Episode 250

Kleros – Crowdsourced Arbitration for Blockchain Applications

Clement Lessage Federico Ast

Dispute resolution is the process by which contracting parties settle disagreements. Whether in the form of litigation, arbitration, or other means of mediation, every contract defines a dispute resolution mechanism and jurisdiction. It is the metaphorical Lady Justice, measuring the strength of each party’s arguments, and reaching a decision based on evidence.

Smart contracts are unique in this sense. Unlike traditional contracts, they are rigid and deterministic. Written in computer code, nuances in human language and vagueness of terms do not exist in this realm. There are no judges, no jury, just calculated execution.

The DOA hack and other similar events have prompted observers of the space to express the need for smart contract dispute resolution. Some have suggested “exit switches” that would allow for human intervention when edge cases appear. But could the arbitration process be integrated into the smart contract and on the blockchain?

We’re joined by, Federico Ast and Clement Lessage, respectively CEO and CTO of Kleros. This dispute resolution layer provides contracting parties with a fast and secure process for arbitration. The system is broken up into courts and sub-courts, each specializing in specific matters like e-commerce, insurance, and transport. In the event of a dispute, parties submit their case to Kleros, where a crowd of expert jurors analyses the evidence. When all votes are cast, the decision is enforced by the smart contract, which may unlock funds, or provide parties with additional time to fulfill the terms of the agreement. Clever incentive mechanisms reward jurors who vote with the crowd, making Kleros resistant to bribe attacks and collusion between jurors.

Topics discussed in the episode

  • Federico and Clement’s respective backgrounds, including a crowd arbitration project called Jury.
  • The vision behind Kleros and the problem it addresses
  • The case for crowd-sourced jurors as a means to find the best judgment
  • The game theory and incentive mechanisms embedded in Kleros
  • Kleros’ hierarchical system of courts and sub-courts
  • How jury selection works and who administers courts
  • The system’s built-in governance mechanism and its purpose
  • The Kleros token, Pinkaion coin, and it’s utility in the system
  • “Doge on Trial,” a clever experiment to find authentic doges
  • The current status of the project and roadmap

Meher Roy: Today, we are having a conversation with Clement Lessage and Federico Ast from the project Kleros. Kleros is a project that seeks to merge the field of decentralized arbitration with RF arbitration with blockchain technology to create decentralized arbitration where commercial disputes and other disputes can be solved through a group of decentralized crowdsource jurors. Federico and Clement, welcome to the show.


Federico Ast:  Thanks for having us.


Clement Lessage:  Thank you.


Meher Roy:  So we can start with Federico. Federico, tell us a bit about your background and how you, what’s your journey towards building Kleros?


Federico Ast:   So my name is Federico Ast. I am from Buenos Aires, Argentina. I studied economics and philosophy in the university then I started working in the online media in the biggest newspaper in Argentina. Then in Argentina, we have a lot of problems in institutions.  You may have noticed that Argentina had problems with currency and problems with governance in the in the history. So people here became quite interested at a very early stage about blockchain technology. So what is this idea of a currency that doesn’t depend on a government and a currency that it’s in control of the people. And at a very early stage, also in Argentina, people started to be interested in the applications of blockchain not only to currencies but also to the governance processes.


Some projects from Argentina, for example, democracy here started to think of how to use blockchain for improving voting systems, for preventing fraud in elections. But no one was really taking a look at a very important part of governance of a society that is justice systems. People don’t have very good access to justice in lots of the world because access is very expensive, justice is very slow. So I started looking into that part of governance. And so that’s what made me become interested in how blockchain technology can be leveraged for making better institutions, right, in particularly in justice. So I started working on a concept I called, about then crowd jury, that was the crowdsource of juries. And so we met with Clement about a year and a half ago and we were working on quite similar stuff at different sides of the ocean.


And we’re introduced by a common friend, I would say, that Bit Nation, this founder. So she met Clement at the hack-a-thon and she knew I was working on decentralized justice. And she said, hey guys, you should meet and start building this together. And so here we are.


Meher Roy:  So what was crowd jury and how did your initial ideas differ from Kleros?


Federico Ast:  So and the idea of crowd jury was basically the use of, the main concept was use of crowd sourcing for creating a court. So in the end, if you think of what a court is, it’s basically a group of men that they have to analyze information about some event that happened or didn’t happen. So in the end, a trial is like, so you have some guys saying something about the other and you have to see what the truth is, right? And we have learned through the rise of the internet that some processes of information analysis can be done faster and cheaper by using the intelligent — collecting intelligence.  For example, Wikipedia, so instead of building encyclopedia like Britannica with one centralized agent and different experts writing. So you can crowdsource the information flow through the internet people.


So I said, what if we could crowdsource the information analysis that’s basically a trial into the crowd? So and that was the idea of the — behind the concept of crowd jury. It was like very, very early stage. So I think when I started thinking about that, so not even a theorem was launched. So it was a really, really early stage idea. But the technology was not yet there, but some of the main concepts were already in place about the use of crowd-sourcing. But the idea of crowd jury did not have and the game theory shelling point principle on which Kleros is built, so that was it. So the engine of Kleros was not present of the early ideas of having it on crowd jury.


Sebastien Couture:  And Clement, why don’t you tell us a bit about yourself and your background? You came from machine learning and transitioned over to cryptocurrency, if I’m not mistaken. Can you tell us about that transition?


