Metagov Seminar Compressed To 0 The Silent Strings Of Proof Of Personhood 04172024
Speaker 1: My end.
Top Keywords
- misha 0.010
- pools 0.009
- puppeteering 0.007
- idina 0.007
- accounts 0.007
- bots 0.006
- identity 0.006
- personhood 0.006
- proof 0.006
- humans 0.005
- systems 0.005
- paper 0.005
Transcript
My end.
Welcome, everyone. Today is the Medigob seminar on Wednesday, April 17. I'm excited, to, present Puja Oliver and pass things off to her to discuss, a recent paper, that was published on, proof of personhood entitled The Silent Strings of Proof of Personhood, coauthored by herself, Mikhail Nikulin and Paola Berman, and I appreciate all of the authors joining us today. So, yeah, please, Puja, I will pass it off to you and, and, yeah, please take it away.
Thanks, Eugene. Thanks for having me and, allowing us to present this paper. I'm joined by Misha, who is the cofounder of Idina, which is the protocol we studied as well as Paula Berman, who is a c o o, radical exchange. I think some of you know some of us already. I this paper, I'll start just give a little quick introduction. It was very fortuitous how this happened. There were a lot of claims I had and intuitions I had about proof of personhood and criticisms I made about two years ago. And about a year ago, Misha introduced me to excuse me. Paula introduced me to Misha, and so this is really interesting cofounder. And he is really willing to look under the hood of his protocol, and there's a bunch of really interesting experiments he's done. And we started digging, and we came up with this first empirical paper. So, you know, it's very rare that someone criticizes their own technology or even looks under the hood. So props to Misha for doing that. But I will stop here and hand over the mic to Paula to let her start the presentation, and we'll talk for about twenty minutes followed by a q and a and just run through the basic findings of our study. And, I'm not sure if anyone's had a chance to read the paper, but this kind of is a is a high level recap. Paula, do you wanna take it over?
Yes. And, Misha, could you share the slides? And apologies everyone for having my camera off. My Internet's a little spotty, but I hope you can hear me well. Maybe see Noah with a thumbs up. Okay.
Eugene, would you would you mind if I share my screen?
No. That would that would be great. Thanks.
Yeah. I cannot cannot do that.
You should be able to now.
Okay. Let's try it. Okay. That's the screen.
Okay. So our start
Sorry. Sorry to interrupt. You should you wanna okay. There you go. Yeah. Thank you.
Great. Thanks. Okay. So our starting point here, Misha, if you can go to the next slide, is generative AI. As we're contemplating this technology, a number of questions are being raised. Is it going to replace our jobs? Is UBI a compelling possible path in a scenario where many jobs are automated away? How should AI models be governed, and can we trust the information that we encounter online? Is it real or not? And so in essence, these are questions about democratic governance. They're questions about economic distribution, and there are questions about authenticity of information, which is essential both to well functioning markets and well functioning political systems like democracies. And so, Misha, if you can go to the next slide. These these issues have largely motivated the development of proof of personhood protocols, and we might be referring them to POP, to them as POP moving forward. And these, as a potential digital infrastructure that could help address some of these open questions. And the goal of these POP protocols is simple. It's to represent each unique human with a corresponding unique digital identity one on one. So this means that on the one hand, these protocols need to be able to distinguish between humans and AIs. AIs cannot be
able to obtain a personhood credential, and
they also need to be able to ensure that the same person can't obtain more than one credential. So these credentials have to be unique. And then if you have a registry like that with unique human credentials, this could pave the way for democratic processes, even AI governance and the distribution of UBI. So today, as Puja was saying, we want to offer a critique of this model, which is based off of this the experience of one of the earliest proof of personhood protocols. It's called Idina, and it's based off of this paper that we released a few weeks ago, which is linked in the in the Slack. And so why did we choose Idina? This is because Idina was the first proof of personhood blockchain with an endogenous network of validators with a one person, one vote consensus algorithm. So we had a democratic governance model to it, and they succeeded in filtering bots and validating humans. And they were able to stop a problem of account trading where participants validate their accounts and sell them on a black market, which is a very common problem in the proof of personhood space, and it's plaguing protocols, that are much larger like Worldcoin today. And they did that using very interesting novel mechanisms. So now I'm going to hand it over to, Misha, who's the founder of Idina, and he will walk us through how Idina was able to achieve that.
Thanks a lot.
Oops.
Sorry. So yeah. So first of all, I I'd like to emphasize that Adena is not just an identity solution, but it's a blockchain based on proof of personal protocols. And Adena aims to to provide an alternative to proof of stake systems and, like, address challenges of centralization of token voting systems. And, specifically, we all know that capital is always heavy tail, correctly distributed. And in proof of stake systems, large staking pools have inevitable economy of scale. So the sole miners at risk of vanishing, giving way to oligopoly. Also, on the technological side, Adena aimed to solve the blockchain's scalability problem by building decentralized network of, like, of validators with excessive number number of nodes, facilitating sharding and parallel execution of transactions. In such network, every node is intended to be aggregated by a unique human with a unique account.
