Kleros, a Protocol for a Decentralized Justice System
Building a Judicial System for the Internet Age
ABSTRACT
In this paper, we introduce Kleros, a decision-making protocol which uses blockchain and crowdsourcing for adjudicating claims. The system relies on incentives backed by the game theoretical concept of Schelling point in order to motivate anonymous jurors to rule cases correctly. Kleros means “chance” in Greek. The system is based in some principles for random jury selection which were known by the Athenians twenty five centuries ago. The result is a peer-to-peer justice system able to solve the disputes of the Internet age in a fast, transparent and inexpensive way.
This is a long conceptual paper. For an abridged version, see here. For a short technical white paper, see here.
By Federico Ast and Clément Lesaege
“Whoever controls the courts, controls the state”. Aristotle.
The world is experiencing an accelerated pace of globalization and digitalization. A growing number of transactions are conducted online between people from all over the world. It is estimated that disputes arise in 3 to 5% of online transactions, totaling over seven hundred million in 2015 alone (Katsh and Rabinovich-Einy 2017: p. 67).
Buyers in eBay claim that sellers failed to send the goods as specified in the agreement, guests in Airbnb claim that the rented house was not “as shown in the pictures” and backers in crowdfunding claim a refund as teams fail to deliver the promised results. As a larger part of commerce and social interactions move online, so do conflicts.
Existing dispute resolution technologies are too slow, too expensive and too unreliable for an online real-time world. A fast, inexpensive, transparent and decentralized claim adjudication system will be a key institution for the Internet age.
In the article The Crowdjury, a Crowdsourced Judicial System for the Collaboration Era (2015), Federico Ast, one of Kleros founders, and Alejandro Sewrjugin (1) presented some ideas on how a decentralized court system could work. Crowdsourcing, they argued, could enable faster, cheaper and more transparent judicial decisions. The key insight was that, as Ancient Athens courts, a judicial system could work on a p2p basis.
The article laid out the big picture, but did not dig into details of how this technology could actually work. Ethereum was still in its early days. Protocols and decentralized applications had still a long way to go.
Fulfilling the vision of Crowdjury requires a number of building blocks. First, a protocol for selecting and incentivizing anonymous jurors to analyze evidence and vote decisions honestly. Second, a protocol for selecting and incentivizing jurors to adjucate claims when anonymity cannot be used. Third, a standard for writing contracts that can be adjudicated by the system. Fourth, a framework for mediation and restorative justice for restoring trust between parties and reverting loss of reputation.
This paper will focus on Kleros, an Ethereum decentralized organization that uses game theoretical incentives for anonymous jurors to adjudicate claims in a fast and transparent way.
What the Greek Knew About Justice
A court proceeding is an epistemic engine, a tool for ferreting out the truth from a confusing array of clues and indicators. There is a procedure where an agent (jury) uses some input (evidence) to produce an output (verdict). The probability (P) of reaching a true verdict will depend on these variables:
P (VERDICT IS TRUE) = F (JURY, EVIDENCE, PROCEDURE)
A true decision occurs if the tribunal votes A when A is truly the case and votes not A when not A is truly the case. An error of justice occurs if the tribunal votes A when not A is truly the case or votes not A when A is truly the case.
Ability to reach true decisions is the critical feature of a court system. Courts that frequently convicted innocents and acquitted the guilty would fail to win the respect from those it governed (2). However, other criteria are also to be considered, such as the cost and duration of proceedings. Reaching a true decision at a low cost is preferable to reaching a true decision at a high cost. Reaching a true decision sooner is preferable to reaching a true decision later.
All this considered, the key problem to solve is:
How to design a system that produces true decisions at the lowest possible cost and in the shortest amount of time?
In order to answer this question, we may find some inspiration in the Ancient Greek. The Athenians developed a remarkable judiciary. Trials were conducted by a large body of ordinary citizens. On trial days, citizens wishing to serve as jurors went to the courthouse where a sophisticated jury selection procedure took place.
Citizens inserted their pinakion (a bronze or wood ID token) into a slot of an allotment machine called kleroterion. After all citizens had inserted their pinakion, an official threw white and black icosahedron (20 faces) dice in a tube affixed vertically on a side of the kleroterion. The tube stopped at the bottom and held the dice in the random order in which they entered. Dice were then released one at a time. Candidates having a black dice on their row were dismissed for the day. Candidates with a white dice were drawn for paid jury duty (Boegehold 1995).
About 25 centuries ago, Athenians knew that courts could work as a peer-to-peer system without turning into mob justice, provided three conditions were met: jury duty was voluntary; jury duty was paid (3); jury selection was done by sortition (kleros) (4). Greek popular courts had no continuing existence or permanent personnel. They were decentralized organizations where jurors and officers changed from day to day following predefined rules. The pinakion, the kleroterion and the icosahedron dice were extraordinary pieces of civic technology that supported a basic concept of the Athenian polis: ordinary people had a right to judge.
Athenians trials from the classical period greatly contrasted with the legal systems that nation states started to develop in the 17th and 18th centuries. Instead of ordinary citizens, most modern judiciaries are based on professional judges and attorneys (5).
