Are we then just a bunch of AI agents talking to each other within the DAO?

The entire blockchain is really what's at stake, not just Lido. So where DAOs tend to fail in my view on the organizational side is often because you have these issues of collusion.

But the truth is, like, an experiment, some fail, some succeed,

and some raise more questions than answers probably. In DAOs, we wanna be in a situation where we're able to coordinate the entire protocol

the administrative scaffolding of what an organization is, we really do need to make these things accessible. We need to make them accessible to the groups who need these systems the most. Ali,

what's the future of governance? Hi. I'm Jamila. And I'm Eugene. And this is the governance futures podcast, where we explore the past, present, and future of decentralized governance.

As an ethnographer, Ali provides insights at the intersection of people, technology, and systems. In this conversation, we talked about what makes good versus bad DAOs, explored knowledge problems,

And now here's the chat with Ellie. Hello, Ellie. We're so happy that you're joining us today for this episode. And as a warm up, I wanted to hear your thoughts on, you know, good and bad DAOs. So DAOs usually are often described as social experiments, and we often refer to DAOs when we talk about new models of coordination, decision making, value creation. And they're often even celebrated as some ultimate fix to organizational problems. They're often called ultimately decentralized, democratic, fair by design. But the truth is, like any experiment, some fail, some succeed, and some raise more questions than answers probably. And you spend so much time looking into so many different decentralized communities. And to open things up, let me ask, when you reflect on what makes a DAO good or bad, first, where do you start? And then how do we know what makes a DAO good? And just as importantly, how do we know when one is bad?

When we're analyzing DAOs, I think it's helpful to break them into two components. So there is the automated component, the part that actually differentiates them from a typical organization, which is the smart contracts or the robots, the machine functionality that enables that coordination to occur through code. And, of course, a DAO can fail on that in many different ways if it's not sufficiently audited, if the smart contracts themselves are flawed, and we've seen many exploits occur on that front. And I would say I'm not the best expert to comment on that side of DAO failures, but that it's clearly, an integral part that we need to be looking at. And in fact, that's where I started was gathering these DAO failures that were happening around the place in a dataset within Metagov called GovBase. And then at some point, they kind of became too numerous to follow. And, also, as an ethnographer and someone who is interested in observing how people and communities behave, it it also became clear to me that the other dynamic that's going on in DAOs, which doesn't necessarily always tie well into those smart contracts or have the expected outcomes, is that human coordination side, which is more like a typical organization and faces many of the issues of typical organizations. So I would define those problems within organizations generally as knowledge problems. And in organizational theory and economics and fields that have been looking at the development of organizations and in the context of institutions, the we can kind of see that organizations are like people, like an individual in many respects. They have their own knowledge. They have their own capabilities to act in the world. But it's a group of people, and it's more than the sum of the individuals. So when you're looking at an organization, you have two kinds of knowledge typically. Well, it's always broken down into knowledge for the sake of production and then knowledge for the sake of governance. And in DAOs, I think we're always kind of also guilty of separating these two things. But I think where we're at the moment is that separation is also not quite right. I think for me, the real opportunity of DAOs is the ability to use the knowledge that's required for production to fulfill the mission to get your layer one or layer two or application, your dApp, whatever it is that you're building. For that knowledge that you're using for production to also be sufficiently organized so that it has the ability to hold the governance of the organization as well. So where TAOs tend to fail in my view on the organizational side is often because you have these issues of collusion or behind the scenes. So people coordinating in ways that is not really apparent. You don't know what's going on there. And sometimes that collusion is deliberate and done for some kind of beneficial outcome such as ringing the whale to get the vote through that you need to because you don't have quorum, which leads, I I think, to the other big issue in DAOs, which is attention. And they're structured in a way that they have these expectations around decision making and voter turnout or delegates, which are very, I'll say, old fashioned, but kinda based in legacy ideas of how you do organizational governance, which don't necessarily translate very well in an completely digital online context. So while that's a bit abstract, I would say that, yes, the two big issues are probably collusion and attention, and how we should be thinking about them, is around both this design of the DAO and how do you ensure that those who have sufficient knowledge of what's occurring are also those who are able to steer the organization, the DAO, in the direction that you need it to go. To do that, you need the organization needs to have the knowledge of who has that expertise or who is willing to step up and be accountable.

