31. Reputation-Based Governance to Power Funding Allocation - Taipei 2023

[Music] all right well thank you everyone for inviting me to the stage I'm super excited to speak today uh so my name is Anu I'm a senior program manager at the Stellar development foundation a nonprofit organization that supports the development and growth of Stellar so at the foundation I lead the uh Stellar Community Fund and this session will actually delve into the fund's Strategic shift towards leveraging reputation of in funding allocation decisions so before we dive into that uh does anyone here know about Stellar please raise your hands oh pretty good veryy good what about xlm okay actually I like that actually many people know both stellar and xlm so usually it's like only one of them but uh yeah so actually Stellar is an uh is an open source layer one decentralized blockchain network and uh it's been around since 2015 actually before ethereum believe it or not and is specifically designed to create Equitable access to the Global Financial system so it provides infrastructure to developers to issue assets build applications and deploy Financial protocols and direct access to the Stell network is available either through a dedicated set of functions available in several uh programming languages or through our new native a rust based smart contract uh smart contract platform which is currently in Tesla net and what's that Stellar part so uh apps can actually integrate with Stellar's extensive on and offramp provider Network to get access to the world's currencies and Sellar actually has the largest uh uh off-ramp uh Cash Out provider Network out of all blockchains out there through a collaboration with MoneyGram where Stellar apps can actually integrate with uh with to offer endtoend solutions to uh real world problems so so the growth of this ecosystem is led by the Stellar Community Fund or SCF for short and I'll be using that abbreviation so SCF stands for Stellar Community Fund uh which is an open application Awards program that supports developers and startups building on stellar and earlier this year the SCF supersized to foster a development on Stellar's new native smart contracts platform which is scheduled for mayet earlier next year so in 2023 we already award wed over oh yeah we already awarded over $1 million in xlm to over 140 projects and uh projects can actually include anything from educational applications to developer tools to financial protocols and can apply for up to $100,000 in xlm at a time for about three months long development all right so different from typical like Foundation Grant programs SCF actually draws on input from the community for both award allocation project support and governance and Community vote in SCF is actually an exciting time for verified members who have proven reputation in the community to signal for which projects they like most so why Community vote so right now SCF actually is Solly fun funded by the foundation so decentralizing decision making uh in funding allocation is actually not necessarily A Natural Choice and so why Community vote for funding allocation in SCF I have three reasons so number one shared responsibility uh it's important to fund the right projects to build a very strong ecosystem but as we all know this isn't very easy so decentralized decision- making encourages a culture where decisions are made thoughtfully considering the impact and implications of The Wider Community second we have credibility so the diverse view points and and Collective wisdom of the community actually validate the choices made and enhancing their legitimacy and accept acceptance of the results third scalability and this is one is very important so as more members uh become involved in governance it cultivates a stronger more active and invested Community which is essential for long-term sustainability um an expansion of the ecosystem now the why is pretty clear but the how is what's difficult as getcoin founder Kevin auki points out in this tweet I kind of like this tweet um that's why I put it on there but I don't know anyone knows this one anyway uh yeah um so yeah so earlier this year we actually collaborated with R&D firm uh blog science to create a novel voting mechanism to power funding allocation in the Stella Community Fund and blog science actually uses the following three-phased process to design solutions to complex uh problems and so starting with kind of mapping the requirements uh and desirables uh followed by the spe specification design and integration of it and finally uh the validation of the system by thoroughly testing and simulating its performance starting with phase one we've analyzed the following desirables of how funding allocation should be governed on SCF based on both Community input and prior experience with voting uh mechanisms so this mechanism needs to be effective and impactful to determine projects that should receive funding the decision-making process should be fair and really reflect Community sentiment and reward beneficial Behavior it's also important that it's scalable uh so that and that it avoids overwhelming voters process should be also transparent and publicly available um but still remain individual voter privacy which is something that we really value so that voters can actually be honest in their decision-making process it should also be dispar uh it should also be flexible to be able to adapt to varying uh quality and quantity of project submissions each round changes in voter activity and also allow for easy modifications in line with the evolving needs and challenges of our ecosystem so finally we wanted to also this system to be Innovative and actually use Stellar's technical capabilities where possible so once we knew desirables we needed to make some decisions on the design process of the overall um mechanism and there's a lot of lot to figure out so for instance like who can vote how are votes waited and also like are votes binary like yes or no or non-binary like with matching Tool uh pools there's a lot to figure out and we've already we looked at the various already existing voting mechanisms out there such as one vote per unique identity this model is quite Equitable and and doesn't but doesn't allow for much flexibility to in like voting power weights uh towards to let's say reward experts uh it's also quite difficult to Check's identity uh as many people pointed out here during the conference and unless you have a very robust like KC type of process it's often subject to civil attacks so there's also um one token