In late 20th century, following the advent of the Internet, national legal infrastructures became unable to cope with the complexity of a networked age. Limitations became clear, for example, when Google acquired YouTube in 2007 and had to comply laws on privacy, intellectual property and defamation in over one hundred countries.
The transition to a blockchain decentralized Internet is likely to make things even more complicated:
“An ownerless network upsets many of the regulatory and legal frameworks now in place for our communication infrastructure. Clouds don’t have a lot of geography. Whose laws will prevail? The laws of your domicile, the laws of your server’s domicile, or the laws of international exchange? Who gets your taxes if all the work is being done in the cloud? Who owns the data, you or the cloud? If all your email and voice calls go through the cloud, who is responsible for what it says?” (Kelly 2016: p. 130).
A new, global and real time economic reality requires a new, global and real time adjudication technology.
A simple website dispute
Alice is an entrepreneur based in France. She contracts Bob, a programmer from Guatemala, on a freelancing platform to build a new website for her company. After they agree on a price, terms and conditions, Bob gets to work. A couple of weeks later, he delivers the product. But Alice is not satisfied. She claims that the quality of Bob’s work is considerably lower than expected. Bob replies that he just did what was agreed. Alice is frustrated. She cannot hire a lawyer for a claim of just a couple hundred dollars with someone who is halfway around the world.
Now, imagine that, at the moment of their agreement, Alice and Bob had picked a standard off the shelf smart contract for website design freelancing (6). The contract had a clause stating that, should a dispute arise, it would be arbitrated in the Kleros network. After Bob stops answering her email, Alice taps a button that says “Send to arbitration” and fills a simple form explaining her claim.
Chief is a software developer who lives thousands of miles away in Nairobi, Kenya. While on the bus commuting to work, he is checking Kleros website to find some arbitration work. He makes a couple thousand dollars a year on the side of his primary job by serving as a juror in software development disputes between freelancers and their clients. He usually works in the Website Quality subcourt, which requires skills in html, javascript and web design. Chief activates 2 pinakion, the token used in Kleros to select jurors for disputes. The more tokens he activates, the more likely it is that he will be drawn as juror.
About an hour later, an email hits Chief’s inbox:
“You have been selected as a juror for a website quality dispute. You can access the evidence here. You have three days to analyze the evidence and submit your decision”.
Similar email are received by Benito, a programmer from Cusco and Alexandru, from Romania, who had also activated their pinakion for the dispute. They were randomly selected from a pool of almost 3,000 candidates. They will never know each other, but they will collaborate in settling the dispute between Alice and Bob. On the bus back home, Chief analyzes the evidence and votes who is right.
Two days later, after all jurors have voted, Alice receives an email:
Bob receives an email saying he lost and that he has to pay. Jurors are rewarded for their work and the case is closed.
The technology behind Kleros
Crowdsourcing: pulling knowledge from the crowd
Adjudication is, essentially, a task of information discovery and analysis. The exponential decrease of computing and communication costs have enabled this type of work to be done by crowds in a massive and global scale (Brabham 2013).
Attempts were made to crowdsource dispute resolution. The Internet Corporation for Assigned Names and Numbers (ICANN) established a nonbinding arbitration system with crowdsourced arbitrators for trademark claims. An early pilot was done in 1999 to adjudicate disputes between buyers and sellers on eBay. In 2008, a community court scheme was built, where sellers disputing negative buyer feedback could submit their complaint to a randomly selected tribunal.
Chinese e-commerce platform Alibaba adopted a similar method for dispute resolution. About 99% of the disputes arising in the platform are solved by a user-based jury system. In March 2016, 920,000 active jury members rendered 150 million votes. Jurors are rewarded with positive reputation which Alibaba uses for computing user credit scores (Katsh and Rabinovich-Einy 2017: p. 66).
Blockchain: enabling transparency
Transparency is a key feature of a justice system. Users need to trust that evidence is not tampered with and that the selection of jurors is not biased. Blockchains, open and distributed ledgers that can record transactions in a verifiable and permanent way, can fulfill the transparency requirements.
First, the blockchain may provide a cryptographic proof that the evidence was not tampered with in any of the instances of the adjudication procedure. Second, Ethereum decentralized autonomous organizations can guarantee that all the processes (handling and securing evidence, selecting and rewarding jurors, etc.) are fully automated and incorruptible. These features of blockchains guarantee that Kleros will be an impartial and incorruptible third party for adjudicating claims. (6)
Schelling Coin: designing incentives for the truth
“You are to meet somebody in New York City. You have not been instructed where to meet; you have no prior understanding with the person on where to meet; and you cannot communicate with each other. You are simply told that you will have to guess where to meet and that he is being told the same thing and that you will just have to try to make your guesses coincide” (Schelling 1960: p. 56).
While any place and time in the city could be a solution, game theorist Thomas Schelling found that the most frequent answer people would give is: “noon at the information booth at Grand Central Terminal”. Nothing makes noon at Grand Central Terminal a location with a higher payoff (any other place and time would be good, provided that both agents coordinate there), but its tradition as a meeting place makes it a natural focal point. For each agent, it is the expectation of what the other expects him to expect to be expected to do.