Yeah.
So can you see the next slide? Because it sounds like
The verifying human slide is what we're seeing.
Yeah. Can you let let me let me
just mhmm. You want me to present the slides, Misha, and then you can talk through them?
Something like hold on a second. I'll try to fix it. K. Okay.
K. Now it's good.
So yeah. So proof of present protocols establish uniqueness in different ways And unlike of wall coin and other protocols, which employ biometric data, Adena relies on online simultaneous cognitive tests, like, it's called validation ceremonies, which happen every few weeks synchronously for all participants. So the uniqueness is based on the idea that you cannot be in multiple places at the same time. And by solving the flip tests, you can prove that you are not bot. And the flip test is is test based on the user generated story, which makes sense in one configuration, but it meaningless in Nava. So for example, you can see the less left sequence tells a coherent story about bee sting, whereas causality on the right side makes no sense. So what happened? I did not launch in thousand nineteen, and accounts steadily increased. But as the network grew, the on chain data began to show some kind of suspicious patterns, patterns of transactions sending rewards from many accounts, from multiple accounts to the same wallets on the same time. So this coordinated one way transfers, the same time to the same wallet implied automation, which would require third party access to the participants' private keys. So either these participants had unwittingly seeded their keys to third party or they just simply never had them. By December 2020, one of the users contacted us and admitted to running a a a a human human form, like a pool of humans. So in fact, there was a high information operators, pay low information participants like puppets to prodigally perform validation ceremonies and verify their uniqueness in exchange for controlling their accounts and their private keys.
So You need to switch the slides?
No. Thank you. Can can you see these slides with a Yeah.
We can see them. We can see them.
Alright. Yeah. Yeah. Yeah. So so that was not a kind of traditional Siebel attack where malicious act actor, would fake multiple Siebel accounts typically using some algorithmic bots. But instead, as the protocol have filtered bots and success successfully authenticated flesh and blood humans, the puppeteering was de facto sibil attack of humans acting like programmable bots. So at best, puppeteers extracted some amount of delta between, the UBI paid by the protocol and the worker workers' wages. And at worst, they could just take all their loss of pocket accounts. But what's the picture so black and white? The pull operator act also claimed that they they have some awakened puppies that were coming back. So they were offering a service to some consensual participants and being remunerated. Earning rewards on IDNA presented number of muscles. Like, you need to run nodes. You need to exchange the IDNA coin to the local currency. And the poll operators could coordinate this task better than participants on their own. In short, they were cooperating and simply providing a service to some of the participants. So really in in reality, there are, like, two extreme possibilities and spectrum in between. So the same transaction pattern with one way transfers at the same time, at the same wallet could indicate, like, both procuring or voluntary operation. And both extremes differ not in how it looks like on chain, but in the distribution of information and control off chain. In case of web tiering, operators have more control and more information than know how the protocol works. They control private keys, and they capitalize their asymmetry of information and control in exchange for for minimum wage payments while maximally extracting of their own advantage. In contrast, corporation has great asymmetry of information and control, so participants may know how the protocol works and they control their private keys. But nonetheless, they find it mutually beneficial to pool resources and delegate control, including their private keys, total operator for greater rewards with economy of scale, but holding these operators accountable off chain. So despite of aiming, like, establishing a a transparent network of individuals with unique accounts, the protocol collapsed into hidden groups and subnetworks. These groups, whether autocratic or democratic, were competing for control over fixed economic pie. In March 2021, rather than fight pools, I didn't have a community agreed to have a hard fork and bring pools out of unmodified shadows using on chain delegation. This delegation, gave pools, economic incentive. So pooled accounts can now earn mining rewards using only one node instead of running multiple nodes for every account. But what's more important, now with with delegation, pool operators could handle operational hassles and earn accounts' worth without needing to know the participants' private keys. So providing this economic incentives, delegation succeeded in making pools transparent, and the green area shows the network pre delegation when accounts could not be differentiated even though they were controlled by the same proliferator. But after delegation, the pool the protocol could now measure the size of the pools. So as the network grew, the growth was notably among large pools, which are red and pink. And while solar counts, which are blue, outlined, considering constituting smaller and smaller proportion of the network up until the peak of the perpetuating crisis in May 2022. The percentage of the solo accounts in blue shrunk from 62% to 27%, the network, and large pools had reverse trend, ballooning from 22% to 61% of the network accounts. And large pools also captured larger share of rewards from fixed economic pie, dumping, Adena coin price, and squeezing rewards from small accounts. As the number of single node pool increased, the node as a percentage of the network in turn significantly dropped, leading to loss in throughput and security. So the purpose of delegation was to make coordinated pools visible, but also, like, to enable pool operators to handle operational hassles and for participants without needing to know their private keys. But oddly, large pools continue to show signs of third party key access and automated simultaneous sequential transactions. And looking at these factors, we examined 31 top pools that had ever been delegated to more than 100 accounts in the prod IDNA protocol history, and all these 31 pools showed signs of third party key access. And moreover, some pools have financial ties. And if we treat them as the same entity or subnetworks, then these 31 pools were, in fact, only 23 entities. So the procurement crisis reached its peak in May 2022. There were only 23 entities representing less than 1% of distant entities controlling at least 40% of accounts and almost half of the protocol rewards. And this prompted us to change the model to so called sublinearal identity staking, which is a combination of proof of stake and proof of personhood. And this will be the topic of our next paper titled, between zero and one.
