self- selection like
(46:50) the people that are good at sort of um overriding their attention and kind of you know forcing it to do things even if they're not interested in them probably have an advantage whereas people that might have like really great focus in their Flow State but it's hard for them to um to kind of force their attention where it doesn't want to be or at a disadvantage in this kind of culture uh yeah to me it's just like I hear this and I'm like we need so many different kinds and you know such a diversity of like learning environments and not this
(47:17) kind of one monoculture that we have which seems so wrong and and uh does so much damage to people and makes you know makes people feel like they're sick or something but the environment is so and we definitely have this culture like blaming the individual as opposed to blaming the system so it ties into a lot of those things as well um yeah and I wanted I wanted to there's there's a term that I learned recently and this is a really interesting line of research uh called Attunement like your Attunement your your the the quality of being attuned to
(47:48) to to something uh um it's comes from like there's there's not a philosophy and ethics of attention I recently learned so I've been following some research around this and um this idea of like where your attention is at is kind of there's moral vices and virtues associated with that um and and you know it's an an individual level but it also is like a societal level because a society might be ignoring something really important and an individual might be ignoring things that are really important um and we can we need to start
(48:18) talking about that especially in an age of of you know information overload that we all like you know uh Dr information we have to kind of make sense of it so sort of the Attunement to information to me seems like such a it's kind of a virtue we never knew was important but nowadays like I would like yeah it's it's G to be it's becoming so important so quickly and and we can kind of see how often times I feel like that's the when I look at people these days I feel like that's one of the Beyond like left and right in politics I feel like
(48:48) there's this quality of uh Attunement like who is attuned to like what is relevant in this current moment yeah I've even the the person that wrote this piece about atunement she even claims that like the wokeness woke is basically like she likes she she characterizes it as a quality of atunement so like wokeness is a kind of Attunement to certain kinds of um you know um um moral I guess like uh transgressions or or whatever um but it's kind of a quality of attention and it's it does sound like that I mean intuitively we can kind of hear that in
(49:20) the in the word um so yeah and again these are these things are s like you know like no one has heard about any of this stuff but it feels like I mean you know we were talking about attention I I can just see that attention is gathering attention you know people are becoming more interested in all these different aspects of attention so I'm I think that's like at least hopeful but but yeah there's definitely sort of uh very little sort of broad you know from the mainstream kind of to these topics I think attention is also deeply connected
(49:52) [Music] to trust and reputation when it comes to sense making if if you have a distribution of potential voices that you could listen to on any particular topic um you know which one are you going to focus into and there's also this aspect that like attention itself is inherently extractive because you cannot really consume the whole of an information so it needs to take out the parts it needs to like like a little scalpel cut it into pieces that fit into the existing architecture bit by bit and construct it up into a compressed representation of
(50:37) you mean it's like a scarce resource is that what you're kind of saying because yeah okay and um so I think moving forward if if we want to talk about collaborative sense making where it's not just you're setting for for each individual the goal of the algorithm but you're saying we're part of a group we're part of a dow and we have these particular goals and so each individual might have their own attention goals but there can be a larger set of goals that have been defined collectively that will affect the information stream that each person
(51:17) is getting and they have the freedom and right to choose to attend to parts of it and it can adjust accordingly but I think that we need to have better systems that are akin to trust or a trust attention for various topics and if we don't want to place attention to this this uh institutional thing where for example like I don't have the credentials really really to get anything that I would write any research paper I I would write into anywhere that could get the type of attention um that it needs even if that research is far
(52:07) ahead of other researchers there are plenty of things that I have written about publicly that three or four years later I'll pull up an article and we'll say we are the first person to ever do this and I'm like well how the hell am I supposed to like what is my recourse here to say well actually my attention was here far far beyond yours I do think we have in language models uh an incredible capacity to assess at scale the semantics of what people are putting out into the world when those voices uh might move against
(52:50) the crowd at first and then as time moves forward it becomes aligned right there there's almost like these epistemic epistemic Hills where it's like this is where most of the blea is and there's people sort of like little ants going out seeking out new sources of information and over time those Hills sort of just move and sit on top of the hill that ant and that ant isn't brought to the top of the pile it's like it's covered up by everybody else and I think that's sort of the mechanism that we need to move into which is well there is the
(53:26) trajectory of the individual ants over time and there's a trajectory of the larger Collective whole and the if we are measuring the trajectory of thought of research of all of these different voices at each moment we can sort of shift or distribute attention in the algorithm relative to the Past relative to well what was this trajectory of these individual ants was this was this an ant that was just tied up in what everybody was else's saying are they always at the peak of the bell curve or are there some voices that are just