Game theorist Thomas Schelling called focal point a solution that people tend to use to coordinate their behavior in the absence of communication. Schelling points typically arise when communication is impossible, but also when, while communication is possible, parties do not trust each other. Based on the concept of Schelling Points, Vitalik Buterin has proposed the creation of the Schelling Coin, a token that aligns telling the truth with economic incentives. Paul Sztorc developed the concept of TruthCoin based on the Schelling point principles.
If we wanted to know if it rained in Paris this morning, we could ask every owner of a Schelling Coin: “Has it rained in Paris this morning? Yes or No”. Coin holders vote by secret ballot. After they have all voted, results are revealed. Parties who voted as the majority are rewarded with 10% of their coins. Parties who voted differently from the majority lose 10% of their coins.
The Schelling Coin mechanism provides incentives to agents to tell the truth. Agents are expected to vote the true answer because they expect others to vote the true answer, because they expect others to vote the true answer… In this simple case, the Schelling Point is honesty.
Schelling Coin mechanisms have been used for decentralized oracles and prediction markets such as Truthcoin, Augur and Gnosis. The key insight is that voting coherently with others is a desirable behavior and has to be incentivized. The incentive design underlying Kleros is based on a mechanism similar to the Schelling Coin, slightly modified in order to answer to challenges regarding scalability, subjectivity and privacy to encourage agents to engage in optimal behavior.
The Kleros Adjudication Process
The Kleros decision procedure is made of the following elements: 1) Contract, 2) Securing Evidence, 3) Jury Selection, 4) Analysis, 5) Voting, 6) Appeal, 7) Token Redistribution.
1) The Contract
Kleros is a voluntary opt-in system. In order to use it, the contract between the parties needs to have a clause stating that, should a dispute arise, it will be adjudicated in Kleros. When they started working together, Alice and Bob picked a freelancing contract template and modified a number of parameters to represent the specifics of their agreement (complexity of the product, delivery date, etc.). They also agreed that they would both pay the arbitration fee in case of dispute.
Parties don’t need to know how to code in order to use one of these contracts. They don’t even have to know what smart contracts are. From the point of view of users, signing the contract is just using a form to state the agreed upon terms in natural language. Projects already working in the development of such templates include Aragon, OpenZeppelin and OpenLaw. As the industry matures, more templates will be developed for contracts in e-commerce, finance, gaming, travel and so on.
Subcourt Selection: Where will the dispute be adjudicated?
In traditional justice systems, courts are specialized in criminal, economic, civil law, etc. Kleros subcourts are also specialized in particular kinds of disputes such as freelancing, finance, consumer protection, online gaming, etc., where each one has some specific features regarding policies, session time and cost, among other parameters (This will be explained below in the section Subcourt System)
Contracts between parties will specify the subcourt where a dispute will be adjudicated, including the number of jurors that will be drawn for the first instance decision. The contract between Alice and Bob stated that the dispute would be solved in the E-Commerce/Freelancing subcourt using three jurors. As soon as the dispute arises, it is sent to the subcourt specified in the contract.
Arbitration Fees: Who will pay the cost of arbitration?
Jurors lend their time and expertise for analyzing evidence and voting. They offer a valuable service and will be compensated for every case they solve. Different subcourts will have different arbitration fees, depending on the complexity of the disputes and the scarcity of juror skills. As will be explained below, this will be decided by users through the governance mechanism.
Proceedings will only start after a deposit is made to cover jurors fee. From the point of view of Kleros, it is irrelevant which party makes the deposit. It could be split equally between both parties, paid only by one party, paid by a third party with an insurance mechanism or by any other scheme. The only thing that matters is that enough money is available to compensate jurors. As long as arbitration fees are paid, different models could work.
2) Securing Evidence
Proceedings begin when at least one of the parties believes there was a breach of contract. When the party decides to send the case to arbitration, the contract in plain English (or the natural language chosen) and all relevant pieces of evidence are sent to Kleros secured by public key cryptography.
Relevant evidence will depend on the type of dispute at hand. In the Alice vs. Bob website suit, it may consist of the natural language agreement and the digital files delivered as a product. In an online gaming dispute where one party claims that the other has cheated, evidence could include a recording of the game. In a car crash insurance dispute, it could include the insurance contract and photos of the crashed car.
3) Jury Selection
Jury selection relies on two basic elements: candidate self-selection and sortition. In order to avoid retaliation and intimidation, jurors are not required to provide proof of identity to be drawn (6). The key challenge is: how to create the right incentives for anonymous jurors to adjudicate claims in an honest way? This problem will be solved by using a system token and a random selection mechanism.
System Token: the Pinakion
Users have an economic incentive for serving as jurors: collecting arbitration fees. Anyone can self-select as candidate to be a juror in specific subcourt by using a reputation token called pinakion (PNK). Users must deposit PNK in order to have the possibility of being drawn as jurors (7). The probability to be drawn as juror is proportional to the amount of tokens a user deposits in a subcourt. The higher the amount of tokens a user activates, the higher the probability he will be drawn as juror.