Okay. So to many of you here that are sensitive, to both crypto and governance, this damning statistic which Misha just showed about third party key access, that would, to many people, is just prima facie evidence or a sufficient condition of puppeteering, and the common refrain here is not your keys and not your coins. Right? But on the other hand, to others, if you're sensitive to governance and delegation, third party key access could theoretically be this voluntary custody as a service choice by a higher information participant, presuming they could hold these pool operators accountable off change off chain. So addressing this argument and this higher standard, we argue that there still was puppeteering because there was this improbable lack of marketing and disputes that otherwise would accompany off chain accountable custody relations. This silence, and also the title of the paper, the silent strings of proof proof of personhood, this silence was this absence of marketing, this absence of formal legal disputes and informal customer complaints. And this was more indicative of puppeteering than an accountable principal agent relationship. And then there were other corroborating factors. The known jurisdictions of the three largest pools were weak in rule of law. They're in Russia, Egypt, and Indonesia. The solo accounts and family pools with strong social ties were flatlining or declining while large pools with weak social ties were blooming, as Misha showed in his charts. Also, there were conversations with the top human pool operators, which was showed earlier, controlling at least 15% of the accounts confirming that they pay participants to perform validation ceremonies. And then finally, the top three networks controlled almost a fifth of the network accounts, and they had meteoric rise and falls, which is also consistent with puppeteering. So based on this totality of facts, we felt fairly comfortable concluding that this intended model of one person, one vote, one reward had collapsed into mostly puppeteered subnetworks. And yet this analysis was very cursory because it excluded 84 large pools holding between 15 and a 100 counts. We only analyzed pools with more than a 100 counts, and it also excluded four eleven family pools with less than 15 accounts. And in total, this was actually 95% of the protocol's pools, which we did not include in the analysis. So the statistics could really only get worse. They could not get better. So I wanna quickly talk about the takeaways from this experiment, and I think this is, you know, in the discussion section of the paper and and really rich and interesting. The first one is that simply by giving human economics incentives to periodically differentiate themselves from bots, even as low as $2 or $14 every few weeks, the protocol gave more informed, more resourceful humans financial incentives to control and puppeteer less informed humans like bots. The second takeaway is when participants traded their time for a paycheck for more, from more resourceful humans, they transformed the system that was intended to be one person, one vote to one token, one vote where plutocrats and puppeteers gained outsized influence. Takeaway three, these economies of scale for the more resourceful to exploit is common to all global protocols. Other protocols validate humans in different ways, for example, biometric scans. But just because participants don't have to run a node or do periodic cognitive tests, that doesn't mean that they don't have hassles. That just means they have a different set of hassles, which intermediaries and operates operators will always be readily happily happy to step in and perform, which takes us to takeaway number four. These exploits might not always manifest as puppeteering, but can also surface as account trading, which Paula mentioned earlier. IDINA stymied account trading through novel novel mechanisms like sublinear identity staking, which are some some of you are familiar with. Macy is is kind of variation of that. But other protocols, for example, like Worldcoin, they have documented the cases of account trading, and they haven't solved that problem. And account trading and protocols should not be treated as evidence of, like, an advanced protocol. To the contrary, account trading actually might signal a lack of them and may be a precursor to puppeteering. Takeaway five, the challenge of filtering humans from bots controlling accounts really can't be separated from the information challenge. So group of personhood seeks to differentiate humans from bots, not their biases. But as Idina demonstrated, when given the incentive to differentiate themselves from bots, humans also have incentives to align, control, and puppeteer other humans like bots to amplify their biases. And it's these biases or really the problem of faction going back to Madison that we care about in checking systems of democratic governance. And our contention is that this information challenge, which we call de facto civil resistance, and contrast that with de jure civil resistance, This information challenge will become more acute as humans integrate biologically with neural interfaces, and this distinction between filtering humans from bots and humans acting like bots will actually blur more and if not collapse, revealing this more foundational challenge than merely establishing biological uniqueness, and that's establishing the informational uniqueness of participants or the extent to which they cluster with the same interests and biases. And going to the next slide, this is not a technical problem, but a social one. De facto sybils, or puppets, or humans acting like bots, are the natural objects of colluders, or puppeteers. And by extension, de facto sybil resistance is a mutually implicated or a mere challenge to collusion resistance, and neither can be solved independently, but both must be tackled simultaneously. And finally, the last takeaway is that thwarting on chain vote buying with certain mechanisms like receipt freeness and advancements and proofs of complete knowledge. This doesn't solve for off chain vote buying in Demetrius space, but it actually may encourage it as a low cost alternative. So Idina undermined account trading with identity staking, so the next best alternative for vote buying actually became off chain puppeteering or simply buying participants' time. Similarly, when on chain vote buying in trusted execution environments becomes costly, we can expect the resourceful to simply move off chain. And I mentioned this because there's a lot of research, particularly out of IC three around receipt freeness, and this actually opens up an alternative attack vector, which is just simply buying participants' time. So the IDINA experiment underscores scores a lot of things. I think mainly the importance of thinking in terms of incentive systems and mostly acknowledging the ties that arise from talking and trading, right, the silent strings. And in future work, Misha and I will examine the relationship between accounts and stake or incenting accounts and stake, and this is what he alluded to in sublinear identity staking, which is a combination of proof of work and proof of stake. And in future work beyond that, I will explore social identity. So thank you for listening. We'll take questions.
