New Directions for Mechanism Design description | Ori Shimony (September 2023)

right hi everyone I'm going to be talking today about new directions for mechanism design how we can do better experiments and have a better understanding of the design space so that we can more effectively create some of the mechanisms that Josh was just telling us about and whatever else you can imagine so I'm speaking uh through the mechanism Institute which is a new project that I started dedicated to researching problem and doing so practically but also with the kind of collaboration and contributions of the academic and more theoretical side but really linking that into practical experiments and so the fundamental premise here is that crypto gives us a new substrate on which to do the these sorts of experiments so previously what you would do through bureaucracy law um government you can now do through code and not only can massive centralized players who have lots of capital run these sorts of experiments but everybody can right even if you're not technical you might be able to use a software interface to click click click configure something and launch an experiment that manifests some room you have of how the economy or how some Poli system should work so I'll pause there and give you 60 seconds to turn to your neighbor and tell them 30 seconds each maybe one idea you've had about how crypto can change the way something works something small something big I'm sure everybody's thought of some wacky idea of how blockchains can improve the world so go ahead to me the main thing switch sense [Music] so I hope that kind of triggered maybe the moment you got into crypto blockchain what a smart contract was and you kind of came up with that crazy idea of oh wow this could solve this problem I've been having at work or this industry's having or that my friend's trying to do something together having and I want you to like hold that in your mind as we kind of walk through this design space and try to feel out how uh what experiments have gone on so far could maybe reignite that dream and so right now like maybe this great idea but there there's sort of a problem with getting it into reality right so one not many people understand the breadth of this design space right so there's people who call themselves token Engineers or crypto economists or mechanism designers but you know anyone can really call themselves that right there's a really wide range in what that actually means and you might be talking to someone who is a cybernetics Enthusiast or who um is a coordination nerd or who is very into governance or who's like a total like Finance nerd right Financial engineer so it can just mean lots of things and it's a high crossover of different disciplines right and then also so far in the space um you know maybe you have some cool interesting idea that can actually improve things like the ones that we heard from Josh but the ones that are going to take off like we heard from the audience are flipping nfts right so these things end up crowding out the space and making a lot of us jaded and maybe even leaving the space we get upset with what's actually going on compared to what's possible um yeah also it's just easier like from the perspective of a project let's say that raised money um they're going to start to be risk averse maybe that one thing they're trying to innovate on like they're bringing some product to Market so they're going to go really conservative of their governance and just copy paste you know what Unis swap did um yeah and lastly there's not really this nice bridge I think between the people thinking deeply about this from all these different disciplinary perspectives and the people that are actually out in the field building right and so we have this like messy ecosystem right now of Consultants BCS and startups and Dows which are you know later stage startups in some cases really struggling to figure these things out on the Fly and so that's the premise for why I think we need to basically build more knowledge and experimentation around the space and that's what we've been doing so I'll kind of go through some of the outputs that we've already created because I think they can actually immediately be practical and you can uh check them out yourself and so these are the different like modalities that we're using at the moment um for exploring this space and so we started with the mechanism Library which I'll walk through in a moment which shows you kind of what's G on so far in the design space and everything's still a work in progress but this is already uh live we also are doing various experiments on formalizing a design methodology for mechanism design in crypto economics that takes inspiration from different um devices and combines them into of an elegant framework that is fairly easy to use as well as looking at modeling so um Financial models simulations and then also of course AI why not so we'll get into the applied research end as well but we'll start with the library yeah so if you go to mechanism. Institute we already have a library up and so we have this very simplistic um schema right now for how you can tag different kinds of mechanisms and so so the the simplest way I think of starting is with value transfering so this is basically different mechanisms for moving value from between parties and this is just you know a small sampling of them we have plenty more there and we can also I'll leave time so that we can like click into these we can like go to the site and talk through any mechanisms you think are interesting but you can see that like some mechanisms obviously cover two of these uh tags like for example the nouns Dow came up with the concept of continuous auctions um which basically uses I think like Zora's auction house um on top of a fork of compounds governance that uses nfds for voting and so this is an example of a mechanism that is used to like in like uh orchestrate a purchase so you're purchasing NFD it sells one nft per day and then also it's a fundraising mechanism uh because that's what's financing the organization because in some cases you're not getting pure utility out of the purchase but an expectation of future utility and in that case it's also uh governance primitive next we have fundraising so there's been a lot of experiments here um you know a few that have gotten very popular but I think many more that have not gotten popular that are buried in like forums and uh prototypes and so I think