Chief, the Kenyan software developer, often works in the E-commerce/Freelancing subcourt. This time, he activated 2 PNK. Many other users from around the world (among them, Benito and Alexandru) have done the same. When activating their tokens, users start in the General Court and follow a path to a specific subcourt according to their skills. For example, Chief activates his pinakion in the General Court and in the E-Commerce/Freelancing subcourt. Clément, an insurance expert, activates his tokens in the General Court and in the Insurance Subcourt. This will allow him to be drawn for insurance disputes.
Jury Selection: Sortition
Jury selection is done randomly among all the users that activated their pinakion in a subcourt. Theoretically a candidate may be drawn more than once for a specific dispute. The amount of times a user is drawn for a dispute (called its weight) will define the amount of votes he will get in the dispute and the amount of tokens he will win or lose as a result of his vote. Imagine that 6 token owners signed up for the dispute and activated 10,000 in total with the following distribution:
For a dispute that requires 5 votes, 5 tokens are drawn out of the 10,000 that were activated. The drawn tokens are number 2519, 4953, 2264, 3342 and 9531. The token owners B, C and F are drawn with a weight of 1. The token owner D is drawn with a weight of 2. Activated pinakion will be frozen during the court session and will be unfrozen after the court has reached a verdict.
4) Analysis
Users that are drawn as jurors will have access to the evidence for analysis and voting a decision. The specifics of this procedure may change in different subcourts. Subcourts will have different parameters regarding issues such as the time jurors have to reach a decision, the complexity of the voting options and the possibility of communicating with the parties (See below in the section Subcourt Policies).
A simple dispute could involve only two parties and only two options. In the Alice vs. Bob dispute, the decision could be: “Who is right in the dispute? Alice or Bob?”. The winner would receive the full payment and the loser would get nothing. A slightly more complex option would be: “From 0 to 100, how guilty is Bob?”. This would allow to split the payment in different parts.
At is early stage of the project, Kleros will work mostly for simple disputes involving few parties and voting options. Over time, as technology improves, more complex disputes will be adjudicated. For example, cases involving multiple parties, multiple issues and multiple voting structures. Imagine Huey, Dewey and Louie are claiming Mr. Scrooge’s estate, consisting of cash, a mansion, a yacht and a luxury car. The jury could have the following voting options: “Pay Huey X% of the cash amount; Pay Dewey Y% of the cash amount; Pay Louie Z% of the cash amount (where X+Y+Z = 100). Transfer Huey the property of the mansion; transfer Dewey the property of the yacht; transfer Louie the property of the luxury car”. The case would also allow parties to plead to the court and jurors to ask questions and request more information to parties.
5) Voting
After assessing the evidence, jurors commit their vote to one of the options. They are also required to provide a justification for their decision. After a juror has made a commitment, the vote cannot be changed. But it is still not visible to other jury members or to the parties. This prevents the vote of a juror to influence the vote of others. After all jurors have voted, the decision is produced. The winning option is the median one, which gives a consensual result and is robust to strategic voting. Any option below the median will be vetoed by more than half of the voters (the upper half), and the same will happen with options above the median (and the lower-half of voters).
In the freelancer example, if 0% gets 2 vote, 20% 1 votes, 50% gets 2 votes and 100% gets 4 votes, the winning option would be 50%.
6) Appeals
Appealing a court decision is a fundamental right recognized by most legal systems. If a party is not satisfied with a decision done in Kleros, it can appeal and have the case ruled again. Decisions can be appealed several times. In each round, a new jury will be formed with twice as many jurors than the previous instance plus one. The appealing party will be required to make a new deposit in order to pay for arbitration fees.
The cost of appealing the decision is defined by:
Arbitration_fee*Number_Jurors.
Imagine that, after the first instance arbitration, Bob receives the following message:
“Bob:
Kleros has reached a decision on your dispute against Alice and unfortunately you lost. The value of the contract will be transferred to Alice, unless you choose to appeal.
Appealing the verdict will cost you: (Arbitration fee * number of jurors)
If you don’t make the deposit within the next 48hs, the decision will be considered final”.
Bob thinks that the decision was unfair and decides to appeal. The second instance court will be formed by twice as many jurors plus one. Bob will have to pay for seven jurors to be drawn and pay:
Arbitration_Fee*7
Again, the court votes Alice as winner. A new round of appeal would cost Bob the fee for 15 jurors. This time, he decides that the cost is too high and accepts the ruling as final (8).
7) Token Redistribution
After a decision has been made, there is an instance of pinakion redistribution. Tokens are unfrozen and redistributed among jurors following a principle based on the Schelling Coin. Jurors will gain or lose pinakion depending on whether their vote was coherent with the rest.
Assuming that jurors were properly selected, that they were correctly incentivized and that they had access to the same evidence, we would expect them to reach a similar verdict. Following the Schelling point principle, Kleros assumes that jurors who voted incoherently with the rest were not properly qualified (they self-selected into a wrong subcourt), that they did not conduct a proper analysis (they voted quickly just to collect the arbitration fee) or that their goal was not discovering the truth (they were bribed).