Clement Lessage:  Yeah, so actually, computer science was in France and the US and even when I was a student, I was interested by crypto even before as obviously, mostly like playing a bit with Bitcoin but not seeing a professional just which is like all this. And when I finished my studies I had decided on which was offering a way more possibility to build application about what Bitcoin was doing at the time I started being a Bitcoin base. And so I was interested by Sadao, I was interested by dispute resolution even at the time of Bitcoin, at a time of Bitcoin dispute resolution was basically with this nature. You have buyer or seller and an arbitrator which each of the key of multi seek address and easy transaction because it’s effective, both the buyer and the seller side and if there is a dispute, as arbitrator assign includes buyer or plus the seller depending of some arbitrator really.


So it was like the early idea of automating the enforcement of a dispute resolution process. Then I got also interested by E-commerce, business wise E-commerce, stuff like Open Bazaar. And I was looking at like, what is lacking in Open Bazaar, is clearly a dispute resolution system because it’s interesting to buy something to anyone on the internet but if I cannot trust him, although I know that I’m not just going to get robbed of my money.  And as a seller if I send a project before getting the money, the opposite is a buyer which can rob me. And just having one source philosophy trader was not a method which was acceptable because we need to both trust one person in common and we can we can do so but only if we have large institution that we both trust, like PayPal.


I trust a PayPal not to rob me and user policy, we’d also trust PayPal not to rob me. But since it was the institution, this company gets so big then they start to charge monopoly rent. PayPal charge them ready I-fee. Unless when there is a dispute, they even charge fee all the time. So it was not technically optimal. And so the question was, okay, all, can I make a decentralized system of disputes or to solve disputes and all can I make it scale? And that’s how I came up with the early idea of Kleros. It’s the only idea was to draw people once and then the second time, if there is an appeal to draw all the cost and then, okay, no, this is not going to scale because even if you have 10% of the dispute which end up in the appeal, it means that everyone has to review an amount of a dispute which is proportional to the amount of dispute, so this one scale.


So the question was, could we scale this kind of dispute resolution mechanism which were also, we should just invent like I think as most important previous work is trust coin which used shedding point to solve disputes. In this case it was dispute, Oracle dispute on prediction market. But they were asking everyone to vote on everything. And the question was, how can we make this scale to make a real decentralized dispute resolution system which will be general purpose, not only to solve Oracle of prediction market disputable but to solve every kind of dispute which can be enforced by small contracts.


Sebastien Couture:  So yeah, I mean, you mentioned some earlier projects like Open Bazaar and like decentralized E-commerce and all this also needs some sort of decentralized dispute resolution system. It was interesting in the early days to see how were they were solving that with multi-sig. And I mean just recently, I was confronted by this. I tried this software like a decentralized Bitcoin exchange called Bisque. And, you know, I just tested it out to basically just sort of experiment and see what the dispute resolution system was. And it actually work quite well, so I mean, but you do have to put some money escrow in order to secure your transaction.

So it kind of defeats the purpose or at least it makes a little bit more cumbersome to have to put up collateral in order to buy something.


And so having a dispute resolution system that, well, you don’t have to put up that collateral but there is some sort of a like a counterparty crowd that can effectively look into a dispute and arbitrate and make a decision is something I think that would be quite valuable moving forward as decentralized platforms become more prevalent.


Clement Lessage:  You only need to have a collateral because smart contract can only move that you like. ETH shall take it and whatever and as being explicitly given control of by someone to the smart contract. So we could say, okay, we make a dispute resolution system and we don’t ask collateral. But when there is a ruling, what is going to happen, nothing, because the smart contract will not have control over the asset of the dispute. So requiring a collateral is like a limitation of work in base dispute resolution system. But on the other hand, it means that when you have a ruling, you don’t need to have it enforced. It’s automatically enforced by smart contract. You don’t need to go, to have a route order, to have the police enforcing a ruling which can be ready, really, really more problematic than the need to deposit the collateral.


Sebastien Couture:  Yeah, what I meant by collateral was that as a as a buyer, I also had to put some collateral to secure the other side of the trade which was then unlocked once both sides agreed to the transaction. But of course if you’re doing something like web design, the buyer would have to put up the amount or at least some kind of a deposit in order for some collateral to be available if there was a dispute or something of that nature. So let’s dive right into it. So the project we’re talking about here today is Kleros. Give us a high level overview of the vision of Kleros and why you’re building this project?


Federico Ast:  We have people building this decentralized platforms now, some of them for E-commerce, for freelancing, others are building for insurance. So all of them are going to have disputes between users, right? In existing centralized platforms you have a customer service team that deals with the dispute between users. But we don’t have this in the decentralized world. So because these companies want to be fully decentralized, and so the vision of Kleros is to build up dispute solution system to which all of the decentralized companies of the ecosystem can plug. So they just send like the dispute, smart contract selects Kleros as arbitrator, agreed between the parties, of course.


And then when there is a dispute, Kleros selects a jury of people who are going to analyze the evidence and they are going to vote who is right about the dispute. So this can be used for example for a website dispute. I hired some guy from another country or whatever for building a website for me in a decentralized platform. We both agreed that in case a dispute happens between us, so Kleros is going to be the arbitrator. So if everything goes right, the transaction is signed and so the payment is done and I get my website. But if there is a dispute, so the money will stay locked into the into that kind of escrow and the evidence it’s going to go to Kleros.  And depending on how Kleros jurors vote, the money is going to go to one side or the other.