(54:01) consistently ahead of the curve they're also saying things that people don't want to hear at the moment that need to be heard and then we can take that as a signal to say this voice even though it's not getting a lot of likes right now consistently they're ahead of the curve so we should inject what they're saying into the data stream even if people aren't liking it like that was yeah metaphor um yeah I really like the sort of the the yeah the ants and the bell curve I we can kind of feel this intuitively again
(54:32) I think having the data out there would allow us to do a lot of really interesting there's just there's so much you can do actually I mean you can really turn those intuitions into something practical if you have more data about people's this kind of cognitive trajectories through information which we don't have right now but yeah um I think we kind of feel it intuitively and I want to give an example of um I think now at least this has been my experience again we're talking about sort of the right like the the war right now in Israel Gaza and in
(55:01) general I think we've seen this in all these kind of world events that oftentimes there's these people on Twitter that you follow and they will never get it into mainstream news you just won't see their voice and they could be saying things that are incredibly important I found this from my personal experience that I I look at news now but it's definitely it's only like one part and not even a very significant part of my sort of information intake like I much more trust different sources like personal people that I I follow and I get
(55:28) information that I would never be exposed to on the mainstream and it really freaks me out because I figure like okay you know this person has let's say 300 followers or 400 followers or whatever most of the world is not seeing what this person is saying I think it's incredibly important I have no way to sort of convey that and it's never going to make it into the mainstream and yeah Collective attention like you say is going to be misat tuned it's not going to be it's going to be missing very important information we don't we don't
(55:50) have any tools to to sort of account for that so yeah it's a huge gap in in our current information infrastructures and that's part of the reason you know we talked about these kind of the different stigmergic failure modes like the death spirals all these things I mean the and we see this you know mainstream Media or people that control platforms can control when Elon Musk tweets something he appears at the top of the pile so like yeah we we totally know that the the information streams that we're currently exposed to are very um prone
(56:17) to all kinds of failures so yeah so that was just kind to the point about like the the person on Twitter that tweeted something and and doesn't make it into the mainstream I think it's interesting so okay so this is kind of circling back to what you you said early in the in the beginning about sort of the attention distribution like Collective attention distribution and how like a DA might manage that kind of thing and um this is some some kind of research I've been working on um in common sense makers with Brad de graph so this is his idea
(56:44) he's got this idea of trust graphs and and a few people have been talking about this idea so a community's trust graph like imagine that a DA like currently how does a DA decide who it's following like what people it's following what accounts it's following but the practice of sort of having a sort of community setting the people they're following is already kind of that that's part of the collective attention governance like that is a very simple sort of aspect of it and so a dial yeah could right now we don't see this really happening but it's
(57:11) kind of my dream to see sort of a decentralized collectively controlled social media account it would have like the whole API of Twitter you could imagine is like decentralized and like voting or whatever you know collective decision mechanism you want to use so that yeah you're likes that this thing makes the retweets the the follows all these things would be sort of a collective decision and then you'd kind of have this ability to sort of share the yeah I think it would be doing I think it'd be an interesting experiment
(57:37) to I've been thinking about this for a while but but this idea of setting a trust graph I think is part of the part of the collective attention because yeah the the when you follow someone you're basically signaling you're saying my attention is on this person I want to see more information from them so it's a very simple Act of kind of uh uh of expressing this kind of attention and I think that you know I'm I'm interested in trust a lot and you know I've looked at a lot of different systems that have some sort of trust layer and usually
(58:10) it's pretty much a whole sale soort of this person or this source of information is trustworthy or not and that doesn't really track in the real world like trust like I trust my doctor for things related to healthcare but I don't trust them to you know fix my or something yeah yeah yeah like so there is sort of a contextual aspect that we can get into where if somebody says a particular thing because it can be placed within a semantic context using uh natural language processing that the algorithm can make a decision about whether or not
(58:55) to amplify that voice in that particular context and that reputation doesn't need to be tied to a a dis direct trust rating of another person it can be related to just the tip their trajectory basically right and it could be inferred from that trajectory such that you can have people who are not part of an institution um who might be saying something ahead of that curve and over years maybe that institution catches up and then they start saying okay we are the first people to ever write this but you can just go back we have the whole
(59:30) record we have we have that whole record of all of those claims over time and we can find the earliest representation of an idea we can find the provenance of cognition where it's like this is the first little seed and this seed got into maybe these three minds and it mutated a little bit and then that seed transferred from from these Minds to these Minds to these minds and slowly as it was as it worked through the collective mind space um and started to grow into various different people's minds until it reaches a critical