Jurors whose vote is within the 25th and 75th percentile of the votes are considered to have voted coherently and are rewarded with tokens. Jurors whose vote outside of the 25th and 75th percentile of the votes are considered to have voted incoherently and are punished with the loss of tokens (the use of the 25 to 75 percentile was suggested by Vitalik Buterin in the post SchellingCoin: A Minimal-Trust Universal Data Feed) (9).
Imagine the dispute between Alice and Bob was solved with 7 jurors. Jurors Ignasi and Julio, who voted incoherently with the majority, have their tokens redistributed proportionally to the jurors who voted coherently. Token redistribution would look like the following:
The pinakion is a critical part of the cryptoeconomic mechanism because it prevents sybil attacks and provides incentives for Kleros to produce true verdicts. The expectation of winning or losing tokens gives jurors the incentive to self-select into the subcourts where they truly have expertise, to analyze the evidence carefully and to vote honestly. A juror who chooses the wrong cases, who does not analyze the evidence carefully or who does not vote honestly is more likely to vote incoherently with others and, as a result, will suffer an economic loss.
Incentives Design: the Combined Effect of Arbitration Fees and Token Redistribution
Within the Kleros network, the incentives for jurors come from two sources: the payment of arbitration fees and the transfer of tokens.
Arbitration fees are payments from parties to jurors as a compensation for the time and expertise invested in analyzing evidence and voting. All jurors adjudicating a claim will earn the same amount. After the jury has reached a decision, a transfer of pinakion occurs from jurors who voted incoherently to jurors who voted coherently.
In all cases, jurors voting coherently with the rest will experience a net economic gain (they will collect arbitration fees and tokens). Jurors voting incoherently will suffer two effects. On the one hand, they will have an economic gain for collecting arbitration fees (all jurors always collect an arbitration fee, regardless of whether they voted coherently or incoherently). On the other hand, they will suffer an economic loss because the incoherent vote makes them lose tokens, which have an economic value.
The net effect (arbitration fees minus tokens) will depend on the difference between the value of the earned arbitration fee and the value of lost tokens.
As usual, some users will try to abuse the system. But Kleros cannot be gamed. Manny learnt this the hard way. He had always thought of himself as the smartest guy in the room. When he found out about Kleros, he saw it as an opportunity to make some easy money. He bought some pinakion and started to activate them in subcourts with high arbitration fees. Of course that, when he was drawn as a juror, he would not even read the evidence. Manny just voted randomly, collected the arbitration fee and moved on to another dispute.
A couple of weeks into his “brilliant scheme”, he realized that he was suffering a net loss. Manny earned some money in arbitration fees. But, since his vote was often incoherent with the rest, he systematically lost pinakion. After a couple of weeks, he realized the net effect was negative and he abandoned his intention of trying to game the system.
Parameters of the Subcourt System
Different disputes require different types of proceedings. Token holders will have the right to make a number of decisions affecting how subcourts work. These decisions include policies, session time, arbitration fees, maximum number of jurors drawn and minimum number of tokens activated.
Policies
A dispute over inheritance rights will be different than a dispute over a freelancing contract or trolling on social media. Some will require communication between jurors and parties while others will not. Subcourt policies will help adapt subcourts to the specific features of the disputes to solve.
Arbitration Fees
Arbitration fees may range from a few pennies to millions of dollars, depending on the type of expertise required to solve a case. Simple disputes that do not require specific skills will probably have low arbitration fees. Complex disputes that require highly specialized (and scarce) skills will have high arbitration fees.
Fees are likely to evolve as to reflect the supply and demand of adjudication services. Subcourts with an excess demand of arbitration services (many disputes and few jurors) will have rising fees in order to attract more jurors. Subcourts with an excess supply of arbitration services (few disputes and many jurors) will have falling fees, driving jurors away. The price system will coordinate supply and demand of arbitration services towards equilibrium in each subcourt.
Session Time
Session time is the amount of time a specific subcourt will give jurors to vote a decision. It may range from a few minutes to years, depending on the nature of the disputes. Simple cases, which may be solved in minutes, will have low session time. Complex disputes, which require more time to properly analyze evidence and come to a decision, will have high session time.
Maximum number of jurors
When the maximum number of jurors is reached in a subcourt because there was a high number of appeals, the higher level subcourt will be called to make a decision.
Minimum amount of tokens activated
Setting a minimum amount of tokens to be activated in order for a user to have the possibility of being drawn as a juror gives parties some control about who can be drawn. The higher the number of pinakion to activate, the more likely it is that only highly committed jurors will be drawn.
Governance
As Kleros reaches a critical mass of users and is adopted in a rising number of industries, new subcourts will have to be created, changes will need to be made in subcourt parameters and the protocol itself will have to be updated to new versions. Since Kleros is a decentralized organization, no central authority will be able to make these decisions. Users will control Kleros governance with a liquid democracy mechanism. Each user will have a number of votes equal to the amount of pinakion he holds. Users will have the option of voting directly or delegating their vote. When a user fails to vote, his voting power is automatically transferred to his delegate.