That is so interesting because you've mentioned that there's distinctive features about DAOs, that there's just a different technological setup. They're decentralized. They're based on blockchains, and that what you've mentioned, it sets us up very differently from other types of organizations or online communities. And you've mentioned as if, you know, the biggest weakness is collusion, attention, and it almost feels like the biggest advantage is having this technological infrastructure. But biggest weakness is having this human element, which is by nature messy, chaotic. So it's crazy. Are we the problem? Are we DAO participants who are creating those problems or, you know, just duplicating them from other more traditional forms of organizations?

I don't think humans are the problem. I think it's a it's an issue of scale. And in a typical firm, we have both the, you know, a kind of size, typically, initially, probably small, that expands and grows. And as it expands and grows, you increase your administrative infrastructure to be able to handle that, which is the theory of the firm, you know, that the firms have all of these competitive advantages because they are able to internalize all of the the costs that would other otherwise be involved in in having to do that as separate entities. You bring that inside of your firm, and you're doing that through increasing, say, your HR departments or your managers and your organizational hierarchy, which, of course, is centralization. And that is not what we're looking at in DAOs. We wanna be in a situation where we're able to coordinate an entire protocol in a permissionless way that enables anybody to get involved. And so in some respects, it's not that people are the problem here. It's that we haven't upgraded the administrative scaffolding of what an organization is or our definitions of an organization themselves to be able to function in these environments, in this digital infrastructure, decentralized permissionless way. And I would argue that the reason why we have DAOs and need them and will increasingly need them is, in fact, that particularly with AI about to become such an important and, I suppose, scary part of society that we also need organizations that are capable of working at that scale. And we need so we need the ability to bring rules, to bring oversight, to bring this kinda old fashioned administrative infrastructure. We need to be able to bring that into our computational infrastructures. And so that is a DAO. And so for me, a DAO is not necessarily a small group of people who have a cooperative or collective or an arts project that they're working on. Although it can be that, it's also potentially a mechanism for people to be able to cooperate in a human machine assemblage that is in many respects a a new form of organization. And that is an organization that is capable of implementing rules and specifications and accountability and human insight in appropriate ways around these new technologies that are about to descend upon us in a massive way.

That sets up us very nicely. Eugene, would you walk us through what's next?

I think that you're correct in that something will be unlocked. What that something is the bit we need to be very cautious about. Yeah. So DAOs may be insufficient without AI, and I would say that they're they're insufficient in that they will struggle to hold and sustain the amount of knowledge that's required for things like onboarding Mhmm. Or things like ensuring that people are across the next issue that's coming over the horizon, or that those who have decision making abilities even have sufficient training around what the legal risk implications are of the decisions that they're making. And particularly when you're saying, oh, well, those who have decision making are based on who happened to be around at the start and have and therefore possess a lot of tokens. Right? So where you have those plutocratic governance structures in particular or where you have, for instance, people who were put into positions of being a delegate because they have the most followers on x. So you have particular issues going on that I think AI can assist DAOs to overcome and that without something like AI, they may never fully be able to get across them, except in ways where you're creating, you know, these traditional team structures of committees and hierarchies within a DAO that is essentially just a regular firm or organization operating with a bunch of smart contracts. But if you wanna get to that point where perhaps anybody can come in and participate and where the ambition of the DAO is something that can be this generalized infrastructure without necessarily having those points of failure around central teams, then I think, yes, AI could be completely transformative. Mhmm. But, of course, you also then have these dangers of, well, are we then just a bunch of AI agents talking to each other, within the DAO? How many of them are spambots or actually hostile agents? And who's keeping check on that? And I suppose, are they even sufficiently across the knowledge that organization needs? And I'm not sure if they're just going out there, if we're just basing these things on a a large language model product that a commercial product out there that is taking all the information on the web and spitting something back at you, is that really organizational knowledge? I don't think it is. Because I think one thing that I find helpful in thinking about LLMs is that and this is work I've been doing, with Jason Potts, is that they, you know, these are mathematical models that are drawing on the products of human culture and knowledge, things that we have produced that which are our own embeddings. We don't just come into this world as, you know, fully formed knowledge agents. That is developed within the human biological being through families, through communities, and that's how we develop our own knowledge systems that they're groupish by nature and through education and even organizations and rules. So we have these kind of ways of structuring knowledge. So LLMs are coming in and just running kind of mathematical equations across that history of knowledge and spitting out something that is called when prompted so that the individual is then kind of bringing its own localized knowledge system to that AI and saying, give me something that I need before then. It's just math. So what is kinda interesting, I think, with this is it's disaggregating that group formation that we have developed in the first place and giving us something back that is devoid of the group. So where we need to get to is ways of bringing that collective knowledge formation and the ability for groups of people to have some kind of control and organization over that knowledge process and to be able to specify this is what I care this is what we care about. This is the these are the rules that we work to. This is the direction that we're headed in to be able to have that kind of expressive ability to direct how those machines are operating for themselves rather than just relying on some kind of commercial product out there. So I'm not I mean, you could think people are going out and training local LLMs and doing that, and I think, yes, they are. But it's not mess that's not necessarily an efficient way of doing it. So DAOs, I believe, need to find ways to be able to create appropriate rule constrained AI's fit for their specific purposes. And we're not I don't think we're anywhere close to that yet. Definitely not. Yeah. Jamila, did you wanna jump in? Absolutely.

