8. A ‘Governance Layer’ for the Decentralised Web - Lisbon 2022

thanks very kind cool so I'm glad that Juan kind of transitioned us into uh design mode and also um has touched on kind of methodologies in a way oh check check all right I'm back wonderful um so I would like to speak with you uh on what is titled A governance layer for the decentralized web and now this is a culmination of over 12 months of research um that I was grateful to be supported by protocol labs to conduct uh but actually it's a major study of my PhD research on resilience in decentralized Technologies and so I'm a social scientist my primary method is ethnography which is uh the study of uh people and culture and community and so when I approached protocol labs and kind of told them about my research I wanted to know who are the people using ipfs what for what are their threat models and then what happens when they do and that data wasn't kind of readily available and so um we'll talk about it today so really this fits within a broader picture of funding the commons and I think this is uh grounded and actionable uh in relation to this question of how is data governed and governable for those that create it or create value from it so what we're doing here in the context of ipfs and filecoin is thinking about decentral ized data infrastructure and so I very much uh approached This research from an infrastructure stud's lens which acknowledges the uh social and Technical uh components of infrastructures as broader scalable um enabling substrates for other activities so as I mentioned uh a lot of qualitative research um ethnography occurs through um observation interviews as well as practice so I joined um filecoin in hackathons in the oceanana region um I hung out in Discord channels I observed uh uh CS disappearing from Whole nft Collections and getting recovered and and all of these kinds of um events in different contexts uh and so I wanted to point you to a working paper that's publicly available on this um the actual final version has been peer reviewed um by i e and will be published um and shout out to Janis from uh resnet Labs with protocol Labs who was one of the reviewers on that with very valuable feedback and so in thinking about uh ipfs as a decentralized technology and a component of kind of broader scale digital infrastructures what's really important and the kind of uh theoretical like Lynch pin of the way that I looked at these was this idea of Gateway moments and now Gateway moments are uh when an infrastructure is thought to be most fragile and that's when previously separate systems are linked whether they're legal political or technological for broader for broader growth and scale of a network or the ongoing process of adaptation to changes in circumstances and environment and so when thinking about uh resilience is obviously very important and when thinking about content addressing you have ipfs which is its own distinct thing uh the wonderful tool that it is which you know creates those CS so it changes uh the way we uh can locate uh content digitally from HTTP which does location addressing to ipfs which creates a CID but then it has to be linked to a storage solution so you could rely on the ipfs network uh to replicate your data but the whole idea of cryptoeconomic systems is that there's some kind of hard incentive uh and so that solution uh might be filecoin uh in a lot of ways we think that the two are are interl but there's a Gateway moment between the two so what do I mean by resilience I mentioned these social and Technical components that are inextricably linked in digital infrastructures and so the understanding of resilience I was bringing to this investigation was this idea of adaptability and transformability in response to threat or crisis but when I started speaking to people I found out that people had different definitions so reset Labs which is a technically focused uh research and development lab uh within protocol Labs is uh defines resilience on their website as uh a network that is fundamentally uncompromised by an isolated failure or a network split the system is malleable adaptable to different conditions and capable of evolving to meet new requirements over time so that's a very technically focused definition uh you know that really emphasizes the resilience of the technology itself when I spoke to Molly McKinley um the head of engineering product and research at PL uh she had a different definition again that kind of talks about this uh the broader sweeping you know vision of of linking uh content that is addressed to provable storage over time but when I spoke to end users in various contexts it emerged that resilience relates to data Integrity in line with the preferences and local context of the provider of that data and so next I'd like to talk to you a little bit about what some of those local contexts were if you're interested in the methodology of how I conducted This research or what it means to find out about um resilience in the broad sphere of decentralized Technology um I I open-source blog um a lot of my methods and kind of initial findings um as I myself investigate and I'm actually speaking on Friday afternoon in the D track um specifically on on methods for researching decentralized governance um so some kind of Dow science stuff um but just at a high level um this was one of the initial mapping exercises that I did uh which which kind of just tries to break down uh this the specific and kind of concrete elements that I'm looking for so I wanted to know you know what's the use case what's the purpose or the objective of the people using these tools or infrastructures uh what's the threat model of these Stak holders what are their vulnerabilities and and vulnerabilities emerged uh in the literature as this uh relation or notion to resilience by which to kind of observe it and then in terms of the action space how are people responding so what's happening in the social and the technical Dynamics or other dynamics of a system to actually adapt uh and and change or or fail or die and resurrect um in response to the threat models and then of course what happens and the role of an ethnographer in communicating these feedback loops so some of the uh the people that I spoke to that I've written into full case studies in that working paper include uh matters news so they're a a a crowdsourced Blog website so they have a frontend um website but what they allow is for crowdsource content