Imagine a decision about updating Kleros to a new version, with the following token holdings, voting behavior and delegation:
In this example, A delegates his votes on B. Then B has 1500 + 1000 = 2500 votes. B delegates his votes on C. This means that C has 500 + 2500 = 3000 votes. Since C votes Yes, the delegation chain stops and 3000 votes are Yes votes. D also votes Yes with his 3000 votes. This amounts to a total of 6000 votes for the Yes option (3000 + 3000 = 6000). E delegates his 1500 voting rights to F and F votes No. The No option gets 4000 votes (2500 + 1500 = 4000).
The end result is 6000 votes for Yes and 4000 for No. The final result is Yes.
Kleros: a Multi-Purpose Adjudication Protocol
In this paper, we have focused on a software development dispute as an example of the Kleros adjucation proceeding. However, Kleros is a multi-purpose protocol which may be used to adjudicate a wide variety of claims. Some of them may be workable in the near future, while others will probably work in the longer term.
Small Claim Arbitration
In the early 1960s, 11.5% of cases in American federal courts went to trial. In 2002, the number had fallen to just 1.8%. The decline is not the result of fewer disputes, but the consequence of the growing use of alternative processes to deal with problems (Katsh and Rabinovich-Einy 2017, p. 14). In order to cut costs, governments have promoted the use of alternative dispute resolution (ADR) in areas such as commercial disputes, consumer protection and employment. Early applications of Kleros will probably include small claim arbitration.
The credit card chargeback system is the most frequently used dispute resolution process in America, but it can be slow and unsatisfactory to consumers protesting a charge. Imagine a credit card holder (Client) who files suit alleging that the Bank breached the Account Agreement when it “failed to properly review and reverse” a disputed transaction. The decision may be done in Kleros and be enforced by the government, such as any small claims court, where the judgement of a third party is already accepted.
Kleros may also be used in a large number of small claims such as car repairs done wrong, landlords refusing to return a deposit to tenants, telecommunication customers billed for a service they claim they did not purchase, an online buyer claiming that he was sent a defective product, among many others.
Online dispute resolution (ODR) methodologies are currently being tested in government courts beyond small claims. In January 2015, a report by an ODR Advisory Group to the Civil Justice Council recommended the establishment of a court the would handle civil disputes with a value of up to ₤25,000 through ODR in the United Kingdom. Cases would be decided online, and would be binding and enforceable in the same way that traditional court decisions are.
Freelancing
The global workplace is switching from traditional employer-employee relationships to arrangements involving distant, cross-border and flexible work by freelancers (Friedman 2005). Disputes in such agreements require fast and simple responses. The Alice vs. Bob case explained earlier is a typical freelancer dispute. Another conflict may involve a freelance writer who gets his project rejected by a client, leaving him without pay, only to find out that the client ended up using the product anyway.
Apart from a few exceptions such as Alibaba and Upwork (10), online dispute resolution processes are rarely used for disputes involving temporary labor. Amazon Mechanical Turk, a leading platform in the gig economy, generates many disputes but lacks a dispute resolution service. Uber offers drivers very little assistance to address problems such as disputed rankings by customers. A large number of horror stories in apartment rental on Airbnb platform can be seen in the website AirbnbHell.
These platforms see themselves as technology companies that merely connect supply and demand. They usually do not take responsibility for disputes. The Kleros arbitration system could be an important solution for disputes arising in these global platforms.
Crowdfunding
Crowdfunding, a rising source of funding for entrepreneurs, raises a number of concerns regarding scams and false promises on a team ability to deliver on the results promised to backers.
A software startup may create a crowdfunding agreement stating that funds will be transferred after certain development milestones are reached. Next payment will be made after V.2.0 is launched. However, some backers may claim: “This is just V.1.0 with some trivial extra features and some make-up”. Kleros proceeding could analyze the evidence and make the decision (11). Future crowdfunding contracts could have a clause having Kleros as default arbitration procedure.
Social Media
As an increasing number of social interactions start to take place online, harassment, invasion of privacy, false information and bad reviews become important problems. Such interactions may result in reputational and monetary loss as well as psychological damage.
Imagine Soushiant and Uri are having a discussion on a decentralized social media platform. As they heat up, Uri flags Soushiant for a comment that violates the terms and conditions of the platform. Soushiant replies: “My comment did not violate the terms and conditions. It’s not my fault if Uri is too sensitive”. The evidence is analyzed by a jury in Kleros, which decides that Soushiant’s comment indeed violated the terms and conditions and makes him lose 20 reputation points.
Healthcare
Healthcare is increasingly centered on data documented in electronic health records (EHR). Medical services are becoming more entwined with the digital environment under the form of telemedicine and apps for the management and monitoring of medical conditions and medications, or the experimental use of sensors and robots (Katsch and Rabinovich-Einy 2017: p. 107).
An increasing number of life or death decisions will rely on data that may not be accurate and on devices (pacemakers, defibrillators, insulin pumps, etc.) which may fail (eg., software failure, premature shutdown, failure to restart, unexpected depletion in battery life, faulty detection of patient events). A large part of medical malpractice in the future may transition to errors in healthcare records (eg., a typo leading to a medication error) and negligence in the use or transmission of information.
Identifying and responding to these challenges requires less the assistance of lawyers and courts than the use of informal technology-based dispute resolution and prevention processes. Kleros could provide assistance in solving disputes regarding the failure of such devices. A jury of doctors or technicians could analyze the log of the devices and vote on whether a failure existed, which would result in compensation to victims (12).