one vote for instance so the more token one has the higher the voting power it's a very popular mechanism but as members often can buy themselves in the system can actually cause misalignment with the initiatives due to also monetary incentives we've also even considered a quadratic voting mechanism which is I guess that quadratic funding is very successful in gcoin but from a previous attempt the community vote actually became more of a popularity contest uh which didn't necessarily result in funding the right project at in our ecosystem at least uh just the ones that were really good at marketing themselves and uh well let's face it not every developer is great at marketing we all know it but they are very good at building tooling so we need to fund the right project which is something that's very important so something actually that really sparked our interest was reputation-based governance as it was something that we were already naturally doing in a more like manual type of way and somewhat worked as well as well as delegation is a very popular mechanism out there and also generally like while working to solve for voter apathy issues now reputation based governance it involves evaluating one's trustworthiness expertise contribution history and even alignment with uh the initiative values to influence both voting power and or to assign membership so in funding allocation uh reputation based governance can have several benefits including more informed decision- making as it could be more voting power to experts for instance more inclusive representation of the community as governance doesn't need to be limited to token holders or validators but can actually include many different types of user types uh also it can boost On's rep boosting one's reputation can also Drive intrinsic motivation rather than often miss misaligned extrinsic motives such as monetary gain another plus is that the implementation is quite versatile so actually it can be layered with other mechanisms and actually there's a lot of rep U mechanisms even the ones that were described during this conference that use some form of reputation based governance and finally I want to kind of make the note that reputation is also a very intuitive thing so every day we make choices based on reputation without even realizing it where we decide like who to PR friend or who to with the task or who to partner with on the project we're evaluating reputations so these judgments are quite intuitive and they're process seemlessly by our own brains however translating this intuitive process into a formal governance system is quite challenging it requires juggling many types of factors which are often um overwhelmingly complex and another potential issue to be wary of is that influence of a certain members who have aced a lot of reputation from the start can remain dominant or unchallenged over time so how to solve for this together with blog science we've designed and implemented the first MVP of neural governance which is a modular transparent and flexible architecture uh for defining defining one's voting power inspired by neural networks the system mimics our brain's information processing through interconnected voting neurons and so each V voting neuron actually consists both of a Oracle version uh Oracle function and a waiting function so the first one in quites encodes all the core logic of the data input and the second one encodes how it should be weighted and so the outputs of from all voting neurons are actually aggregated together in layers and then pass them as input to the next layer until a final value is reached so for the initial MVP for the sell Community Fund we've actually uh implemented three neurons you can see them here on the diagram so uh the first one is um is reputation like expertise which are um kind of the input comes from SCF or from Stellar uh challenges badges that people have accured over time uh then uh there's a voting history neuron so the more like someone voted some more time someone voted the higher kind of that weight is then finally we have a trust NE neuron um where kind of other people can assign Trust two members and you can actually see that it's layered so uh the first two neurons are collected before voting and the second one is actually collected during voting so it's a lot of flexibility and these are actually some of the more I guess like three MVP versions of neurons I think like going forward we're going to see like oh how can we actually um reflect Community sentiment and reflect reputation um in a bit more yeah of a sophisticated way all right um so by adopting this neural network approach neural governance can simplify and uh simplifies and structures the evaluation of reputations it also breaks down the intricate process into smaller more manageable Parts it offers a plug andplay experience so you can actually add remove or change voting neurons as part of a more adaptive governance process and it also enables peace wise transparency this is something very important if let's say something like voter privacy is important so you don't want to reveal the Identity or you know the particular projects that a particular user has voted for you can but you still want to have the entire thing be somewhat transparent and have a public record you can enable peace wise transparences so that some neurons might be public and others might be more obis cated so overall neural governance can also be used to enable more like inclusive participation in governance as it can also be used to determine membership some potential um use cases include public is funding so neurog governance has the ability to incorporate evaluation criteria Beyond profit potential such as prior Network participation prior voting Myster experience with public goods and more and it all also allows a community to easily adjust weights on those signals uh second is domain specific expertise reputation so similar to the one before neural governance allows a community to set neurons for a particular governance action to be higher weighted towards individuals with a particular set of expertise such as technical engineering work or social impact work or even academic background oh almost um third is adaptive governance as neural governance accepts inputs in at each layer actually transforming them into weighted outputs for the next layer it allows for um crossr adaptations and also prior voting actions so let's say like project success or and other outcomes can be an important signal for uh future voting power allocations so Something That We're actually