a lot of these are worth taking look at because uh if you look at the the projects that have succeeded today with fundraising um most have have just gone a full conservative route right like privately raising from um VCS and these more experimental uh measures have have had less traction largely because of regulatory reasons but I think there's an in between there that people have to go deeper on yeah then we have signaling mechanisms so these have to do with like information uh curation of information um yeah incentivizing uh the surfacing of Truth in different ways governance is pretty self-explanatory and I think one one Insight here that I'll touch back on later is is the need for algorithmic governance mechanisms and also incentivized governance mechanisms because today there's kind of this naive token weighted voting which is deeper in the library which most projects are using but there's plenty of you know critiques out there and I think what what's very interesting to look at is these more algorithmic approaches to governance like conviction voting uh where different people are staking their preferences and then over time the the protocol or the algorithm is making the decisions on behalf of the group so you don't explicitly do so um like most protocols or systems do today kind of borrowing from like let's say like uh government and corporate governance where of yes no voting on proposals yeah next our Market mechanisms so things like auctions um matching systems uh currency design Primitives and lastly value capture so how do um how do systems that are designed designed that somehow like bootstrap resources facilitat and transfers of values between different parties somehow govern that you know how do these systems return value back to different stakeholders right so we can dive back into specific mechanisms I'm sure there's a lot of questions there later uh but I will go next to cover some of the design modalities uh design methodologies we've been working on and so one space that I haven't heard talked a lot about crypto is actual like the game design literature so there's really extensive literature from video game designers and board game designers on how do you go about like building a system with certain tools that then various players are going to participate in and how do you predict the emerging outcomes that that will have and so I think one of the most heavily cited papers in this space is this uh paper called the MDA framework the mechanics Dynamics Aesthetics uh framework and the idea is that that the game designer has control or the perspective that they have is on the mechanics of the game right so what we've been fing crypto the mechanisms or the coordination Primitives and so that's what the the mechanism designer or the game builder sees right but then um when you actually run the game you you have these higher order dynamics that emerge that's the runtime behavior of the game which they call the Dynamics and so these you know you could predict them and you could through looking at the experiments that have happened so far start to have a mapping between combinations of mechanisms and then Dynamics you expect to emerge and then lastly the thing that's further away from the designer is the actual impact that the game has and so in in this context they thinking about liter like games that people are playing to have fun and so they call this Aesthetics like what is that like experiential thing that the human feels when they play the game and so right now we're working on adapting this model to crypto and looking at the Third third side of it the third step is more so the impact that the game has because if you're designing if you're you know designing a universal liquidity protocol the impact you want to have is like very low slippage trades if you're making a refi protocol maybe the impact that you want to have is reducing carbon emissions and so the player we have to note uh has the opposite perspective they just see the experience so the impact that something's happen and they might have no understanding of the mechanics and so there's a lot more in this paper that you know the game design industry has found useful that I think we can adapt for our needs and this is just an example that that we put together of how you could take something like the the business model canvas and then map something like the mechan mechanics Dynamics impact framework onto that when you're designing some sort of cryptoeconomic system so you can think about the technological platforms you you'll be using The Primitives uh which we have mapped out in the library and various components and how that then you expect that will result in certain dynamics that will have the impact and maybe some externalities on the players um yeah and I think um one uh thing to note here that a lot of papers that came after that original MDA uh paper uh were very critical of the fact that it was kind of like deterministic and almost like waterfall style and so the part of the framework that we're working on is also how you can actually apply this framework and feed empirical data back into it right so you you do like a really quick and dirty prototype uh you do some modeling you roll it out in the field and then you update and then you update and then you update so you actually run experiments to get better data on what will work um yeah so another kind of like orthogonal direction is just thinking through uh tokenization more systematically so okay people talk a lot about tokens and they lots lots of times handwave you know what the tokens are for uh right the tokens are for governance uh or they're the access token for the system to the currency but I don't think there's like an infinite uh design space for what a token can do I actually think there's a very uh like limited set of things that like a tokenized rights can give you um and this is like a starting point for thinking about it I think um and so I think the the really exciting I guess aspect of tokens is that they take us beyond what like shareholder like stock uh corporations have which is a government economic RS into the ability to program complex access rules and then um another like misconception I see generally is that people when people think token they think um you know transferable usd20 or Bitcoin um where then maybe as of the last two years they'll think oh nft right so this like collectible art piece that I can like send to anybody but I actually think it's useful to think