to be uploaded to uh to their platform and so this is really valuable in parts of Asia where uh perhaps content might be censored or or might not be sort of persistently available just through a static front end and they had a very interesting setup they uh automatically upload content to ipfs when members um upload a Blog and then they run their own servers that are actually physically geographically distributed with uh ipfs nod so they have multiple ipfs nodes running and they are uh running an application that um that mirrors that data so it it follows specific cads and makes sure that it's replicating them in contrast in another part of the world in Latin America omn lingo is an early stage research project into uh recording uh people speaking in their local languages to actually use mozilla's common voice um database to train um an app to help people learn those languages so it's actually a really important exercise in cultural knowledge preservation but what we what I found here is that uh the the founder uh who was also a University academic was incredibly skeptical of cryptocurrencies and his kind of anecdote uh in a research interview is that I hate cryptocurrencies uh I got paid you know a grant uh around my integration and to support my project but by the time I got the cryptocurrency to my wallet it went down and I can't afford to fix the hole in the roof in my house and so that's so true you know and and that you know bridge between um the ideals of uh you know cryptoeconomic networks and the actual you know feeling someone said on a call this week like you have to know that your heart beats faster when you send a transaction kind of out of Terror and excitement I was like very true and so for omn lingo you know in in these um in this case in this local context perhaps um a better idea that was shared in this interview was uh you know what would it look like for the people actually recording their voices to maybe run an ipfs node and and you know make a local copy on a on a hard drive of that data and then maybe the local Language Institute could also operate an ipfs node and make local backups of that and and we can have this you know Community to make sure that the data is replicated but it's actually uh controlled and maintained and managed within its local context rather than abstracted to a you know the world's largest decentralized um you know storage network but you know you might be negotiating um storage deals you know with someone on the other side of the world you don't know kind of where that data actually ends up and how it's managed uh and of course I don't say that as a generalization I know you can upload very small amounts of cont content to um filecoin um and there's also amazing you know specific case um applications that are emerging like nft storage and web 3. storage but I want to bring it back to this idea of resilience in the local context and so uh one of the other examples I talk about is styling lab and Jonathan Doon is around here somewhere actually the founder of that and now that's a a broader scale experiment really um based uh at Stanford uh in partnership with the USC shower Foundation uh and they're trying to look at uh provable uh data throughout the kind of life cycle of a data pipeline from where it is captured so they're working with um kind of Frontier Hardware um development so how can I take a photo uh for example of a war crime and there's um a fascinating article I was just reading today on their website about um work they've been doing in the Ukraine uh with a time stamp a geolocation stamp Etc and then kind of map that data um you know encrypt it but have it across all like this patch work of um decentralized technologies that each kind of contribute a different thing so you know content address it on ipfs upload it to filecoin you know make a copy of it on gun Network and so on um and so again that's really about um resilient data meaning the Integrity of that data in line with the preference to be able to bring it to court as evidence potentially you know or to not have history changed when there's elections and you know tweets kind of come and go or photos are you know potentially doctored um and so on and so we see these passive and active threat models you know state level uh censorship of your content is very different to um you know omn lingo who's paying uh uh server as a service provider in Germany €35 a month for the actual storage component which is completely um insecure uh to you know to styling lab and and kind of having to to engage with legal requirements and so really at this Gateway moment of bridging content addressing and storage we see these very bespoke linkages and so these were kind of my high level findings and and takeaway from this research I think number one that the social coordination layer around how data is managed is inextricably linked to the actual technical infrastructure and so one of the first times I I spoke on this uh to a very um techn technical audience of software developers um and kind of protocol researchers uh someone you know in the audience asked the question like so you're telling me that the technology is good but the social breaks down and I was like no because they're inextricably linked you know um the the infrastructure studies literature talks about you know a tool is is only a tool when it's used and so same with um with infrastructure part of it becoming infrastructure is the active process of infrastructuring which means that people use it and actually the coordination around it is inherently um incredibly important and so uh institutional infrastructure is um these these processes that that surround that coordination to enable and constrain the operation of infrastructure uh and I also wanted to point out that uh you know while there are you know there is this kind of um this finding around these further opportunities to think about potential patterns for coordination around data governance uh that some people found that uh you know even though there's kind of limitations and the network is under development or they didn't like certain um components such as cryptocurrency um ipfs still provided a solution that was more resilient than Alternatives such as hgps uh for certain threat models which is why these people were using it in the first place and so here we transition from the kind of empirical findings to the more sort of speculative open space for further research directions and kind of