Great. Thank you so much for presenting and for the three of you for walking us through that presentation. So just as a reminder to folks in terms of the flow here, so there have already been a couple comments in chat, so we'll just give a chance to to let those folks voice any additional questions or opinions from that. But, please, if you do have a question that you wanna bring up, either raise your virtual hand in the reaction section or just type question in chat, and I'll call on you in order. But, yeah, initially, I saw, Philip, you had dropped a comment kind of early on. I just wanted to see, you know, the the idea around not calling it an identity solution for Adena. Was there anything else you wanted to bring up there, or was that more of a pure comment? I guess
if I was gonna add to that, I would just say that some academics have concluded that the phrase identity means so many different things to so many different people in so many different contexts that it's almost lost its communication power. So I hesitate to even write the comment with the word identity in it because I didn't invest good time to define what I mean by it. In fact, I have wasted so much time in conversation talking past each other because it turned out after a couple of hours that my interlocutor had a totally different context in mind to the one I was working with. And I just maybe would find that as a critical concern if anyone on the call here is relatively new to questions of identity. I'm just trying to learn the comment I made there. We're talking about personhood here, which means if there's an identifier in a system, we know it uniquely represents one physical corporeal human entity. It doesn't constitute identity as I define it in my work because it doesn't allow any other party to recognize or make sense of a particular individual. It just says this identifier has been distinguished as human from not human as it as the current definition goes in terms of tissue and de facto. That's obviously an extension to it. But yeah. So I just wanted to throw that in because it might help the conversation.
Yeah. I I completely agree. The starting point in in the first page, we recognize we we kinda qualify what we mean by proof of personhood that it's limited even even beyond identity. Like, what is personhood, like, philosophically? Like, these are really rich questions. And so we limited it to the context of biologically unique, you know, blockchain protocols because we are seeing a rise in arguments for them, like Worldcoin, as if they can sort of magically solve these problems of, you know, democratically governing AI or facilitating a UBI or, you know, offering information provenance without really examining the incentives that these systems create. So this was a critique, but to start the critique, we have to contextualize with what's, you know, a paradigm that is being essentially, you know, pushed out there and created.
I forgot to add, Richard. I really love the paper. I forgot I I should have led with that. I really loved reading it, and I respect the idea and the team on the work and the explorations that they've conducted. It was it's fascinating work, and I I just loved it. So thank you.
Hey. Thank you, Seth. I know you had a question. I got some engagement already in chat, but, yeah, I wanted to to see if what else that you might wanna discuss with the group on audio.
You know,
I might have just been preaching to the choir. Thank you so much, Puja and Paola and Mikhail. Really wonderful collaboration. Maybe I could pass my question along to looks like Liz just posted a question. Can I can I hand off?
Yeah. Of course. Yeah. I know Liz asked for clarification with what was Pucci's statement towards the end around takeaway five or six around boundary blurring. Liz, I don't know if there was any more you wanted to add there. Otherwise, we'll yeah, just wait for the clarification there.
Oh, so okay. I mean, what I wanted to say so one of the terms that we introduced in the paper, we introduced this distinction between de facto and de jure civil resistance. Like, de jure civil resistance is, like, just biologically unique humans, and we're trying to differentiate them from bots. Like, let's just separate the humans from the bots, and that's a huge step in, you know you know, progress. And a lot of people and and that that that is a valuable step. But when you put economic incentives around this, you you and and you and it's associated with power, whether it's economic power or voting power, you also create this incentive to for other people to control other humans like bots, and this is a de facto civil resistance or or people acting like bots. Right? So it's a sort of deep irony when you, give incentives for people to differentiate themselves from bots. You also give them incentives to control other people like bots. And, you know, we see this in, like, many contexts. I mean, this is kind of like the perennial, debates and campaign finance law and so on, right, and the influence of money money in corp in contributions in voting and attention and advertising. So it's just, you know, the same sort of principles apply to to this context of of civil resistance. I I don't know if I answered your question there.
Yes. She said yes. Thanks in the chat, so thank you for that clarification. Steve, I know there were some things you brought up, but also so you mentioned you're at a conference. I don't know if you're able to come off mute. Did you wanna ask any anything to the group, or should we jump to Daniel?
No. I'll just say hi. You know, I just I didn't catch enough to really make any comments, but I'm glad I was able to pop in. But I actually gotta leave now and go to my next thing. So thanks.