So you've mentioned knowledge systems and the necessity of maintaining those knowledge systems. And to me, knowledge systems present themselves as kind of like the umbrella term that includes so many different factors like values, I don't know, vision, you know, goals, etcetera, etcetera, expertise. And my question to you, what are your thoughts on governance systems that models decide to attract? For example, expert groups that come with their own knowledge to reshape or to define governance system of particular community. And what are your thoughts on this preferred way of, for example, organizing someone's governance as opposed to perhaps letting the community shape it in whichever form with the knowledge

that they have? As someone who's been on the boards of nonprofits, I actually think the processes around strategic planning and constitution review and the rest of it are really important. My concern really in DAOs is how you hold that. So you can go through that process, but then you have a whole bunch of new people come in, or the team changes hands or whatever it is. So I think that and, typically, within an organization, you would just have a timeline of review for doing those processes. Every year, you'd check over it. Every five years, you might do another major review. I think that while they're useful and important, they're not necessarily binding in the way that we currently deal with them. So you can create a constitution that says these are our values, and we expect all of our validators to do these things, for instance. Unless you have actual processes that are going to be executed and which are going to, I would say, reward and punish behaviors in that DAO context, you don't actually know if that's occurring. So

If I could rephrase here. So if it's Uh-huh. Externally, it's not authentic, it's just not gonna stick within the community. And sorry for this rude oversimplification.