what I hope to engage all of you in and so what could these decentralized institutional infrastructure models look like um are they Dows and I loved um a talk earli this morning uh which kind of spoke about you know Dows to govern the comment but what does that mean we know that a dow is not a defined thing there's no kind of conrete pattern you know it utilizes a blockchain somewhere but in what ways we don't know um and in many cases governance is relegated to voting processes when we know that there's still a whole lot of you know social processes that are um inherent in in the politicking of facilitating those governance processes uh could it look like data trusts and actually there's an incredible amount of value in looking at existing Cooperative governance models um that exist pre- crypto um so I've got some working papers that are on ssrn um on cooperatives and Dows um and then I've mentioned the one here on Dow's data trusts so the initial version of that was written in 2021 the link there is actually 2022 and the idea there is uh there's really fantastic uh research and actually practices going on around you know bottom up uh you know kind of vulnerable constituencies being able to work together to have uh you know like legal representation Advocate on their behalf as a kind of a fiduciary for data and that paper kind of says hey this is what Dows want to do instead of appointing a lawyer to act you know in line with the law on behalf of a a group's preferences and represent them in how data is utilized you know um sold monetized used for research and so on uh the rules are just encoded in in the rules of the Dow and you could have protocols interacting and doing that for you uh so that's like a very active area of research and then some of the reviewer feedback on the initial version of that um ethnographic research was you know what could this look like as protocol level governance and obviously that's not my area of expertise but um they mentioned um sort of how how the kind of fundamental Protocols of the internet works um but again someone pointed out today that that still requires people to fly around the world and you know negotiate those slas and and operate that infrastructure which has become very very centralized and so uh as part of this kind of um speculative design space uh I wanted to transition to to talk about uh pathetic dot Theory so um this is more uh focused on regulation specifically as kind of a subset of thinking about governance but really what I'm pointing out in this slide is that um there's more than just technical forces that uh influence and regulate Behavior so unfortunately each of us is the pathetic dot uh in this uh structure yeah hands up uh and so you know thinking about these forces we can also think about them as levers when we want to design effective models uh in different contexts as patterns uh to govern dat to govern data and uh when looking at regulation um I wanted to point out another piece um entitled algorithms as policy and this is still in the um the blog post uh phase it was uh we haven't transformed it yet into a research paper um but it was co-written with my colleague from block science Michael zarum um and he's actually done a funding of the commons talk on specifically um this topic of algorithms as policy uh but the idea is that um in regulating you know we're actually you know taking law as code in a lot of uh cryptoeconomic systems and trying to kind of translate policy level decisions into algorithmic systems so then we have to think of algorithms as policy which means going back to the policymaking processes that acknowledge the social dynamics and so um I thought this was a really nice quote uh that was on 20 years are reflecting on Les's work on on TW on so sorry on the laws of cyberspace and so you know coming back to that broad lens of how do we get you know more resilient decentralized infrastructures um rather than you know having this last minute policy approach like oh we need governance or the social stuff is messy and hard and political uh perhaps we can you know paint those objectives uh to start out with and actually think this through now before just implementing the things and you know moving fast and breaking things and hope we don't destroy the world and so this is my uh very initial thinking on um on what some of these models look like and it's an area um that I'm intending to focus my research on for uh 12 months to come and continuing um in relationship with protocol Labs on that and so here I'm invoking this idea of self- infrastructuring and this is an idea that comes out of my PhD research as kind of a core finding and it's that resilience in decentralized Technologies is the idea that you're infrastructuring so you're engaging with infrastructure is an an active process of doing but it's it's self infrastructuring the whole kind of Crux from the history to the uh kind of modern day manifestations of of decentralized Technologies is that you know people want to operate their own infrastructure or not have a a third party that could be you know corrupted or co-opted do it for them and so coming from this lens we need to be sens to those local contexts with ipfs as well of what does it look like for people to be able to self infrastructure but still bridge that Gateway moment to store their data uh in line with their preferences it doesn't mean storing it forever it might mean storing it for a very short term and actually being able to um to ask nodes to take it off um ipfs Network and so what some of these pictures mean uh and I've got time to talk to them which is cool so um what some of these pictures are are initial models and so uh rather than just talk about you know a broad data Dow as a model I mentioned there's not kind of great patterns or definitions of what a DA actually is um I've drawn a very bad picture of um the viable systems model so another piece of research um by Michael zagam and myself uh looks at how um cybernetic principles align with decentralized governance and what vssm gives us is pretty much a map and a tool to look at any system and see what essential components to make that system viable are missing uh and so often times that could be um uh conflict resolution you know so so every Dow to be a viable system needs conflict resolution it needs forward planning so it needs the ability to uh look forward and look outwards at environmental factors and think strategically about those and plan for