Intellectual Property
Decentralized music platforms such as Ujo Music will enable musicians to upload music and receive payments using smart contracts. This will generate a number of claims regarding copyright infringement.
Imagine the rock band The Misfits uploads its new song on to Ujo Music. Another band, The Holograms, claims that the song infringes their copyright. In a centralized platform such as YouTube, this claim is solved with the proprietary algorithm Content ID, which blocks any content identified as copyright infringing by matching between materials submitted by a user and copyrighted files. Concern has been raised about the lack of transparency of Content ID and the lack of accountability mechanisms (13).
In decentralized music platforms, the dispute is sent to Kleros and jurors produce a decision: “The Holograms are the rightful owners of the song. All revenue generated by the song is automatically redirected to The Holograms account”.
Online Gaming
League of Legends is a very popular online battle video game with over 100 million players every month. The 2016 World Championship had a total prize pool of over 6 million dollars. To make sure that the game happens in a safe environment, the creators introduced a system for using the community for disciplining bad players. The tribunal is composed of voluntary players, who are empowered to vote on punishment for abuse. Penalties go from a mere warning to a ban. A similar system was implemented by Valve. Kleros could work as a similar, yet decentralized, system for solving disputes in online gaming.
A Justice System for the Decentralized Internet
A large part of our lives happen online. Our social and economic connections are increasingly intermediated by global Internet platforms connecting content producers, consumers and advertisers (Facebook and YouTube), buyers and sellers (Amazon and eBay), drivers and riders (Uber) and travelers and hosts (Airbnb).
If the blockchain promise holds true, the coming years will bring the disintermediation of centralized global platforms and their replacement by distributed models. People will transact goods in decentralized versions of eBay (e.g., Open Bazaar), will contract rides in a decentralized Uber (e.g., Arcade City), will stream music in a decentralized Spotify (e.g., Ujo Music), will contract labor in a decentralized Upwork (e.g., Ethlance) and will connect in decentralized Reddit platforms (e.g., Steemit).
These platforms will use some shared building blocks. For example, they may all use IPFS for storing files and Golem for sharing computing resources. When disputes arise between the users of decentralized platforms, efficiency considerations suggest that they should also use a single decentralized adjudication network. We do not need two separate networks for decentralized storage, for decentralized prediction markets, nor for claim adjudication.
Kleros will become the unified, decentralized, network of jurors for adjudicating claims in different industries. It will become a fundamental part of the infrastructure for the decentralized Internet.
The Kleros network will enable the widespread use of smart contracts in a growing number of economic activities. Smart contracts self-execute when the predefined conditions are met. However, automatic enforcement by code is at odds with a key principle of the philosophy of right: all contracts are incomplete. At the moment it is signed, no contract could ever foresee every possible situation that could arise until the time it is to be enforced. Sometimes, strict enforcement may result in an unfair situation. And this cannot be solved exclusively by computer code.
The Greek called epikeia a moral principle that exempted a citizen of strict compliance with a positive law or contract in order to be faithful to its spirit. Modern legal systems also recognize that parties may be relieved of the obligation of compliance if such obligation has become unreasonable after a change in context. Massive adoption of smart contracts require an “escape hatch” mechanism for when strict compliance would produce undesirable or unfair consequences. But how to create a procedure for an escape hatch without using a centralized decision maker that introduces a new single point of failure into the system?
Kleros can become this method for “error correction”, a decentralized escape hatch to revoke smart contracts when compliance has become unreasonable. It can comply with this, without reintroducing arbitrariness and corruption into the system.
Conclusion
In the past years, e-commerce has been growing at double digits and is expected to total a 2 trillion dollar market in 2020. Key components of the sharing economy (travel, car sharing, finance, staffing and streaming) are expected to reach $335 billion in spending by 2025. Around 2,000 platforms of equity crowdfunding exist in 2016 and the World Bank predicts that crowdfunding investments will be a $96 billion a year market in developing countries alone by 2025.
The legal systems of the nation state era were successful in creating an institutional framework for economic growth and social prosperity. In the wake of the digital revolution, however, they are reaching their complexity limits. While the new economy requires a deep institutional rethinking, few people are conducting research on legal infrastructure from a system level perspective. Lawyers research what the law is. Economists study what the law should be in order to promote trade or improve incentives for workplace safety. But hardly anyone studies how law works as a system, what determines the system’s costs and efficacy (Hadfield 2015: p. 215).
Inability of legal systems to solve the disputes of the Internet Age led platforms such as eBay or Alibaba to develop their own arbitration mechanisms. However, no horizontal system emerged to be used across the board and that could gain from increased specialization over time. Kleros seeks to become this system.
While Kleros relies on cutting edge technologies in blockchain, cryptography, game theory and collective intelligence, its fundamental logic is still based on a principle the Greek knew 25 centuries ago: justice can be served on a peer-to-peer basis. Bitcoin is helping advance the cause of financial inclusion. Justice inclusion is an equally important goal. Just as Bitcoin is bringing banking for the unbanked, the Kleros promise is to bring justice for the unjusticed.