about the whole design the spectrum of what we could call token as encompassing things that we wouldn't normally think about that aren't financialized right so you have a spectrum where you can code in restrictions on transfers of tokens or like block or prevent uh transfers entirely like in the case of uh your passport which is non funable it's unit to you and you can't it doesn't have validity if you just sell it to somebody and then on the top right quadrant we have uh units that are fundable so every one is like any other one but they can't be traded between people they wouldn't have the same significance like Reddit Karma let's say or Uber like rating for and so contining on the tokenization theme I think the third um the third side of this that again there's patterns that have taken off that people then just mindlessly copy is how do the tokenized rights get distributed right so after we know uh what the tokenized rights represent how we're going to uh actually represent them in data structures and then we have to think about how people acquire rights over time what do they do to get them and so there's you know lots more in the library on ways to do this but these just this is just like a basic example looking at again un swap okay so that's a little sampling of some of the design methodologies we're looking at and quickly I'll show you just uh Snippets from a project into a getcoin on modeling the impact of different mechanisms so thinking how you can then take conceptual Concepts like token saing burning a token um right like bonding curves um and to pered Registries and then actually impact like model the impact these different mechanisms will have in different uh values that you put for those parameters what will they what impact will those have on your overall goals on that impact that your system is trying to achieve right so if you're trying to like prevent Cil or capture value in your your governance token then you have to understand uh um how the different variables will affect those outcomes and lastly I'll I'll touch on the um the AI uh bot we're making so this is a bot that's trained on the library and we're constantly updating the library and then uh needing to retrain the the bot and when I say train I mean run it like on an embedding of the library so it does a semantic look up based on a query in the library to see which mechanisms um are most relevant to the thing that trying to build to the coordination of problem you're trying to solve and it might like suggest uh a few mechanisms and then tell you examples of how it's been used and then lastly create like a a Proto specification of how your mechanism overall will work like a step by step of how the different Primitives could fit together and so this is one of those things that is very rudimentary right now you can demo it if You' like um but it just gets smarter over time as we have more contributors to the library and we have more empirical data just basically plugging in all the case studies that have already happened in space and so this is like uglier what it actually looks like that was just a a mock up here and so yeah you can just ask it these sorts of questions and it might like grab the right entries from the library to suggest uh an overall uh protocol flow the different participants the key Primitives that it's using and this is just like a first part of that but we can like take uh audience later and yeah so those are a couple of the materials that we're working on and putting out and some of which are already out there now this is kind of where we're where we're going and where we think the the biggest need is uh for crypto isn't actually like doing the experiments so for example getcoin we designed for them uh like a civil detection game for passport that lets uh various actors of data scientists uh stake GTC to sub submit lists of suspected Bots and duplicate accounts and then the smart contract will take the highest incidence over those and then slash those passwords automatically and reward the um the Civil Hunters who put that U we submitted those those uh accounts and so that's great in theory we have the Prototype it's public on GitHub but um we actually need to run tests on that right like you're not just going to throw that into production and then like uh potentially jeopardize uh the identities of like hundreds of thousands of users and so you know some next steps for example with that project is to actually organize that an experiment put real money real skin in the game get like beta civil Hunters together and then run these like sandbox experiments and so we want to do that sort of thing not just with particular protocols but then also um yeah with communities with organizations that are using mechanisms already and so we have like the whole like uh research road map there that we can talk on later design competition I think is another use useful modality so kind of defining problems maybe even outside of crypto that we think mechanisms can help with and then putting together individuals from like teams that that do mechanism design that understand the importances of smart contracts to solve those problems let's say for example let's how do you prevent human rights abuses with mechanisms what are some like problems in that space maybe you talk to Human Rights Watch and amnesty and get them together and then you bring the mechanism designers and we can like Workshop different solutions based on experiments that have already gone on with crypto um and yeah don't have much to show on the others but uh yeah basically working this all into like an overall curriculum and then like I said doing these experiments in the field not just with protocols but with communities through established uh research methodologies and these are some of the specifics I may have mentioned a few of them kind of these areas where we think there's like uh low hanging fruit where we can put in like a little effort uh in the next few months and actually come out with uh with solutions that that move the ball and these like kind of difficult problems that maybe maybe people in the space have got a bit lazy on and just like moved past because the crappy Solutions work work well enough for the established players and we can dive into any one of these uh that people are interested in after yeah it's just a bit more about where we're going very very B you know without actual so that I can't get off to account yeah and that's it I wanted to leave as much time