No. I think no. That's right. I think it needs to come from within, and it's something that needs to be anchored in everything that you do. So, an example I would give was when I was doing ethnographic work on a project called the Validator Commons, which was trying to create governance standards for validators and delegated proof of stake blockchain systems. So, essentially, mostly cosmos blockchains, which are effectively DAOs. Right? Where validators and token holders vote on the direction of the blockchain. And they go into forums and debate things endlessly and aggressively in in the Cosmos ecosystem. So in those delegated proof of stake blockchain, you have situations where you might say, okay. The validators in these blockchains need to follow these certain rules. But then you need to be able to persuade them to do that, and it might be that we expect them to always vote. Something as simple as that. You know? You're earning rewards of our blockchain. We want you to vote. How do we get you to do that? And so the way that would typically happen is that the foundations of those blockchains would give them tokens, and if they didn't vote, they'd take those tokens away. That was the stick. But then you end up with a situation where you're essentially getting these alliances between foundations and validators, which is starting to look a bit like collusion and elite rent seekers, which is how I've described this in my written work. And what was really interesting for me throughout that process was validators themselves are very much aware of these problems and wanted to mitigate them because they do care about the longevity of the blockchains. And, of course, they do because they're making money out of it. They should care. And they're gonna care more than the average token holder, to be honest, who can just sell their coins and move on to the next thing. Whereas validators are actually invested in the capital infrastructure required to run these things. So there's all these reasons why this is not a bad system. But at the same time, you have these problems. And where they kind of arrived at, which I thought was super interesting, and I don't think it's necessarily been enacted, but is that you kinda need to go back to the code and the protocol itself. So you need to change the underlying institutional rules. So this isn't the organizational rules. This is the institution. This is the framework that enables functioning organizations to exist. Like, within the nation state, we have functional organizations because the nation state is putting all of these rules around how organizations have to behave. In blockchain, that's the underlying code of the blockchain itself. Right? And other people have written about that. But the so they kind of got back to that of, well, maybe our maybe we actually do need governators and not just validators. And maybe we need to think about, you know, how that are we requiring validators to vote on everything, or can we actually accept that some of them have liability concerns around voting? And and for instance, that it's not necessarily great if Coinbase has all the votes or whatever it might be. But you do that through those rules that rule organizations, not necessarily through telling the organization my constitution is gonna be really great, and we're gonna agree to do all these things. I think that the whole point of blockchains is that we can get to something more reliable, something hardened that and that means we don't have to fall back into some kind of strategic planning, road map, or whatever it is, or even a internal organizational constitution, which is really just often an expression of your values, but can also be a written document that says, when this happens, this is the consequence. But is it enforceable? That is what I would ask. And that's where I think I probably depart from some of the more constitutional thinkers in our research world. For me, the constitution is a blockchain, and that's what we need to be working with and through. And these are questions of security. Right? So I would take, you know, the Ethereum blockchain is so fascinating because you have this very elaborate validator ecosystem now with the ability to do staking through staking providers and and these commercial validators, but also home staking being reasonably accessible and getting more so from what they're talking about from their road map. But also now, you know, the DVT and item layer and all of these, you know, layers of means and organizational formations around how you uphold security of that blockchain. And even though their decision making is happening elsewhere in their all core those call calls, etcetera, and within organizations like the Ethereum Foundation or the client teams, whatever it might be, it's the security of the blockchain is the thing that really matters. So you start to come up with these. So that raises these questions which we've seen around, well, is Lido a threat? Because it has almost 30% of all the tokens, etcetera. So how do you then look at that blockchain and say, we want to ensure that our code prevents that? And I would be on the side in their case of, well, you need to be thinking about that validator ecosystem really carefully and the roles within it, who is doing what, who is providing security that you need, and then and then adapting for that. So this doesn't sound like a DAO issue, but it is when you look at Lido. Right? Lido is a DAO. And when you get deep into Lido, you do have, you know, however many at the moment, 30 something commercial val large validators Mhmm. Who providing that service. And then you but now you also have the ability for smaller validators to create squads with OBL and join Lido. So so you so it's kinda creating these constellations of approaches and solutions. But, yeah, but then you have this massive governance issue within Lido Yeah. Of how do you ensure that it's robust? Who has the voting power to decide that system isn't gonna fall down? Are you gonna automate, as they were discussing at one point, the selection of validators, and then what threat is coming with that? And so they're trying to come to all these very complicated responses around, say, state based holders having some kind of checks and balances over LDO holders in their governance system. Perhaps you can actually do this through the Ethereum code base itself. Maybe there are better ways. That's all I'm saying. I think in this case, when you're actually talking about this the entire blockchain is really what's at stake, not just Lido. And so you need to look at that constitution.