Yeah. You're in an identity conference, which feels very appropriate for today's conversation. That'll be I'll pass it off to Daniel then.
Thanks, Eugene. Thanks for a really interesting talk. I mean, this is a very cool project. I have
a question, and feel free
to engage with it as much or as or as little as you want. It's related to what Seth said. Basically, you know, there's this idea that there was one token, one vote, and that was bad. So we're moving to one person, one vote, and that's the goal, and that's good. But implementation is sort of undermine the whole thing. And then I feel like there's a third approach which gets some play, and I'm curious. And this is kind of, like, a a graphical based so almost reputation system where it's not token based. It's not one person, one vote based, but it's a bit more of, like, a representing soft power with a number approach. And and and and it and, you know, with the graphical approach, a lot of the the the failures and exploits with one person, one voter kind of avoided. Was that track considered and and sort of dismissed as part of the process? Was it never considered? How does the Idina team think about those kinds of approaches, both their value as reputations or identity systems and as their feasibility in terms of their ability to avoid some of these exploits?
Yeah. Thanks. That's a good question. Like, I think that it's it's quite challenging to put to represent reputation system on top of the blockchain algorithm. The back consensus algorithm actually requires some kind of a common common knowledge, you know, that is shared across all the participants so that, they could just vote for that, like, whether this participant is is eligible to vote or not. And this kind of based on the agreement of the whole all network. And rep reputation is quite challenging to establish on, across, the whole network. I think it's it's it can it can be only contextual. Right? And, and the only thing that currently, is kind of available in the web the three space is, actually tokens, like like the systems based on token voting, like proof of stake systems. And now we have, like, a proof of personhood. And the next experiment that we are currently running, like, what's IDNA is currently based on is a combination of those two, like, proof of personhood and and token voting. So it's like it's called so called identity staking with sub neural economy. But I think that the the third part of the paper will be kind of more, like, related to this question of reputation and, like, a contextual maybe Puja can can can add add something there.
Yeah. I mean, IDINA is, like, is a global protocol. And so with global protocols with one sort of verification method, there are there are just limitations with that. I do think the web of trust approach makes a lot of sense, but how you actually do that in a way that is not laborious and is easy and intuitive is is a challenge. But I think in in general, like and I think there are ways where kind of maybe Idina can incorporate you know, have aspects of of local groups kind of merge with it, kind of with this with this global protocol, and have, like, different layers to it. But that's a kind of architectural conversation that is much much deeper than than this call. But but for now, the, you know, the trend is, you know, every technological network or operator of a technological network, especially if they're, you know, backed by venture capital. Right? They wanna get the maximum number of users and, you know, the broadest reach and the broadest breadth. Right? And so they always go for, you know, global. And and so that that that alone is sort of a problematic first step when you're building systems, I think, especially reputation systems.
Yeah. And, Juan just typed out a question. I please feel free to come off mute. I know you said you're driving, so you might not be able to. But I know you just mentioned, you know, why would a technology for solving social problems be easier than a meat space pollution resistant, something along the lines of democracy, just as a clarification question there?
Yeah. Sorry. I mean, I'm sort of just agreeing violently, like, the the the the problem isn't technological. The problem is that human time is for sale cheaply in half the world. Like
And so I guess building off of the oh, please, Puja.
No. Yeah. I mean, it's it's, yeah, it's an aspect of why humans are just cons not not only is, like, are there differentials in the value of time, there's all kinds of differentials, right, in education, and information. And, you know, you one of the points that we tried to emphasize in the paper is, like, you can be, quote, unquote, an agent, right, in in the home we cannot in in the kind of classical sense in one context, but be a puppet in another. Right? So, So, like, my dad is very agentic when it comes to medicine, but he's a complete moron, right, when it comes to handling, like, private keys and what or anything related to crypto. Right? And so we can't be, you know, agentic and quote, unquote agentic in all these contexts. Right? And there there's always just differentials in in different contexts that we operate. And so when we're designing systems, one of the goals is to always kind of encourage information revelation across the most adversarial parties, so people just don't get stuck for too long in whatever kind of information rabbit hole or or spin zone that they've been stuck in. And that's kind you know, that's the work of plurality, of bridging bonuses, and correlation discounts, and so on and so so on, is getting groups that are very different to to talk to each other and achieving consensus across difference. So, you know, we avoid these situations of, you know, puppeteering and and kinda getting on out of control. So the solution is not to, you know, fight social coordination, but to actually encourage social recombination through bridging bonuses and so on. I don't know if that addresses your what what you were going for or your question.
Yeah. You said it does. Thanks. And I I also wanted to just ask the question, you know, thinking about some of, your other work, Pujan, writing about soulbound tokens, and you just mentioning the idea that, you know, at least in a lot of the current models that we're seeing, there's kind of this VC back pressure or irrespective of what is causing it. But there's this pressure for more global protocolization as opposed to focusing in kind of smaller scale solutions. I mean, do you do you and and Paula and Misha have a sense of what kind of more smaller scale experiments could look like in order to make it more feasible to architect web of trust based systems and find ways of, you know, capturing some levels of reputation that can be very localized and and slightly bridged up to, you know, let's call it regional protocol, if not a global one. Or are you just thinking about it in a different way in terms of other kind of experiments you'd love to see as these kind of things just get, you know, instantiated into the world?