And of course all the process of selection of Kleros jurors is going to be fully automated based on smart contracts. So like nobody can tamper the evidence, nobody can put their friends as like jurors. So everything is fully transparent, fully, yeah it’s quite fast and cheaper alternative method. So I think that Kleros going to solve a very important problem that is at this point preventing the growth of the ecosystem. That is like the lack of some dispute resolution mechanism that is accepted by lots of people. Different projects started to build their own arbitration system because they sought that as important thing for their platform. But building an arbitration system is quite hard to do like a site project.


If you are building your website, so your platform for E-commerce, so you have to focus on your platform, you can’t just build arbitration as a site project because this is a very hard thing to do. And we spend all our time building this. And so we want to build the arbitration system for all of the decentralized ecosystems. So that’s like vision the Kleros.


Meher Roy:  You have already mentioned that one of the reasons for building a decentralized arbitration system is that other decentralized platforms need a decentralized arbitration system because they cannot really depend on the government courts in order to settle disputes like imagine if Open Bazaar needed to defend on the State of New York to settle its disputes. That would be a little weird, wouldn’t it? So therefore you need decentralized arbitration. Are there other advantages apart from this feature that you think justifies the case for decentralized arbitration? So you mentioned that you can make arbitration cheaper, right? Or you can make it faster. Give us a sense of why it could be cheaper or faster.


Federico Ast:  Why arbitration, why decentralization, right? I think that who explains this very well, so there is a very good article by Chris Dixon, the title I think it’s Why Decentralization Matters? Why should we do this decentralized? So of course one of the reasons is like, so this cannot be controlled like by us because this should be control by the community because this is going to deal with lots of very high value dispute. So this should be in hands of the community and also because in the long run, so the goal is not to have Kleros do everything but have people build platforms or like applications on top of Kleros technology. So one of them is going to build an application specifically targeting E-commerce, others are going to target like insurance, other finance.


If you are going to build something on top of a platform, so you want to be sure that this platform is going to be a decentralized one. So that you’re not going to be locked in so by a big like monster, like a big future Google or Facebook of arbitration. So in this article, Chris Dixon, he argues that, yeah, you have in all this centralized platforms, you have this typical curve where you have an early stage where the platform tries to bring users or like developers on top of it by giving subsidies. And then when it has like enough market power, it becomes like a monopoly and starts to charge monopolistic rents. So the point of building this decentralized way is that everybody that’s going to build on top of Kleros can know that they have a say in to the evolution of the platform and how it’s going to be working in the future.


So I’ve been thinking like the same reason why we build this on a theorem because it’s decentralized platform and not on top of like Google. So that’s an important feature of why Kleros needs to be built in a decentralized way in order to be the focal point, I’d say, of a future ecosystem.


Meher Roy: Of course the idea of crowdsourcing jurors is very interesting, right? So the idea with Kleros would be that if, let’s say, Federico and I, we are the buyer and seller and we enter into a dispute, then the dispute is going to be resolved by a juror or a set of jurors and these jurors have been crowdsourced by the Kleros Network. They have been picked out of a group of relatively pseudonymous or anonymous people. The underlying assumption in crowdsourced juries is that the scale to resolve disputes is widespread in society because like anyone can be part of this crowd that wants to be jurors. And therefore anyone can be selected to be one of the jurors that would arbitrate our dispute.


Do you think the scales for arbitration are widespread in society or do you think that this is a skill that is specialized and it needs its own specialized people or specialized community?


Clement Lessage:  So for now we are focusing on the low-hanging fruit. So dispute which do not require, I amount of knowledge to solve. So in the pilots in the example of created lists of that feature, you don’t need any kind of particular skill, also a skill that you need to learn is very easy to learn to be a juror. So I think processing, it’s really efficient when it’s tasks which do not require expert in this particular task. Due to crowd-sourcing that also mean that you can. If you have more argument for a dispute resolution, your kind of people refining the system, people editing the system in way smoother manner compared to a company or a state-based institution. And with crowd-sourcing, you can lower cost because you can have a draw which I’m not actually expert.


And they can design activity, they can walk before, okay, I have some, like one hour trainer dispend and I’m going to solve disputes in this one over in the trainer. If you were in a centralized company, you will have to hire someone, put in some office hour and you will have to be paid whether or not there is dispute. So it’s way less flexible. And so flexibility should decrease cost and also the fact that this is decentralized system and no one is the locked in it, means that you cannot charge monopoly rent as Federico explained before. So you choose this for reason. We can lower the cost of arbitration by you processing it for easy to resolve dispute. So easy to resolve is not necessarily linked to low value.


So for example, if you make some financial contracts and contract for difference on the price of the ETH, it’s something which can be also median into dispute. But the resolution of the dispute is quite easy. What was the price of ETH at this particular time? And no one requires specific skill to do so. So you have dispute which are good for purchasing, some like allow to purchase and follow, we will look at the one which are good and easy to purchase.