Yeah. Okay. So COI stands for knowledge organization infrastructure, and it is an architecture and protocol that was developed by block science. And the Medigov has been trialing and testing because Medigov is a very decentralized research network. So it's holding a lot of knowledge and has a lot of expertise, but it's not necessarily well organized all the time. And my team of ethnographers have been involved in trying to assist both MetaGuard and Block Science in understanding how knowledge organization infrastructure, their particular protocol, but more broadly, changes organizational dynamics. So what COI is specifically is the ability Michael Zagham describes it as duct tape and w d 40, which you use to stitch together the knowledge of your organization that currently lives in very dispersed different places. So some of it is in your Google Drive, and some of it is in your Slack channels, but some of it is offline even. And so how do you not only be able to create a common reference system for all that knowledge, a bit like a library catalog. It's a way to think of it. You know, you have these call numbers that you can apply to all the books. So COI is so it has something called our IDs, which do that and says these are our knowledge objects, and they live in these places. These people might be associated with them. And so you have this kind of common referencing system for that knowledge and where it lives. And you can also then have nodes within a coin network, and you can put particular functions around under these conditions or this particular organization over here is able to access this, but that other groups cannot. Mhmm. So it it can provide a standard and a referencing system to be able to create knowledge networks where you can, say, have in in our discussion here, two different DAOs might agree that they have something in common. If you can think of this, say, a cosmo local region scenario where every you know, the knowledge that's being held there, in fact, region network is also experimenting with using this technology. You know, the knowledge of this local place is specific because of its particular type of biodiversity and land management practices and indigenous knowledge systems. But it also needs to be able to coordinate with all these other systems, similar knowledge groups that are doing, similar attempts that want to be able to come together on a particular issue or share share their particular knowledge without having to duplicate it and replicate it or become a centralized organization. So you don't all need to have the same knowledge ontology to be able to coordinate. You can just share your reference identifiers, create the permissions and the access that's required to be able to coordinate. So this is the bigger vision of what we're we're trying to get to with COI. But the key thing that is important, particularly in the context of DAOs, is that you can put context around how that knowledge is used. And you can essentially govern your knowledge system itself Mhmm. And come together and say, alright. We we feel that our we're happy to share with this little district over here and the work that they're doing, but we're not gonna share with Monsanto. Also, you know, or whatever whoever it might be. So you can begin to enshrine rules and boundaries and direction and context into how it's being used. And an example that I will give because, actually, my little research team of ethnographers is the first real demonstrator project Mhmm. Of koi, which was a bit accidental. So we have a project called telescope. You see the telescope behind me. This is a little research tool that we created with Medigov that just enables either an ethnographer or people within a DAO, within a community to put a little telescope emoji on a post on in something like Discord or Slack, and that is signaling that this is an important piece of knowledge that someone has created. It's important for the discussion. And then the bot the telescope bot asks for the author's permission to include that within a in our case, qualitative data set. We began to realize that on its own is it still requires all this manual ethnographic work of adding context and linking that to other posts so you can see it as in the context of its conversation, you know, saying this relates to this particular theory out there or these other events were happening, and these people were involved, etcetera. So that's the work that the ethnographer does. Right? Mhmm. And so what we just what I started doing was using an obsidian vault to and got Luke Miller from block science and meta gov to help me create a plug in which brings those telescopes in. I add all the context. But now we've been realizing, well, my colleague, Kelsey Navin, and my other colleague, Brooke Coco, they're doing similar things. What why can't we share that back? So you're basically saying each ethnographer in this case is a little node, and we're collating, we're agreeing to share knowledge back, say then to the community that you could then use an LLM or create an AI agent that's able to see the contest as well, not just those little snippets. So what we're the way that we're describing this is it's not so much that we're putting the human in the loop of how the LLM is then treating that knowledge. We're building the loop itself Yeah. Through COI. So we're enabling these systems. And so so we're ethnographers, but a DAO could do this. Right? We we realized it's essentially a DAO tool where that ability to describe what's occurring and to add context. And then and, specifically, you could add rules around then how that is treated to these groups or that the agent can only use it in these ways and put, like, very strict directions around what it can do with it. Right? It can only perform certain actions based on that data. So where we're getting at here is that these are organizations, DAOs, are these human machine assemblages. Mhmm. And it's not just about making human oversight. What you really need are these elaborate layers of of how that knowledge is organized, who and then creating systems around that enables those who have the ability to organize it to have some kind of decision making over how it's treated. And where I would go with this is that you're actually getting to something that I would call a contribution system, which we can talk about. I just wanted to clarify.

Just to make sure I and our listeners understand as well, you've just described putting this little telescope emoji as, you know, making a bookmark almost. Or you see often people on x tag certain whatever service provider, like save that for me for later. And basically, that's what it does. It's like gather all the knowledge and then you has you've described ethnographers become those little knowledge wizards and that can gather that knowledge and share that knowledge back to the community. Did I understand that right?