Paula, do you wanna take this one?
Sure. The way that I think about this question is that people don't come together to authenticate their identities or to give each other a certain reputation so that they build that reputation for themselves online. They have to be they they need to have an authentic motivation. So some of the things that I'm excited about in this direction are data cooperatives or local community currency initiatives where people can come together, exchange information, and trade with each other for different reasons that have nothing to do with building their identity. But through doing that, they are building the kind of substrate that would serve as a foundation for a more interesting, more informationally rich identity paradigm. Does that answer your
question? Yeah. Yeah. No. That's it it's just yes. That does. And it's always interesting to kinda think about what keen to see more of both what happens in Adena and the follow follow-up in terms of writing. I see Rick had a question or, I guess, two questions. Rick, are you able to jump off mute and ask, or should I read on your behalf?
Oh, that's fine. No problem. You know, I understand the the I'm sorry. I I'll
just do a very quick here that might be interesting for some of the folks in this call, right before apologies, Rick, right before you, ask your question, which is that Radical Exchange is actually running what we're calling a plural money experiment in a couple months at this in person community, which is gathering in Northern California in, I think early June, for the full month of June. So if some of the folks are around, this, you know, is this is something that is, going to be happening and that sort of bridges with our thinking on on identity. Sorry. We please go. Okay.
Yeah. No. I was just reflecting. I mean, I I I think the issue about you have to define what identities is within this within this context, and there's gonna be variations, and the clearer you can be about better. But the association that came to mind was was personas and architects. That's not up for discussion. That's for, you know, for your consideration if it's worthwhile. But the other the other question I had was, if you were to write press releases for a tech audience, the general public, what, are the future implications of this paper with regards to the future of democracy?
Wait. There's this whole section on what democratic governance means in the age of AI. So I would you could even skip the paper and just read that. But I think what is what is power? It's information and control. And when we're built when we're trying to decentralize power and build democratic systems that harness information from different, you know, aspects, we have to think about ways in which people's information is correlated. In this case, what's a puppet? A puppet is someone whose information is controlled by somebody else, by a third party or a colluding group. And so acknowledging these clusters, right, is really important, and it's actually a new way to kind of think about democratic governance because democratic governance was there was a there was a an assumption in it, a few 100 years ago, which was that geographical, location correlated with a set of interests, and systems of representation were built on top of that. So your representative could kinda reconcile all the different perspectives within this geographical region and then, you know, with a bunch of other representatives could reconcile in the in the nation's larger public interest and common good. But as humans integrate more with technology, as we kind of are more blue door phones, as we are attention fracked or engaged in perpetual attention auctions, which is actually most of the time what it is, our beliefs and desires get correlated in very kind of unpredictable ways, which break that assumption of geographic geographic distribution of beliefs. And often, it leads to, because we are in this sort of polarizing attention auction, we end up in these very divided groups that are just really independent about, like, where we are living, like, locally. And, you know, people are just outraged about all kinds of things that are not really even closely relevant to them. So I think introducing this you know, acknowledging how the the role that information and control plays in, you know, our own beliefs and desires is really and and, you know, the pipes for that is really key to thinking how to renovate democracy and and and put it on a stronger technological footing going forward. So hope that makes sense. I see Josh has a question. I'm very curious what Josh's question is.
Good, folks. Great paper. Well, great paper in the sense. I'm scanning it as we're giving the presentation. I apologize I didn't get to as much as what I was hoping before. One quick thing, just a citation thing. Maybe, like, this paper definitely, like, reminded me of the the Robert Dahl classic, who governs, in which, you know, we define he defines the idea of pluralism. You know, it's the idea of, like, factional, the factual conception of democracy, right, or, let's say, democratic governance in, you know, many types of societies. And, you know, the the relationship I guess, like, one one issue maybe I had is just, like, this idea of puppeteering. This word, the usage of the word puppeteer, you know, is a very clearly sort of normative it kind of generates a kind of normative stance on, like, what's going on. Whereas you could argue that and I I think you do argue in the paper. You know, you position it, you know, later down, that it's a what is it? You know, it's it's it's there can be a version of this that is quite, you know, that is quite good. Right? You know, it's about, like, people coming together to form political factions or groups in order to sort of express power and sort of compete with each other, you know, in some sort of playing field. The the specific I guess the other question I had is just I was wondering if you guys had any thoughts or lessons learned about, let's say, the modularity or compositionality of identity sort of protocols on top of, you know, other economic or political playing fields, whether it makes you think that identity is less composable, more composable. Yeah. I'll just leave it there. Quick question.