looking to explore is and in the uh is to expand the voting history neur ontak success data of projects so that actually uh extra voting power goes to those who voted for successful projects in the past now there are several other projects that are leveraging some form of reputation based governance in the blockchain space one example of that is uh Colony let me see I accidentally clicked next on my on this one instead of this all right here we go all right so one example of that is colony and colony is a dow which offers tooling for others to build Dows uh reputation in Colony is an attribute that can actually quantify a particular individual's influence and this is calculated based on the sum of work that has been completed within a colony or Dao uh reputation actually decays over time so members need to stay engaged to keep their status and this is pretty important in some of the drawback that I mentioned earlier so that people who have gotten uh in in a colony very early on and acur a lot of reputation actually need to kind of stay engaged to keep that status and not automatically keep it so the other mechanism I mentioned earlier that we're interested in for SCF is delegation uh so for the interest of time I think I'm like five minutes out or something um I'll go over this pretty briefly as it's a popular concept but in short delegation allows members to entrust of their voting power to another trusted actor and the most important benefit I see if a delegation is that it reduces attention cost uh technical requirements and also administrative overhead for individuals that said there are some uh challenges with delegation most notably The increased risk of collusion and centralization and these are the very things that we try to solve for in blockchain so how do we solve for that Forum delegation this is uh together with block science we created a a different take on how delegation can be done in governance mechanisms this is called we call a quorum delegation which allows users to passively vote for uh by delegating their voting power to a quorum or a group of users and this is actually inspired by the Stellar consensus protocol which is our uh which is a proof of agreement type um consensus mechanism where the validators actually Define their trusted Quorum of all other validators in order to validate transactions on uh this Stella Network so voting power in this case goes to the uh majority decision of the chosen delegates who can be changed each voting round and you can also choose to delegate per decision so there's quite a bit of flexibility uh with with this type of mechanism ISM and also you don't need to it's without having to reveal any information about the quum users can automatically vote so these are kind of anonymized quorums as we say so this can reduce the attention cost for users while still allowing for pretty high flexibility now I understand you know it takes quite a bit of time and effort to set up these s systems initially uh figure out what neurons that you would like to create figure out the waiting figure out the parameters of these systems um I'm not going to lie it does take a bit of effort but uh to help out anyone here to experiment with the system blog science actually created an open source tool called cat cat U maybe some of you have known about it uh but anyway it allows for anyone to do research and simulate uh different outcomes and an example of this cat cat simulation is actually this one for Quorum delegation um that shows kind of the delegation over time in a particular voting period uh it's pretty interesting I guess I don't think we're actually going to have 100 days of voting but I guess you can do the math like it's overtime uh things can change all right so let's put it all together um SCF implemented the initial version uh initial like MVP I would say of neural Quorum governance on our new smart contracts platform and this actually combines both the neural governance that I talked about earlier and Quorum delegation so this kind of like resembles some kind of like a liquid democracy I would say with the best of both worlds so on the one hand you have the reputation inputs uh to determine voting power also with the option to delegate so if you don't necessarily have the kind of necessary expertise to evaluate a particular project or you don't have time to like vote in a particular round or like let's say you don't have the tier to be be able to vote but you can still delegate so yeah setup collectory could really solve for some of the common issues that we experience across voting mechanisms like inaccurate Community sentiment like voter fatigue and apathy and also lack of uh growth in core group of users uh by Distributing voting weight in a fair way this also will align hopefully with all of thef objectives that we put out uh encourage engagement by empowering members can actually help build up their reputation and greater uh it allows for greater flexibilities on voter responsibilities and also offers some accessibilities newcommerce all right so I'm almost done I'm almost done uh some next steps so we've actually started early this year with the ideation and development and we finished the inle implementation this quarter and documented everything in several blog posts and resources and so next up early next year um are simulation calibration and adding new neurons accordingly uh We've also are going to put out month free reports based on the community vote runs that we offer and then and uh yeah the goal is to really fully rely on this mechanism U to power funding allocation by the end of June 2024 so wish luck uh so if you're interested if this was interesting to you definitely um get involved and if you scan this QR code it actually goes to the governance modules Library where neural governance and quum delegation are both um documented as well as where the cad cat implementation is available if you're interested if you're more like a builder you're like hey actually I want to like submit a project to SCF because it looks pretty cool exell I want to know what it is definitely go to community fund. stellar.org the next deadline is January 14th um yeah and if you're really interested you can become a verified member and join our Discord so yeah if you have any questions I went over it pretty quickly but definitely talk to me after the uh talk and I also like I'm pretty active on telegram I think I put my name out there I don't know you couldn't have miss me so definitely like reach out all right time's up so thank you so much thank thank [Music] you