Sebastien Couture:  There’s kind of things that I’d like to unpack here. So one is, when I was reading your whitepaper, reading through, I think one of the examples was a website design. As a former person working in web technologies and having worked as UX designer and work clients and work with designers in the US, in Europe and also in Asia, one thing that we observe is that there are at least in this particular space, there are some cultural sensitivities that are inherent to each region or specific countries of this sort of thing. So this sort of brings up the point that the shelling point or focus point, in my opinion, for certain types of decisions may have cultural implications that would make it difficult to arbitrate. So I’ll give you an example, so take for example a European company that hires a designer somewhere in Asia, let’s just say Thailand.


And that designer produces a design for a website or a mobile app. And there is a dispute. So let’s say there’s a dispute about whether or not the design is something that would be attractive to customers. And if there could be a fault in the system where, for instance, if most of the jurors were coming from parts of Asia or where they would have similar sensitivities to a Thai designer and perhaps little experience or knowledge of some of the sensitivities of European customers where those jurors might favor, might arbitrate in a direction that would go against the practical reality which is that, you know, this design would not be attractive to European customers. So that brings us to the point is that shelling points, you may or may not have some cultural or local biases, we could also think of other examples, but what’s your thoughts on so the way that people would converge around the shelling point for things that may have some subjective or cultural biases?


Clement Lessage:  It’s not just asking whether this design is okay or not. So you have the contract, the plain mutual contract between like the European customer and the Thai designer. And you can also have specialized support. You can have support which are specialized by, it’s a kind of dispute but also by way to solve them. So you could have a support with some kind of like Asian sensitivity and someone with European sensitivity in terms of which customers for the design appeal to. And this can solve the problem where you have different culture which would give different answer. So in this case when you make your contract you decide that it’s with this particular support or it’s been this contract that you decide that the customer shall be happy or that the website should be appealing for a European customer target.


And this information have to be taken into account by the juror and about the support, you will only get juror which shows to arbitrate in a support with this kind of sensitivity. So that is a way to resolve cultural difference.


Federico Ast:  And there is another thing that we could think. So the fact of globalization of business is a reality. So there are contracts between Western people and Thai people. Some of these contracts are going to end in some dispute. So what other alternative for solution is there for this and is that alternative like non-bias in the way that supposedly the shelling point is. It’s like a deeper question that affects, of course, Kleros but any other system for arbitration between cultural sensitivities.


Sebastien Couture:  Right, okay. I see what you mean. This other point I wanted to bring up which occurred to me while reading the paper because I saw that you had a section there about, there was a sub-court which specializes in insurance or the insurance could be one area that this could be interesting for arbitration. And last year, I attended this conference in Munich which was specializing in insurance. And one of the things that I thought was really interesting about this industry is that there’s a major push in the insurance industry to arbitrate using machine learning and AI. So there’s an enormous amount of startups that are building machine learning systems that do things like look at pictures of a car and assess the damage or even listen to phone conversations and determine whether or not someone may be being dishonest about an insurance claim or some other matter.


So an insurance is one industry where machine learning is potentially disrupting a lot of people that are doing this kind of arbitration work. There might be other industries where this is the case but in terms of cost and in terms of efficiency of arbitration, what are your thoughts on the potential for machine learning and AI to really make arbitration much, much more efficient than any wisdom of the crowd-type knowledge could ever aspire to be?


Clement Lessage:  At some point you always need some human input because machine learning can be used to help you in your decision but not to make a decision instantly because machine learning algorithm are generally easier to trick than human. So we’ve seen some research on the reverse engineering, some machine learning algorithm and putting some input which make them make completely obscure decisions. So you gave example of a picture, of the car picture to assess the damages. If someone is able to reverse engineer the machine learning algorithm evaluating the damages, it will perhaps be able to do it by changing luminosity in few places of the car to completely make the machine learning algorithm give completely wrong decision. The same for the phone call, by using this particular tone, it’s going to be able to treat the machine learning algorithm into thinking that you are honest while you are not.


So I think machine learning already good decision helping, Couture, but you cannot rely on machine learning as to have the final say on problem which are on this side. So if you want to classify images and just to work on them is fine to use machine learning. But if you want to classify images where you have one per person which have an interest into the image to the A and someone who was interested of the image to be B, they’re going to try to bias this machine learning algorithms.


Meher Roy:  Very interesting. So we have sort of walked through or we have walked some of like these interesting design choices that Kleros has. Like for example crowdsource judgment. You have mentioned this idea of shelling point as the correct judgment in Sebastien’s questions but we have not actually walked through that in detail. Maybe this is a good time for us to go through how Kleros, the platform works or is supposed to work at the end state. So assuming that the dispute is between like, okay, Sebastien and myself, and one is a website developer and one is a company giving work. How would the Kleros platform settle our dispute and what would be our user experience be like?


Federico Ast:  Sure, so before the dispute starts, so you agree with Sebastien. So you’re going to build this website for me. These are the conditions and this is what I expect. And so we agree on that and then we agree, okay, so if something goes wrong, Kleros is going to be the dispute resolution mechanism and for this. And so you make the payment and so this stays into an escrow, smart contract and it says lock there and if there is a dispute, the money stays locked for and it stays into the contract. And you both have to pay an extra fee in order to pay for the arbitration, for the juries to be paid. And you have to make an extra deposit. And so now, this goes to dispute resolution and maybe Clement can explain better the system or the selection of the juries, how the token works and how all the crypto economics work for juries to make honest decision.