I think Misha so I I'll just say this, and I wanna hand it over to Misha. So when Misha first presented me this, we actually had these debates about puppeteering, and is that really the right word? And and as we as you say in the in the paper, we first say, okay. This looks like puppeteering, but, hey, maybe this is actually cooperation, and actually there's a spectrum. And the point is in any and this is what I was going back to the example of, like, my dad. In any given interaction, right, there's always a spectrum of, you know, asymmetric information. And in some situations where, you know, more, quote, unquote, agentic and more puppeteered, And and the same thing with these groups. Some and even within those groups, right, there's a whole different, you know, the you might have a cooperative family pool, but, you know, there's, like, more you know, not all the members might be fully understand really what's going on or the ins and outs of the protocols and so on. So I think the point we try to, you know, introduce that nuance in the paper and the point about that this is a spectrum, But nonetheless, even given the spectrum, there was and this Misha, you should jump in here. There's just so much damning evidence that it was less cooperation and much more of the kind of nefarious, like, people just treating this as, like, a job and doing what what they're told and less less positive. But, Misha, I'll let you jump in.
Yeah. I mean, exactly what what you what you just mentioned. Yeah. So what we observed is that actually, it's a probably, it's might be a function of, like, of of the size of the pool. So the larger the pool, the more evidences of, of of hearing we observed. I we we actually didn't dive deeper into this. Like, a small smaller groups because it's quite extensive, you know, like, work, analyzing the on chain data. But in general in general, yes. I think that's it's most mostly like a package during. So I think that even what what I've what I've seen in the, like, in Discord server, like, like, following discussions of community, even like like, small family family, like, pools with, like, 10 members. They they had also kind of a picturing, you know, like, when you have, like, family members who just do some some, like, exercise, right, like, at specific time to validate their accounts. And they're they they don't know what they're doing, and they just, you know, do what they told to do. So it's kind of puppeteering, and there is a of course, there is a spectrum, but yeah.
I I'll also add this, Josh. Like, if you were to let this experiment run out, I think in the end like, say say, like, hypothetically, we all you know, there's some information prominence crisis. We need some way to authenticate biologically unique human beings, and everybody, you know, gets goes to an orb and scans their eyeball and joins Worldcoin, and they become the, like, global service provider. Eventually, like, with a you know, there would be, I think, oligopolistic market structure that emerges here. And it it would maybe once people become more familiar and they get over the hump, then they sort of understand what it's like, kinda like how, you know, you vote a couple of times. You start to understand how democracy works. And but the point is it wouldn't be this egalitarian one person, one vote, one token, right, system that we end up with. We end up actually with subnetworks and most likely oligopolistic subnetworks. Right? And so we should just be very clear about what this system is, that it is it does not end in egalitarianism. It it ends in oligopoly at best. Right? Or at yeah. At best. And and then maybe, you know, singularity at worst. And as for your other question, like, I don't even know where like, modularity, composability, those words sound, like, great, but I just don't have much context to really really answer it coherently.
Paula Do you wanna oh, sorry, please.
I just wanted to give Paula a chance to say anything.
No. I don't have much to to add. Thanks,
Yeah. I just wanna level set. I know we're running a couple minutes over the usual schedule time, so folks do have to jump. That is totally understandable. I see there are a couple questions that are that keep jumping in between Kate raising his hand, Josh asking why isn't FICO puppeteering, and Rick coming in with another question as well. So, if the three speakers have a few more moment moments, we can run over. We generally push to a Slack thread where folks can ask, any follow on questions or those who weren't able to join today can ask questions. So I'll kinda check-in with the three of you of, do folks have to jump, or would you wanna stick around for another five to ten and keep the conversation going?
I can stay.
I can stay also.
Maybe it makes sense to before we push things even further to do our tradition.
Yes. So yeah. With that, for those who do have to jump in case anyone is not able to, before I pass it off to Kate, we just like to do a small tradition of everyone. Please go off mute and just quickly thank, in the form of applause, thank our speakers. So we'll do that on three, two,
1. Woo hoo hoo.
Thank you guys so much. Yeah.
Thank you very much. And, yeah, no worries for those who do have other obligations at the moment and have to jump. But, yeah, very big thank you to Puja, Nisha, and Paola for the conversation today. But, yeah, with that, Kate, I know you wanted to to ask a question, so please jump in.
Thank you. Really good paper. We're gonna we've been devouring it as part of the work that we've been doing. But I have two questions. One is semi related to what Rick just asked, which I'll ask first. And then there's another one that's a little bit more general to the kind of the group to sort of propose not just to the presenters, but to the wider group. So the first one is similar, I guess, to Rick's, but slightly different. How does this kind of puppeteering and sort of civil attack really differ from traditional or tried and true methods of democratic violence, like through coercion, fraud, astroturfing. The ones the only things I can really see here are kind of speed and scale and perhaps safety for the attacker as somebody who doesn't have to carry AKs around and and point them at people. So I'm wondering, how does this differ as a system from that kind of attack, more meat space attack? That's the first question. And then should I ask the second one as well? Let's answer your
let's answer
your Cool. Answer first one. Yeah.