You did understand that right. And I would say the knowledge wizards is actually something that all of us are doing all the time. So ethnography is just a more systematic approach to that. But in fact, we're all kind of, as you say, bookmarking things and creating notes around why something is relevant or linking something to something else that's been said in our response. Okay? So the job of an ethnographer is a kind of synthesis of knowledge and events and actions. And it's but it's also about and this is probably more active network theory. It's about tracing the pathways between things that enable certain actions to happen at all. So I would say that a knowledge organization infrastructure is enabling all of that, and it's what it's doing is it's making these processes, which I'm saying are happening all the time. It's making them machine readable. Mhmm. So that the LLMs and these Dow contracts and maybe even the constitutions of the protocols. So this is all legible with these massive computational infrastructures that we're now involved with. And it enables us to do it in this decentralized way without having some software as a service product. Because, of course, all organizations have had knowledge infrastructures and knowledge man management systems, etcetera. They're just run by Microsoft or They're siloed somebody. You know? So so they're yeah. Yeah. The and we're just conforming to them, or we have these we have some kind of business process management company come in and organize those flows for us because we don't have sufficient tools or frameworks or architectures to be able to do that across across all our systems Yeah. And then to be able to network that with other organizations. So why it matters in the context of DAOs and networks, and there's a great piece around knowledge networks written very early on by block science on this Mhmm. Is that once you can share that across, oh, like minded organizations, maybe then you're kinda getting those capabilities and efficiencies and scale that a traditional thermal hierarchy would have, but without the traditional thermal hierarchy. Yeah. So we're increasing the capacity of an entirely different kind of organization.

Well, because of our funding constraints, our current answer to that question is this is open source software. If people have the capacity to contribute to the development of the infrastructure and wanna get involved, then and are able to actually help build it out Mhmm. And test it with us and trial it, then that that door is open. I think, ultimately, though, we do and, well, block science in particular, you know, to be able to provide the help and service to another organization who wants to just implement it off the shelf. There's a whole lot of work that we would like to do, particularly around, say, UX and some of the more automated functionality of this as a cute demonstrator of video out there of a a coin network building itself using MCP. So there's there are all these possibilities that we're working on that we would love a foundation that's aligned with those visions to come in and support. But it is open source software, and we are certainly open to collaborations. I do think that the work of that I do, the more kinda ethnographic work, the contextualizing, say, the more informal discussions that go on in a DAO and bringing them into a koi and testing how, say, an AI agent might be able to work with that in a governance forum, etcetera. That really depends on the DAO. So does it have already community managers or people who are pay in paid roles to do that kind of synthesis already? Or is that something that you need trading in? And yeah. So so I think that we're we are in some ways talking about a new literacy around the around how you run your organization in a way that is fit for purpose for the kinds of automations that are coming down the line. We're still trying to answer those questions about the about how that evolves. And, unfortunately, as someone who spent, you know, a decade or two of my life looking at digital inclusion, I know that's not a trivial problem. It we we really do need to make these things accessible. We need to make them accessible to the groups who need these systems the most, which are those who don't necessarily have that the finances and formality to be able to centralize their operations. Yeah. So we're we are pushing uphill here, but we think that if we can get to the point where this is easy to implement, it will be transformative.