Clement Lessage: As a juror, you deposit your token in particular as a quote that you think that you have the skill to resolve the dispute of the support. And when there is a dispute, one token per juror is drawn. And if this token belongs to you, you are juror in this dispute. So you’re going to see evidence of both parties. You’re going to see quote policies, you’re going to see the plain English contract. And with all those information, you’re going to take a decision which show, again the binary like party A wins or party B wins or perhaps more complex, see if we have more party involved or more nuances decision possible. And if you watch it, the certain way as a final decision, you’re going to get rewarded with arbitration C and potentially some token of the parties which voted in opposition to the consensus.


So if it’s just a simple dispute with only one session where you have two jurors which say, party A shall win, one juror which says party B should win, and in this case party A wins the dispute. So the smart contract executes as enforcement. So in the a case of a website dispute and party A is a buyer, it will be reimbursed party A. And the two juror which were in the majority wins some arbitration fee and wins, some part of the token of the last juror which partied for party B. Obviously, you have a key mechanism if as a party in to dispute, you believe that ruling was wrong because the juror were mistaken or because the juror were in bribe, you can by PDC and appeal to get more juror to go through dispute again. And in this cases there is an appeal, it’s appeal reserve to be shared and are going to be taken into account by your smart contract for enforcement and by Kleros for token and arbitration C distribution.


The specific of the interface can be tailored by case. So if you are in with this website dispute, you may have to put the link to the website, the link to the source code, in to planning this contract, you would have specification. So the user interface is something which are going to depend on the kind of dispute you will be participating in.


Federico Ast:  Yeah, for example now that we launched the pilot called Doges on Trial, the jurors have to decide if an image is a Doges or not a Doges. So basically, in this case, the interface is like some screen where you have to see a picture and say, okay, vote. This is a Doges, not a Doges and that’s the interface, can be something as simple as that in this case.


Clement Lessage:  And the interface which displays this evidence, it’s not something which is part of the Kleros application. It’s something which is part of the arbitrary application. So we say that like, Doges on Trial, it’s part Doges on Trial, in the case of freelancer making website disputes, it will be part of the freelancing application. So people which are building or plugging into Kleros have a lot of liberty in all to display cases and evidence sort of jurors.


Meher Roy:  Of course like one of the interesting, so there are many different components to your system. So the ones that I could spot is there’s a component of jury selection, deciding who gets to be the jury, there is a component of depending on the case like gathering all of the evidence and like passing the judgment. There’s a component of economics which is that once you have a group of juries and they have put in their judgment, rewarding the juries based on how they followed the process or did not follow the process. So there’s a component of economic incentives. And the fourth component that you have in your system is basically the component of specialization of some kind, like so not all jury members are suitable for all sorts of cases. There might be some jury members that specialize in insurance, others in website disputes and things like that.


So there’s something that Kleros does for, in order to have duties also specialized. So like Kleros is a pretty complex system which has all of these inter moving parts that like connect together to make dispute resolution system. Specifically, I would like to deep dive into the jury selection. Okay, how does jury selection exactly work? So when Sebastien and I are disputing, how do you decide who gets to be the jury and how big the jury would be in this particular dispute?


Clement Lessage:  So as a juror, you choose a pass in the support tree. So this means that you can say, okay, I want to be a juror in general cases, in E-commerce cases and in website freelancing cases. So you specialize and specialize as you go deeper in the in the tree of juror, which would be in this case the freelancing, website freelancing support and the apparent one is the E-commerce support, and the apparent of the E-commerce report is a general one. So this means that you can be drawn in those three different quotes. When you are registered to be a juror, you have to deposit tokens. So when you deposit PNK into schools and your friends to be drawn as a juror is going to be proportional to the amount of token you deposited. So it mean if you put more token, you are more likely to be drawn. Obviously, if you put a number of token which are really, really high, you’re going to be drawn too much and you may not have time to solve these other disputes.


So even if you are really rich in token, you may not have this much of time to solve the disputes. And when there is a dispute, the arbitrary contract decides almost zero without basis. So in the other trial, we start with three jurors but this is a barometer. You can say, I’m dealing with this dispute already of really, really low value and I don’t want to pay that much in arbitration Christ, so I want to start with only one juror. Or you can say, okay, this dispute as quite high value and I don’t want to lose a lot of time in appeal because I know that small jury are more likely to get drawn easier to bribed and be won. So I have a dispute about the price of ETH for financial contract both on median. So I’m going to start with 100 juror in order to not to lose that much time in the resolution of my dispute. So, yeah, number of juror is something that’s decided by arbitral contracts.


Federico Ast: The main insight about the selection process is that, first jurors self-select themselves into the court where they think they can arbitrate. So if I am an expert into websites, I’m going to. So I have an incentive to self-select to arbitrate disputes into websites. If I self-select to insurance where I don’t know anything about that or if I select myself into a court and then I don’t really take the time to make the vote, I just vote randomly, so I am going to lose money because of how the system is structured, the tokens that I deposit to be drawn, so since I’m not arbitrating correctly or honestly, I’m going to lose them. So in the long run, the system is structured for those guys who try to abuse of it are going to lose money and eventually leave. And maybe it’s interesting also to mention that this is loosely inspiring how courts worked in ancient Greece.