as question thank you very much for who will be the first on questions see hand that is another hand s to go through hey I was actually really curious about on the list of uh short-term projects thank you the harburger Tex licenses uh can you say anything about who's working on that software licenses what I assume what yeah definitely so how many people here are familiar with the concept of a harburger test about half the room so actually SE and I were e Berlin 2018 did a project on like harburger tax at space and so this is a throwback shut up um but yeah they kind of died in space since then like Glen Wild wrote about them in radical markets and that's all I said they were cool and so y did a prototype of uh selling address space through harburger Texas and I don't think anyone ever used them be on like some art projects which one sash in N okay awesome all orss orbs yeah so I have orbs in the library actually as a as example should hard yes so I'll explain it really quickly so it's like this this auction mechanism where you you have some property and you set the value that you actually value it at and then you get taxed at some percent of that over some frequency but the catch is anybody can buy the item from you at the self self proclaimed price and so it incentivizes you to like um note the True Value that you value the asset that so it was first proposed in the context of housing because um property tax is really like arbitrary institution where like a assessor comes and like decides how much your house is worth and looks at the previous purchase price and anyways it's just a very inefficient market and allows like uh people who pull lots of capital to just sit on lots of property and keep it idle and so this isn't meant to be set of mechanism to prevent that so anyways back to the question um yeah I think that kind of mechanism is probably going to be hard to apply to like people's houses but we could start to apply to more like virtual uh Goods where there's these like distribution problems for example intellectual property right so this is one of these um spaces where uh you have lots of innovation like like in Creative comments and you know different like uh uh open source licenses but how could you actually take something like patents right and then issue the the ownership of the IP as a harburger task so that you don't become like this Perpetual or like whatever it is 20-year monopolist uh with like the ability to withhold your your invention from the market or just like sell it to like um Ser prices how how do you do more fair pricing mechanism for IP basically or copyright and then the resource you said is like environmental resources so you know El or osto is like examined in in depth like common pool resources like Timber uh fisheries and things like this and there's various licensing schemes to limmit um the amounts that these resources can be extracted from so that it's below the uh replacement uh rate of the ability the resource to provision itself so anyways you could you could program those kinds of constraints and conditions into some sort of hardworker tax just to have some more optimal allocation of the rights to use environmental resources that are limited thank you any last question last so so this is a lot about mechanism design and from the way you present it you kind of imply that mechanism design is a thing that is absolutely a good thing for the world you can mathematically prove this and it's just a case of applying it in some ways and I'm just wondering if there's any sort of um naysayers or kind of like doubt yeah about whether a mechanism a mechanism designed world is what we really want and what the drawbacks are you know I hear a lot of positive things about it and when I hear so many positive things I want to know what's the brother yeah no I definitely agree with the skepticism and there's definitely dark mechanism designed for like status quo enhancing or worsening mechanisms out there and the two that come to mind that you're all probably very familiar with are like um cases where there's coercion where it's like a central bank digital currency or it's like an orb device to scan your iris that's owned by a billionaire who controls our information like creation systems and so there's dark mechanisms design in one case I think when there's coercion um in the use of the mechanisms and so you don't have an opt out I think the beauty of like the way I'm the way I use mechanism design is that these are optional systems right so that's a key piece right the ability to exit um of course Network effects can have U implications here but in open source world and in like a multi-chain interchain world these concerns can be mitigated so yeah I think that's one area to like be be careful of like don't help authoritarians design mechanisms that they can coers to use uh another um which I'm sure we've all been victims to to one to some extent or another is like these like super uh op fiscated mechanisms that are just like Ponzi schemes with extra steps right so you can they we have very good and some of the best mechanism designers I know have like unwittingly contributed to like very complex ponu schemes that they didn't understand what Ponzi is but the people didn't understand it doesn't literally to be a Ponzi but something that doesn't have fundamental soundness but there's so much hype around it that people go in anyways and so yeah I think we need better mechanism design because this is you whether we like it or not authoritarians are going to use it and scammers are going to use it right so I think we need to have a more robust understanding of the design space to know what's going to lead us towards better outcomes potentially most capitalist ones um or ones that can make us flourish more or mitigate existential risks versus mechanisms that are going to exacerbate this cool thank you thank you for the tool really really interesting yeah I'll just add something there uh not really a question but rather kind of uh to add on kind of like how it could go wrong yeah just when uh Orie was talking about um kind of honies with extra steps um like Tera Luna might be an example of that um you know maybe something where you're talking about uh bad outcomes or like second ughter third ughter Anor effects uh you might say something like big clout or friend. Tech um one could easily see this leading to dystopian outcomes uh so yeah mechanism design can establish a game but uh this makes me think about kind