A contribution system is an institutional mechanism that enables a group of people who are working on a shared mission to be able to identify the contributions to that mission and then to apply whether it be through some kind of algorithm or smart contract capability, some kind of form of compute, working over those contributions, seeing those contributions to implement system for value creation, a way of determining what is valuable through contributions, what holds value in a community. And it might then extend to things like rewarding contributors for their work. So we could take some early examples of this that have occurred. The one that I did a lengthy ethnography of was source credit. It was one of the most elaborate and interesting ones, which is the page rank algorithm and plug ins from GitHub and Discord, and and then and from its its particular governance forum as well. Then applied the PageRank algorithm to that information that these plugins were receiving. And the idea was that the community that could then wait what was significant at that point in time, and the contributors would be rewarded. And people got paid pretty well through this system, and it lasted a while, couple of years. And, eventually, it wound down because they had some problems. I think we're further down the track in understanding those problems now, and we have a massive variety of these systems out there in the world. Another one which is extremely significant, of course, is Protocol Guild, which supports Ethereum developers to work on the maintenance of the blockchain. And it it has a very simple method for doing that, which is, you know, do the other members agree that you're a contributor, and are you full time or part time? And if so, you can share it in the rewards that are donated to Protocol Guild, which is a basket of assets, essentially. So this is all kinda automated in terms of how that is administered. It's a single wallet, etcetera. So you can have a a large variety. But aside from these specific examples, what is in what is interesting about a contribution system and why it relates to knowledge organization infrastructure is it's essentially a substrate that is recording not necessarily actions of who has done what. It's enabling the recording of the reception of those actions. So how were they received? And if you think about academic citations, they're a good example. So they're they're an existing contribution system. I write a paper. Maybe no one ever cites that paper. Mhmm. So it's so in this case, the object of the contribution is actually the citation. It's someone acknowledging that I wrote the paper and that it was important in the work that they did, and then you have this big chain of knowledge and how it was constructed and who contributed to this particular knowledge club around a particular topic. And it is also significant, and the academic citations example, in that that then produces reputation for people who are highly cited Mhmm. And promotions in the there is an actual reward. And there's so many problems with that system. I'm not gonna say it's perfect. It's far from perfect. But it's far from perfect because, okay, journals are a hugely problematic, exploitative industry, which other people have written about and so have I. And there's a whole lot of vested interests going on all the time in this. But, you know, DeSci, etcetera, there's work going on to try to solve that. But a contribution system, you can think of it as a kind of substrate then, a recording of who has done what, but who has built on what. And like a citation, I might not get rewarded until ten years down the track when suddenly people discover my work and I get given the Nobel Prize. Alright? Yeah. Highly unlikely. But okay. So you get Fingers crossed. You get this fingers crossed. So you so what is interesting about a contribution system is actually it carries through time, and it's almost pulling from the future. So as an institutional mechanism, it is enabling us to hold something for what we're gonna value in the future in a way that our current economic systems don't, in the way that price and markets are very much of the now. Even except the little tiny narrow things like futures markets. Right? There are these little windows that are trying to do things like that, but they're just the narrow slice of particular finance technologies and markets. So we're we're interested in how these systems and by way, I mean, my collaborator and husband, Jason Potts, and other collaborators. We're interested in how this creates future institutions as in not the institutions of the future, but institutions that are doing future all things that are enabling us to deal with problems of a scale and size across time and space that our current systems don't deal with. But they also relate to COI in that for how do you build that substrate? How do you hold the knowledge of who did what? And in DAOs, organizational memory is incredibly important, and it's one of the main problems that they face. How do you ensure that people coming in are not gonna repeat the same actions and attempts and problems of people who were there six months or a year ago. You need to be able to hold knowledge. You need to be able to attribute that knowledge. You need to know that those who are really working on something and have a deep understanding of it are the people who have the governance power and decision making power. So you can't I would say you to get to an optimal DAO, to get to a good DAO, we need that substrate. Mhmm. We need to be able to know who is able to carry something out. And you also need to be able to stay, for one of a better word, aligned. Mhmm. You need, as a group, to be able to stay on message, and you need to be able to determine what that is and to share that. And so you need to be able to stitch together your knowledge, and you need to be able to associate it with certain people. You need to be able to associate it with certain events, and you need to be able to then, give people certain rights and benefits to do things. And so I would say a contribution system can actually do all those things, whether that's, say, reputation tokens that give you governance rights, which have been experimented with. Or, you know, it can also be reward for contributions. So that citation that gets picked up in ten years' time, actually, I'm gonna get some kind of return from that. So I'm gonna make that contribution in the first place. I mean, this could be a way to change how we work in the world, and projects like SourceCredit had those ambitions. But it might actually be more about, I would argue, if you look at, say, things that we don't associate with work, like voluntary work, like the clubs and things that we do for our own fulfillments. It might be more aligned with those things than work. It could be many it could be many things. It could be our intellectual pursuit. But we the more that we can record these things, we the more that record these things, we the more that increases our capabilities to be able to know the what the direction of that DAO is, who can steer it, and also then how other machine agents, the machine actors within our DAOs can behave accordingly as well. And they might be part of the contribution system. Right? Absolutely. If your agent is also a known contributor.