So the Athenians of the Classical Period, they have this interesting concept that every citizen of the police have their right to judge. But they didn’t want for anyone to go to the like the Agora and become like a mob of justice kind of stuff. So they developed this very sophisticated mechanism for selection that was based on token that every citizen had, like it was an ID token that you had like your ID as a citizen and you went on the court day and you put your token into this machine like a two-meter high machine with slots called Klerotherion that means the altar of randomness, so it’s like the place where randomness happens. And this was an allotment machine that selected. So some guys threw some white and black dice on a cube affixed on the side of the machine. And if you had on your row a white dice, you went to the jury, if you have a black dice you went home.


So that dice, that 20 faces dice we call Icosahedron, that’s one of the Platonic solids, it’s actually Kleros logo. So Kleros system is like loosely inspired, it’s something that was tested before like 2500 years into the past. And so that’s where the name comes from also.


Meher Roy:  One of the differences between the Athenian system and Kleros that I can see is that in the Athenian system, when each citizen gets a token, it’s one citizen one token and then the tokens are used to select the jury. Whereas in Kleros the token is the is the coin, Pinakian coin, and the holdings of Pinakian coin can be different between people. So somebody might have very little Pinakian. And if you look at like how the other tokens have panned out, majority of coins are actually held by the crypto hedge funds. Now you see a system like, there’s also definitely, I don’t know like 10% of the supply is owned by poly chain because like poly chain is the biggest crypto hedge fund. And there is no way around such an unfair distribution. So in Kleros, my odds of being selected as a juror proportional to the number of tokens I have right now.


So somebody that has a lot of tokens in a particular code, let’s say I own 5% of the supply and I go to the website design code, there’s very high odds that I’m going to be selected as juror repeatedly even though I might not be the best juror. So do you actually feel that like with this incentive structure, the jury will end up captured by all the crypto hedge funds because crypto hedge funds might end up owning 30 or 40% of your supply pretty easily. And so the crypto hedge fund managers end up becoming the jurors. Do you see that as a risk to your selection process?


Clement Lessage:  Yeah, I always kept my amount of words it can provide. So if you get 20% of the token, you can deposit all of them and get drawn 20% of the cases. But ruled as a crypto hedge fund, do you have the time to solve all those cases? Probably not. So it’s all you can ask hold them in a sloppy manner and be incurring a lot of time and lose money or you’re going to need to best the admirable people to solve disputes instead of you. So no, it start to look like more like putting than someone having 20% of the vote. In the Ancient Greece model, you have had different is that you can easily get identity of people because you get to see them while in the crypto world, if I split my balance into two accounts, no I’m two person and you have no way to know if I’m two person or if it’s just me which is to be that I came.


So you cannot base the chance of being drawn on being someone because you can be lot of people online. You can have lots of identities at the CD attack. Now, because of the non-one person one chance to be drawn, you can see, okay, this means that there was a problem as a reason kind of democratic problem. And the democratic problem is solved in a different manner or it could be solved like the Ancient Greece and kind of like state system, because of the possibility of working. If you’re in Ancient Greece and there is a dispute, you have no choice. You’re going to be brought to this Ancient Greece quote and you have no say in who is going to rule on your case. While on the centralized system, on the version world, everything is optional. So the democratic choice is not made at the juror level but it’s made as a party level in the dispute which you can say, okay I want the system to arbitrate my dispute and I don’t want this one because I think this one is a corrupted system and I don’t, if you only accept this system as a dispute resolution mechanism, I don’t want to make business with you.


So that’s where are the decisions, the democratic decision comes from, from parties in dispute and not from as a juror side. Juror are more likely, it look more like a dial, like a dispute resolution comparative and they need to be looked as — it does not look like a state based institution at all.


Federico Ast:  So we use this as inspiration because it’s easier to understand. But so, it’s not that we are like just replicating the Ancient Greece institutions into the crypto world. So it is loosely inspired in the sense that this bigger solution uses from this crowdsource model. So the Greek have this very interesting insight about every citizen has their right to be a judge. And Kleros is based on this peer-to-peer philosophy of everyone who wants to participate in the arbitration process as a juror can do so under these conditions that are defined by the crypto economic system. But it’s not — so of course there are many, many differences between Kleros and the Ancient Greek system.


Sebastien Couture:  So in the whitepaper, there’s an interesting section that I want to ask you about and so I’ll have the link to the whitepaper in the show notes and anybody looking to understand at a deeper level how close works under the hood should definitely read it. And that was a section about random number generation, the process of selecting jurors implicates some randomness. And there there’s a mechanism within Kleros to generate random numbers in a decentralized way. Could you explain how that works?


Clement Lessage:  So in the current release and I used some blog hash but in the next release we will use a sequence chart proof of work. So the idea of sequence full of walk is that you start with a value that everyone can modify but no one can fix to a particular value. So this means that you may start like zero, we should broadcast or whatever and then anyone can give local random number to make this value change, but you cannot fix it to a particular value. Then what people have to do is to apply sequence chart full of work on it. So they have to hash it repeatedly, some amount of time which are going to take computation on time. So you don’t know the result instantly, you have to dedicate one CPU to solve this result.


And by the time that someone solve this computational problem which is a concern, so this means that you cannot put more computer into solving it, it’s not possible anymore to modify the scene. So it’s to determine the process but it’s to determine the process where anyone can change a point where we start from. And the time to know the answer is long enough to avoid anyone from knowing sense, we know the answer to be able to change the seeder. So you can know what the number is going to be but when you know it is already too late. And that’s a way to generate randomness and really basic trust assumption which is just you have at least one honest party which want the number to be random. So basically it was saying there is no one honest, you just have to do it yourself.