of like the uh the Jeff gold buum and Jurassic Park thing like just because we can should we you know and like can we think about like the intention of these systems uh and really design for what outcomes we want oh hi there super interesting talk um so I think like right now we have a lot of spare block space which a lot of l2s are trying to sell and we are also seeing a lot of fragmentation in the space like as a Dap if you build on one chain it's hard to kind of work with other chains and whatnot so I was wondering kind of if you can go to the slide of future research directions um I was wondering like how many of these U kind of Primitives that are trying to build on like identity and fundraising mechanisms and whatnot would be chain agnostic able to work across chains with a public key or something else versus those that are specific to uh L2 or any particular chain yeah I would say almost everything we're looking at is chain agnostic it's just working with the concept of um Humanity now has smart contracts what can we do with them and you know of course mechanism design can go below the layer of smart contracts you know when designing smart contract platforms themselves l1s that's less work we're looking at um there is some useful full um analoges we can draw from those for example the dynamic uh parameter setting mechanisms of bitcoin's difficulty algorithm or you're block space algorithm we can like take those and transpose them to the smart contract layer to like determine the values of a perimeter of a a governance system for for example so I think you can learn a lot from the Bas layer uh but we're kind of agnostic in that sense and I think there's different people working on the problem of uh bridging mechanism or rather bridging daps um and instances of mechanisms between chains like you know sharing data uh or liquidity across chains hey um really amazing collection great ontology probably one of the best I've I've seen um I was wondering if you've given some thought to um in this community there's PL of people showing you know their protocol their Dow it all works great and they're and we're not really looking at when it fails or when it goes wrong and in terms of mechanisms and the collection of that um propagating mechanisms that I would say don't work but aren't well applied yeah have given some thought to that and how that that applies here with this collection yeah I mean the one of the most fun aspects of this project is to document the failures and I think that's one of the areas people should look more deeply at in order to learn what to build is looking at what not to build right so we we don't have extensive case studies at the moment most of them are a few sentences but that that's a big part of thear like the learning that we want to be a part of is for example um curation mechanisms people were really into that in like 2018 2019 you know tons of blog posts tons of proof of Concepts like not many of them worked right why is that right like we need to understand that we can't just like we're not just like promoting oh yeah use tcrs use curation markets you should like understand the you know why the ones that were tried worked and didn't work and it's hard in many cases because a lot of times it's depending on exogenous factors like the personality of a founder or Regulators or the market hype cycle but yeah at least we can try to like document what like worked about it what didn't work about it from a mechanism perspective coming coming and how do you manage to finance this sort of collaborative and open research and Langs yeah good question so right now we work like on a project specific basis but the aim is to avoid becoming like a mechanism design consultancy that's like designing mechanisms because there's plenty of like cook and Engineering Consultants mechm Design Consultants and we want to like chart a space that's more more uh advancing the science mechanisms and the design methology of mechanisms and so there's a few ways like were going about that one is um like one project many supporters right so in gitcoin case it was our first project so it's just gitcoin and just like civil detection and like publicist token research but like for a lot of the like research topics we've mapped out we have several like partners that could like collectively fund any one of those um and then the other modality I would mention is more so training um individuals companies communities and how to like do mechanism thinking right so more so almost like Education and Training and coaching um as we're making the resources we can then be like working hands- on with people to like shift the way they think so that they understand uh the design space better in the work that they're doing so rather than like just building a mechanism for someone handing it to them and saying here code it up uh yeah yeah more so like doing um a trick training and like support that way and then and then another does it help Finance yeah yeah this this is something that people need and so you know one approach is the create a useful service for the market and then the other is to like partner with existing research institutions Rob foundations donors and then corporate sponsors protocol sponsors in some case who are interested in like the actual Mission and see some value for the supporting that so we're doing with combo thanks last question I can ask for can you quickly elaborate on AI trading Frameworks uh yeah so I guess this is broader than training just generally a lot of the problems with like AI governance so you could call it AI safety um yeah basically uh the problem of um compensating and recognizing the provenant of data that's used in training and so there's been different models proposed that would use like cryptography and that flow um and then when it comes to AI um yeah another model is just like uh people exposing uh clusters of gpus running uh like using models right so basically the open source AI problem where either use a proprietary like centralized uh solution or you have hardware and you run local models open source models locally and so there's like this concept of like GPU Dows and these like local AI clubs and basically looking at how we can have mechanisms that can have the crowd sort of like rival the the be the monoliths in terms of AI there's a few other sides to the AI question that are not popping to mind but yeah it's a very one yeah we can talk in later cool thank you so much already Round of Applause to