(1:00:08) breaking point where suddenly goes viral and it says oh this new article has come out that says and proves definitively um this new piece of information we can track back over time and find non-institutional sources of information those people with only 300 sources or I mean three 300 followers and really lift up their voice but only in the context within which they have proven themselves and I do think that it's very important for that a part of that aspect to be to separate passive posting meaning I'm just around with the I need
(1:00:59) something stronger to to indicate to the algorithm I need a mechanism that says hold the up you need to pay attention to this over time and you need to continually assess relative to this post and maybe I'll discover that I'm wrong down the line but I don't have that mechanism I don't have a mechanism other than just repeating content multiple times and even then like it's based off of my basic understanding of how these algorithms work and not that I actually have access to how they really work I I don't know particularly how the Twitter
(1:01:39) algorithm distributes information I know that when I post something if it's like a skitso Buddhist non duelist post like very quickly those people the people who are associated with that are alerted to that but is is that not just a bubble of misinformation where it's only presenting the people who are already in the similar part of the semantic field how do we push beyond that and give people the tools to say okay this is this is a smellier scent marker and I'm choosing to leave this smellier scent marker and you should smell this
(1:02:17) smell because it's very important right now and it won't just get covered up it won't just get buried in like the millions of grains of sand of other posts that are there and it can actually bubble to the pot bubble to the top of of the of I guess the collective mind space when it is actually necessary yeah I I mentioned earlier the fourthought dialectic which is part of the cognizes tool set and each sort of claim that you you would stake has two Spectrum along which they're ass associated with vote which is I guess you could call it
(1:02:59) Varity and veilance and it's Varity is like the spectrum of certainty with a full confidence that a piece of information is false on one end and the center just complete uncertainty and on the other end a full confidence that the piece of information is true and then there's veence which is uh is it is it good or bad or neutral with the same type typ of spectrum within it and the notion being that we need to balance between hearing things that are true that we don't want to hear that may not be good um and we
(1:03:40) do want to minimalize sort of like the misinformation that we get even if it makes us feel good over time so there's sort of a a balance in that Vector of which direction we want to go in society and I think markets for example really lean into the veillance of want short-term want does this make me feel good and not so much on is this true um or and in particular not we don't have a society that leads us into those degrees or leads us into uncertainty um the last thing that I want to bring up is maybe not directly semantically
(1:04:20) related but it's just bubbling around in my head which is the degrees of freedom that we have in our attention um because I I kind of think that our cognition is sort of in this semantic field and I know from my experience that I can run from thoughts a little bit but they're still somewhere out there and eventually they'll come or they'll bubble back up and they'll have to to deal with whatever that thought is trying to tell me until I let it sit in the center of my attention for a while and just let it I I call it like the thought needs to
(1:04:56) feel heard uh or it won't stop even if it's like a crazy thought I just have to sit with it and be like okay and so there is like I would say that we have at least the the ability to move away from a thought so there's sort of a directionality to the lens of attention in our mind but the very underrated other aspect of attention is the widening or the closing of the lens and I really feel like that's what's lacking in a lot of the discourse right now that a lot of what people are saying if they just zoomed out and took a broader picture and then
(1:05:41) sort of zoomed back in again it would alleviate a lot of the emotional based posting that the algorithm seem seems to love there there's a tight relationship with our personal relationship with attention and how we are able to modulate it and and the algorithmic aspect of it so we both need to be able to work on the larger algorithmic aspect of our attention modulation and gain the knowledge about how we can strengthen the muscle that allows us to direct our attention Within our own minds there's also this big problem that
(1:06:30) like I think when I look at elon's behavior and how he's sort of re constructing Twitter is that he Maps it very attention very closely to physical energy and so a lot of the design decisions that he's making are simply from creating an energetic bubble where it feeds back into itself like some of the first things that he did was like take away links outside like you they tried to penalize people having a link to other Platforms in the top of the thing and now yeah yeah because they want to keep you inside and when you only having such
(1:07:09) a short type of content you're going to keep that attention smaller but you want to have circumstances where it can sort of widen and open to consume in more detail um information and maybe that that that's going to come in the future maybe he's 37 dimensional chess master and all it's all going to work out in the end but and there has to be a sort of energy that comes back to get them the profit that allows the platform to continue existing and allows the um you know maintenance of the site and so they're always sort of intrinsically connected
(1:07:54) and I I sense I sense that we're at a sort of Point similar to you know when America came into being like Declaration of Independence and the constitution of there is there is evolving the replacement or the a new fundamental type of system that was once what governance was doing except for now it's happening every second and we've built it based off of corporate incentives and there needs to be something like a big moment or a big document or a big a sense that it is fundamentally of a greater weight and importance to our society
(1:08:48) than just what it is which is fun little social networks and people need to understand that these system are actually guiding democracy more than democracy is guiding them anymore so I think that's a pretty good uh place to put a pin in it especially if we lost 17 minutes of this um but this was a really good [Music] conversation [Music] no don't