And so since you know that’s why you’re self-honest, you know that the member is going to be random.


Sebastien Couture:  The whitepaper also makes mention your governance mechanism, could you like lay out governance mechanism? What it’s used for specifically and how it works?


Clement Lessage: It’s used for creating some quotes, specialized quotes and setting parameters. So for example arbitration fee and the amount of token you can win or lose in disputes, if you are current or not. It’s a voting system but it’s not a classical one, it’s a liquid voting system, which means that everyone has the right to vote in proportion to the tokens they owned. But it’s a right, it’s not an obligation or something that’s expected from them because they vote or they can delegate their vote to someone else. And this someone else can also delegate his vote to someone else. So you can have transition delegation, so A delegates to B, B delegates to C, C delegates to D and then D votes yes.


So this is like if A voted yes too, because we wanted two characteristics of the bottom system. The first one is that we want everyone to have the right to influence every decision. So you need to let everyone the possibility to vote on everything if they want to. And the second one is that you don’t want the system to be wasteful, you don’t want to waste people’s time. So you allow people to delegate when it’s a decision that they are not specifically interested in. So basically you’re getting both of the advantage of a direct voting and election.


Sebastien Couture:  Let’s presume that I want to make a new sub court about the quality of podcast editing, something like that. What would that look like for me as a user? How would I propose this change and then what were the process of voting that new sub court look like?


Clement Lessage:  So you would propose to create the sub court. You will decide what is apparent of this court. So you may have like apparent quote which is a media court for example. So you will say that this court is a child of the media court. You will write basic policies of the court, you will determine almost visual of the court after we paid, will determine how much tokens they need to put at least and which percentage of those can be lost if a they solve dispute in an increment manner. Then it will be put for review by anyone which can claim that your proposal is invalid. So if you’re just like, I don’t know, like write or will be showing some people in your proposal, someone who challenges proposal as an invalid proposal, and it would go to dispute resolution.


And so if no one challenges proposal, everyone can express interest in your proposal, still in a liquid voting manner. And if your proposal gets more than the interest forum, it’s put to vote and anyone can vote for it and people which did not vote for it, their vote couldn’t as if they voted the same way as a delegate or the delegate also delegated, et cetera. If they delegate also it was. But if everyone gives a vote end up into a yes, obviously, the court is created and now people with dispute can specify that just mark on tried are going to be arbitrated by this party also court.


Federico Ast:  Actually, so the podcast quality sub court could be quite interesting idea for implementing for a created list use case, because Kleros can be used to arbitrate disputes in created list. So imagine you can upload, you submit your podcast to the list and some user can say, yeah this podcast does not have enough quality to be on this list, so now there is a dispute between you and the user and this could be adjudicated in Kleros. So even if we usually focus on the use cases of escrow disputes, because they are easier to understand, so Kleros has many applications. One of them is to solve disputing in created list and them like the experiment we are running now, Doges on Trial is based on a created list use case of Kleros.


Sebastien Couture:  There seems to be kind of an intersection here with, once you get in these types of use cases, they sort of start to overlap with prediction markets in some sense. Can you talk about that and maybe explore how arbitration can merge into prediction markets and vice versa?


Clement Lessage:  So prediction markets needs dispute resolution. So prediction market, this Oracle dispute is, you ask question about the state of the outside world, which is a dispute resolution use case. And I treat like the project which is most similar to Kleros is probably Ogier. It’s not a dispute resolution project but in the projects they have the problem of the Oracle problem to solve the prediction market. So they came up with mechanism which have some kind of similarities with Kleros. It’s not the same because they don’t draw people randomly, instead they ask people to stay on dispute that they want, so everyone can basically self-select into a dispute. And we don’t think it’s optimum because it means that if you are malicious, you can target your attack on the particular dispute while if we use a random draw, you can try to attack but then you cannot really focus your attack on something.


But it’s also something which is Ogier is, basically you have to put a stake in the dispute and if you’re your side gets the biggest stake for some period of time, you win the dispute. And anyone can put the stake in the other side and they have to put the stake in the other side so the stake of the other side is at least twice the stake on the previous side. So if the question is, has it rained today? You might put random web token on yes and then someone say no, I disagree, he has to rep token on two, and then if you have another one, it’s raining, yes, you have to put three additional rep tokens and such. You have four on the yes and if no one does anything during one day or one week, I don’t know the timeline, the winning side wins the token of the other one. So in this case yes side has four tokens, the no side has two tokens, so the yes side wins the two token of the no side, and yes is considered to be the result of the dispute.


So this also look like a bit actual in points because if you put your token as majority, you get rewarded, if you put them like in the minority, you get denied. And with that they also have some forcing mechanism which starts automatically, that we also have in Kleros. But in Kleros, it does not start automatically, someone has to say, I think that the chart level of the court is too corrupted, so I won’t follow the system. But yes, there are lots of similarities between the Kleros and Ogier, yeah.


Meher Roy:  Very interesting. Of course, like if you look at like current legal systems for nations, you see that like many times disputes emerge because contracts are not drafted correctly. Like the contracts under specify what should happen and the courts need to add interpretation on what the contract parties meant when they were writing the contract, like sometimes disputes are clear. It’s a yes and no question. But many other times like disputes are more complex, the contract parties did not foresee the particular situation happening, a particular situation happened that is not specified in the contract, dispute goes to court, now the court must interpret what the parties meant when they sign the contract.