So on the coercion thing, like, a gun to your head coercion, and, I mean, this is a meat space attack too. It's just paying people to do something out out out of protocol, off chain. Right? And so it's not coercive in that sense. Right? In another sense, when you kind of, like, take a systems thinking view, it depends on, like, how you think about coercion as well. Like, I've I've had debates with a lot of people about this. And so if you take a more expansive view of coercion, then some of these situations, like, if you know, because if you don't if you don't do this, then you don't have food to feed your kid in Indonesia, for example, and that's coercive, right, then you might you might want to apply that label. But one of the things that we also emphasize, particularly in, like, a conclusion is that when you take, like, a systems thinking approach, right, also what's kind of cooperative or what's puppeteered or what's consensual or what's coercive can kinda depend on the perspective of the observer, right, and how these nested networks are and your position within them. And we don't have a god's eye view. We shouldn't have a god's eye view. That's a bad system system to design. But, you know, there's there's subtlety there. Misha, do you wanna say anything about coercion?
Well, I would I would also I I would only add that, actually, the procuring like, the nature of procuring is kind of driven by the Web three d space where you have, like, lots of projects airdropping their coins. And, also, the Adena Adena started with a a model where you had, like, one account, one reward. So it was a equal distribute distribution of newly minted tokens in the product also. This is something that actually fueling the predatory and, like, you know, creates all this, like, economic relationship relationship off chain. Right? And the coercion is, I think, that's the kind of extreme way of relationship. That that could also can kind of be driven land fueled by the by the economy of airdrops and, like, equal distribution like UBI.
Thanks for that. And then the second question would be, how do you reconcile the again, this is more general, so it's sort of opening it up to the wider group here around it's more of a general question around Web three, but has, I think, particular potential implications for the kind of work that you guys are doing around these attacks. There's this idea that sort of data collectives or other sort of cooperative movements should exist in order to manage identity systems, where people make decisions about how they're represented in these systems and then how that is then exchanged for services, UBI, things like that. But how do we reconcile this kind of the oppressive streamlining that's inherent to sort of creating identity systems at scale, where the more that you scale and the more people covered, the the the kind of fewer the the more that complexity is inherent in, like, edge cases and things like that. So you have to start to sort of really generalize large populations in order to get there. So that's one part of it. And then the second part of this is that there's a technical literacy required to participate and contribute not just to the using of the system, but also to the development of the system. And I've always, I wonder what democratic systems look like when there's kind of an abstraction a technically a technical abstraction layer over the top of that and how that navigates, sort of informed consent or informed decision making or the ability for nontechnical groups of people to to influence the protocols themselves in different directions.
So I can I can jump in here and say something? So if you read, actually, the you would find there's a lot of stuff about and I think especially in, like, the footnotes about, like, informed consent. But, like, the general idea is, like, if if if my account has many memberships to different groups and those groups are, say, like, very socially distant or even adversarial with each other, right, it's much more likely that I'm you're you're you're not going to be able to pull a quick one over me. Right? And I'm less likely to be coerced. And so, actually, it's encouraging social relationships across distance that makes us that increases the cost of influence. Right? So, like, a a puppet in Idina is somebody who's, like, very cheap to influence and control. And so the question is, well, how do you increase that cost of influence? And you do that by encouraging memberships and affiliations with people who are not really affiliated with each other. Right? Because with their if you're talking to people who aren't talking to each other, you're actually getting diverse sources of information, and you're more likely with diverse sources of information to not be as sort of easily influenced or persuaded because you're hearing multiple perspectives. And this is comes back to sort of the lessons of plurality, which is encouraging multiple perspectives, encouraging social recombination, encouraging prosocial behaviors precisely to address this problem. And the more we do that, kind of ironically, the more civil resistant we are. The less we do that, the less the harder it is to actually achieve the kind of traditional civil resistance that everybody wants. So it's an information problem and a social problem.
Oh, okay. Ask a clarifier on that?
Yeah.
Doesn't doesn't that range of of motivation get somewhat shortened or narrowed as a result of the financialization component of these kinds of identity systems?
Yeah. I'm I'm not saying that these systems are the solution. Like, the the global protocol UBI systems are do do that at all. I'm just saying if you if you're building systems in general, not not necessarily global systems, but also local, regional, nested systems eventually, what you wanna do is encourage bridging to distant groups and conversations with people who, you know, aren't necessarily talking to each other, and that's how you get this collusion resistance.
Thank you so much. Subversive
in a healthy way.
And I see we're
at the fifteen minute mark, so I wanna be respectful of people's calendars. I I do really appreciate, folks joining. I know there might have been some other, comments popping up in chat, but I think at this point, just in respect of time, I'll ask folks to to jump into the Slack to continue the discussion. And if anything, I'm happy to to kinda coordinate with with authors of this paper out of Slack if if they're yeah. They're just not in in actively hanging out in the meta gov one. So, yeah, again, thank you so much, Puja, Nisha, and Paula, for joining today. This is a really fun conversation and a great paper. So, yeah, thank you all very much, and I hope you enjoyed the conversation. And thank you to everyone who partook in the in the chats today.
Thank you. Thank you. Thanks, all. Take care.
Have a good rest of your day,