those um and this also shows um how operational units in a can be um as autonomous as possible while still kind of accountable to the to the greater whole or create wholeness um in the system through the process of metag governance so you can go read all about that but that provides one generalizable pattern of of a data D which has you know the components needed to make a viable a viable system of of governance or viable organization uh a local first Consortium so coming back to that idea of omn lingo you know maybe because it is you know culturally sensitive data um specific to a geographic uh community that doesn't need to be abstracted to a global storage network uh people can engage in you know operating ipfs nodes um within the local community and then they can all um store that on a on a local hard drive on their local computer um and that could be uh a generalizable pattern for for where something starts and you know one of my research questions moving forward is you know at what point does something become scaled you know so what what kinds of data or what kind of contexts need to be local first and what kinds can be yeah like absolutely amazing put it on ipfs you know engage in the cryptoeconomic incentives of that system that's you know that's the right thing for that data set um marketplaces for storage and retrieval so filecoin is um one example of that and I assume people are broadly familiar with how that works um and then you know Self Storage so that's another use case uh example where you know things don't need to be uploaded to a p network um obviously ipfs is not natively encrypted um and maybe you just want to store it yourself and so I've mentioned some of these further research research directions as I've gone um what is local local and what is scalable um data Dows um in practice uh super recent area um I know there's workshopping going on now on on one of them um but you know how how do you build a data di um what are the limitations of that as well uh and then thinking about those four forces of of Regulation so the you know the market the social norms the technology and the architecture so the infrastructure and um and kind of structures of society as kind of tools in how we sort of pattern out um these different generalizable models and so if you're interested in this I am told that there's like an unconference session on Saturday afternoon as part of ipfs Camp um and I would love to get together and kind of research these and kind of generate some kind of patent language that we can publish and and hopefully with the support of protocol lab share out uh and you know this whole kind of exercise of looking at you know end users and real use cases of ipfs um and resilien or through a lens of resilience is um about this broader mission to improve the internet and Computing generally thank you so I do have a few minutes for questions if anyone has any uh when looking at resiliency of networks like ipfs how much how important do you think it is to uh sort of have the network self-reflect and by that I mean um essentially be able to tell how much a data set is replicated across the network so you can determine if you need to incentivize it uh its replication further and how impactful that data set might be to the rest of the network and the more impactful the more it should be incentivized to replicate so uh yeah how important is it to self-reflect as a network yeah that's a that's a good question question I know that there uh about 3 months ago in July when I last checked with um Janis uh from protocol Labs there was incentives out around um tools to actually map the network so I think there's uh you know kind of over a year ago when I started the research there was very little information about you know where ipf nodes are hosted kind of geographically around the world but I think there's um more and more tools out there to help kind of um and and map the network pretty much um when you mentioned um reflective I was thinking about the ethnographic practice of reflexivity and I was like yes people have to reflect on their practices um but that's another question I guess in terms of uh you know how much should be replicated uh across the network and should people do that out of the you know that's a social Norm um um model really around just out of the goodness of your ipfs best node keep hosting it and and replicate other people's stuff um but really the networks it's not designed to do that you know long term or or in line with people's kind of context or preferences of how that data is stored or where or by whom other questions thank you um really nice talk um I I guess what uh stuck out for me in the vssm model was the anticipatory system and I'm just uh I'm just wondering about your thoughts on uh the modes of forecasting Futures in order to reduce potential contingencies um by which I mean uh data getting taken offline or you know what parts of your infrastructure are necessarily s self-hosted versus other parts that are not yeah if you've got thought that' be great yeah that's a wonderful question um I point out in um you know in writing about um the vssm and and drawing on cybernetics I'm not a control theorist I I worked with one to um to do that research but uh what you're referring to there is called meta governance in the viable System model um and so that's um the kind of uh it's compared to the human body so that's like the kind of higher order thinking part of the body that's able to look you know across the system itself and and see how everything's tracking each of the operational functions but also able to look outwards at the external environment and so in a traditional cooporation that would just be like you know the SE Suite uh you know and and the annual kind of strategy meeting and quarterly reports or whatever um you know in a dow what does that look like for people to think beyond their daily operation I think that's um yeah that's probably an open uh area of research but you know drawing on these kinds of provocations around you know even the threat models like I said you know if if people are uh thinking well my you know state government might censor my blog content because I want to write about what I want to write about um that's hard to forward forward plan for um but obviously yeah very um uh very like has strong implications for um the resiliency of that data itself so a good question for further for the research thank you cool we're at time thank [Applause]