And so to do things like this, the courts have different ways of handling such things, right, so the court forces interpretation on the contracts based on the laws of that particular jurisdiction. So like many of the laws are needed because when dispute arises and the contracts are not specified well enough, you need a legal system to give meaning to those contracts and still be able to handle these disputes. Is there an element like that to Kleros where even an under specified contract could be arbitrated?


I think that this is the point where Kleros is really, really useful because if you have a contract that is perfectly complete and perfectly clear, so you don’t need Kleros because that’s called enforcement like yeah, was this payment done or not, was done, everything enforces right. So there is some principle of philosophy of rights as you say that every contract is incomplete because they cannot foresee every situation that could arise in the future. So there is some kind of subjectivity in should this contract go to A or to B? Who is right about this website thing? The website dispute is not completely subjective because there was a contract stating how many pages their website should have, what timeframe should the website be delivered on or et cetera, but it is not completely objective either because like the website can be done in different ways and many of these ways can be correct.


So the world we are building of smart contract and quote enforcement still has not solved this space of subjectivity that is really important for a legal system. And this is where Kleros and where humans can do a very important contribution because humans are better qualified that probably AI for this kind of subjectivity questions. And about the question so, so how should the contract be interpreted at which jurisdiction? So now you have in ordinary contract, so maybe in some financial control, it say that if there is a dispute, this is going to be settled in New York arbitration court. So now, imagine you and I do a contract about the website and we agree at the beginning, so if there is a dispute, it’s going to be solved into a sub court of website with this characteristic sub court.


So the jurors who are going to participate into that court, they know that the policy of the court says, so contract should be interpreted in this way, in this way, this way and so when they do the decision about who is right, they’re going to use those rules as their guidance for the decision. So that’s the jurisdiction where the contract is going to be settled.


Sebastien Couture:  Interesting, so as we wrap up here, let’s maybe talk about the current state of the project. Can you tell us where were things at right now and what’s the roadmap looking forward?


Federico Ast:  So well, about the current state of the project, we did our token sale between May 15 and July 15. We use a new method of tokens they call Interactive Coin Offering, that was suggested by Vitalik and Jason Toth for making distribution more fair between participants. So this end July 15th, then on July 30, I think or 31, we launched our project and our product into the entire illuminate. This is a experiment called Doges on Trial where users or jurors have to arbitrate if an image is a Doges or not a Doges and there is a lot of different mechanism for incentivizing people and to try to see how resistant the Kleros is to different type of attacks. So, people we invited you to come to the experiment and submit images and try to fool the system. If you can fool the system by making a cut into the list, so you are going to get to it.


So it’s a good incentive for you to try to game Kleros jurors. And so now we are focusing on that and all the learning that we expect to have from those submissions on this experiment. And then of course trying to go into the main, so in the real world use cases because we already have a number of partnerships for example with Ink Protocol to start using Kleros into E-commerce, a website E-commerce disputes like in the Ink Protocol, they have this dispute. We have other partnerships with the crypto currency exchange called Diether. So the next step for us is try to prove that Kleros can actually work in real-world use cases, can actually solve the problem that we expected to solve.


So now we are recruiting people, we are completing the team and trying to make this very, very exceptional team to tackle this very, very big problem.


Clement Lessage:  So on the side of the room and so we have obviously the best version of Kleros which is with, we thought support but with the dispute resolution mechanism, the random drawing mechanism, the token redistribution mechanism, which can be used already by project which we’ll plug on it. But for now we just have this — it’s a funny game that Federico explain to test the system, where we want to test the system in adversarial condition, where we give people incentive to try to break it. Next version will have support, we will have some improvement of user interface. We would have the governance mechanism because for now the only point which prevent the system to be really decentralized is the lack of the governance mechanism.


And on like election poll, we have improvement of Kleros and the integration with all the project in those use cases.


Meher Roy:  Cool, so the place at which people can actually use your system is with Doges on Trial, right? So that’s the public facing experiment you are doing that all listeners can participate in right now.


Clement Lessage:  Yeah, you can already use or you can already Doges on Trial. You can already use 10K to become a juror on Doges on Trial. And we got something like almost 200 submission for the trial.


Meher Roy:  Okay, that’s really cool. I’m actually curious what your first focus area is going to be? Are you going to like focus on some kind of cases at firsts specifically or will you try to build a general court first?


Clement Lessage:  So we already built a general court, it is that we have a general court which now is only plugged to one application which is Doges on Trial. But if you want to plug another application now, technically you can do it. It is that since the system is still in the early day like it has been launch less than a month before, we advise to wait perhaps one or two months after the launch to be sure that on those test case it’s work as expected before putting real use case with money at stake. But if you see on the system like operates for two months straight without any problem, I think it will be ready to put real dispute, yeah.


Meher Roy:  Very interesting. Thank You, Federico and Clement for joining us today. It was great to discover Kleros. It’s a super interesting project, one of the leading projects doing justice systems on the blockchain. And we look forward to having you back, let’s say, a year or two later when your system is further along and actually processing commercially useful cases.

0:00:00 | -:--:--

Subcribe to the podcast

New episodes every Tuesday