Yeah. So, a project that is already grappling with the questions that you're asking is deep funding, which I think is fascinating project. It's more focused on protocol contributions and code. So how do all the dependencies so so the app that I've built, what does it depend on? What are the apps are required? What other protocols have gone in? What development of what things have gone into making my thing possible, which I think are easier to trace than some of the more human and organizational contributions I've been talking about. But the way that they've been approaching it is it still requires this form of what we call waiting. You still need to wait the importance of something in order to be able to reward it. So the idea is that, you know, a portion of whatever my app earns can go back to all those things that came before, which are important to it, to its existence. And so they're what they're doing from my understanding is you have a human jury, which makes which maps out some contributions and weights them and decides. And then you have agents compete and to do the same thing, and you end up with a system where you kinda can arrive at agents that are behaving the way you wanna behave. And so you select for those and that they then can start doing some of that work. So this is a kind of human in the loop approach. And I think that we need to we do need to be considering this because from the ethnographic work that I did, the burden of waiting was something so there was a project called govern that unfortunately didn't continue during bear market, the last one, but was you know, people would have to kind of express what if one thing was more important than another, etcetera. So there's but there's a lot of labor in that. So you also you need to be able to overcome some of that labor and weightings. I would say you also need to be able to recognize problematic bots and agents that are there and be able to punish. So contribution systems are not just about reward. They can be about punishment when something is active. So so maybe your agents need something at stake. I don't know. These are just thinking off the top of my head here. But these are the kinds of questions that once you set up a contribution system, you need to grapple with. I mean, I guess as part of that, you know, having both the recognition of positive negative contribution is important for having, you know, that punishment, so to say, especially when it's for agentic behavior.

I think if you can have agents that are acting according to specific rules, you may find that these systems are less gameable than that they have been found to be in the past. And sometimes when I say gameable, that might be someone starts, and this happened in source cred, it was a criticism within the community itself, that people started building tools for the community and its communications, say, things within Discord or something, as opposed to building the product because that was something that humans saw and could understand and easily value over the complicated work that those who are creating the code were undertaking, which for the average community member was pretty opaque and hard to not very legible. Right? So maybe some kinda agent would do a better job of contributions in that respect. In terms of what this means around the broader steering, construction, governance of systems, I like to use this word terraforming, and it's a bit odd. But where I'm kinda getting at here, if you go into more theory and the like, this notion of how do you get people to apps the way you want them to rather than explicitly forbidding them or locking them up, you get them to govern themselves. Right? So you kind of this is the the way that society essentially we're always governing ourselves all the time. But in the context of these systems, really, where we wanna get to is rather than enforcing rules on individual players all the time, we adapt the systems themselves so that people are kind of working to the needs of that particular community or group that your system itself is rewarding or punishing. But you're not kind of going after individuals. You're just kind of changing the parameters of what things are seen or valued or whatever it might be. So I think that the kind of governance so the reason why I call it terraforming is, in the terminology, this is environmentality. You know, you're controlling the environment around people as opposed to people themselves. I would put a positive spin on this, and in in terraforming is the community doing that themselves collectively in a more kind of common thing way. So this is what I think contribution systems are about, is that having legibility and sufficient feedback loops to be able to decide what the community values, what it's prioritizing, and to be able to kinda adapt your system through these kinds of tools that enable you to function more efficiently and effectively.

Thank you so much. Yes. As we head towards the close, we'd really love to hear your one worded answers to our final set of questions that we like to ask every guest. So take your time. Yeah. Just go with what your heart feels like. There's no really right or wrong answer here. So let me start with the first one. What is the hardest part of your work as ethnographer in one word?

Abundance of information. What is one problem in DAOs that AI won't be able to solve? One word. Community.

The most surprising thing you've learned in your research?

Hope. There's a lot more optimism and hope out there than we would generally assume.

That's great. And that brings us to our last signature question. Ellie, what's the future of governance?

COI. I love it.

Thank you so much. That's been really incredible and super informative too.

Hi. You're welcome. Thank you for having me. Always good to talk to friends.