Want a shorter version of this paper? Go here.
Want to learn about the cryptography, math, economics and computer science behind Kleros? Check out a technical version of the paper here.
Want to know more about the project? Interested in joining our team? Visit our website.
The Authors
Federico Ast graduated in economics and philosophy from the University of Buenos Aires. He holds a PhD in management from IAE Business School. He was a participant at Singularity University Global Solutions Program in 2016. He hosts a Coursera program on blockchain to be launched in January 2018. He is founder of Crowdjury and Kleros, projects that leverage the use of blockchain and collective intelligence applied to justice systems.
Clément Lesaege graduated in computer science at the Université de Technologie de Compiègne. He holds a masters degree in computer science from the Georgia Institute of Technology. He follows blockchain technology since 2013 and has found vulnerabilities in several smart contracts.
Footnotes
(1) Alejandro Sewrjugin is a blockchain writer and researcher based in Buenos Aires. Lately, he has focused on the economic applications of blockchain, in connection to income distribution. See his project Phi Economy.
(2) Legal epistemology is a subfield of social epistemology that studies whether the rules that govern judicial proceedings are genuinely truth-conducive (Laudan 2006).
(3) In the 420s BC, jurors were paid three obols for a day’s work. While it was not a grand sum, it was enough to make a difference in the way a man lived and more than enough to sustain mere existence.
(4) Kleros, the name of the project, comes from the use of sortition for jury selection and transparent justice. In governance, sortition (also known as allotment or demarchy) is a method of selection of political officials as a random sample from a larger pool of candidates.
(5) Notably, the United States judiciary kept some of the spirit of Athenian courts in using citizen juries. This participatory feature of the American political system was greatly admired by Tocqueville: “The jury, and more especially the civil jury, serves to communicate the spirit of the judges to the minds of all the citizens; and this spirit, with the habits which attend it, is the soundest preparation for free institutions (…) It invests each citizen with a kind of magistracy; it makes them all feel the duties which they are bound to discharge toward society; and the part which they take in the Government” (Tocqueville 1961: p. 336–37).
(6) Previous attempts were made to use the bitcoin blockchain for arbitration with multi-signature (multisig) addresses. These addresses require more than one key to authorize a transaction. An initiative in this sense is Bitrated that leverages multisig transactions where a trusted third party is selected to solve disputes in case of fraud. However, the limited features of the Bitcoin blockchain are insufficient to handle more complex transactions.
(6) A smart contract is a set of self-enforceable instructions built on blockchain technology.
(6) Pseudonimity is a critical feature of the Kleros adjudication process. It may be seen both as an advantage as well as a disadvantage. It is an advantage from the point of view of protecting jurors from possible retaliation from the parties. However, the lack of identity may turn the system undesirable for cases when individual accountability of jurors is required or when jury must be drawn from a pre-defined pool of candidates.
(7) The name is a reference to the pinakion, the bronze plaque that each Athenian citizen used to be drawn in popular trials. In the Kleros network, the pinakion represents a possibility to be drawn as jurors and therefore to earn arbitration fees. Pinakion will initially be given to people who have taken part in the crowdfunding of the decentralized court. A lesser part of them will be given to project contributors.
(8) Had Bob chosen to keep on appealing, the number of jurors would have doubled plus one again and again in each round. Generally speaking, appeals will be uncommon. However, the appeal system is an important mechanism against bribes. Bribing a small jury is relatively easy. But since the victim always has the right to appeal, the attacker would have to keep bribing larger and larger juries at a steeply rising cost. The attacker would have to be prepared to spend an enormous amount of money to bribe jurors all the way to the General Court and would very likely lose in the end.
(9) Token redistribution is done according to the following formula:
alpha*min_activate*weight.
The alpha parameter determines the number of tokens to be redistributed. The alpha value is decided at the court level depending on the dynamics of the voting environment. The weight parameter is the number of times that a token holder was drawn for the dispute. The amount of tokens a juror will gain or lose will depend on the voting power he had in the dispute.
(10) Upwork is a freelancing platform for technology, writing, translation and design, which has an online dispute resolution service. Contracting parties are instructed to communicate through the assigned workspace so as to document all their agreements, communications and updates on the freelancer’s progress. Failure to document agreements on the workspace will result in the loss of Upwork protection for parties if a dispute arises in the future. Parties use the tools provided to resolve the problem on their own in 3% of the jobs. Users contact Upwork customer support only in 1% of the jobs. The types of problems that arise usually have to do with nonpayment on the freelancer end, or payment for an incomplete job on the client’s end; the average value of disputes is $200. (Katsh 2017: p. 143) Upwork offers a payment mechanism in which money is placed in escrow. After milestones are met and portions of the work are performed satisfactorily (and it’s the client’s obligation to review and object to payment within a specified time), then money is released.
(11) It is possible that a contract of this sort could have been useful to return funds to backers in the DAO hack event in June 2016. Money could have stayed in escrow and jurors could have decided that the DAO had failed to meet some security milestone. After Kleros decision, funds would have been returned to backers.
(12) “Computerized medical devices can fail in many ways, including through programming errors, incorrect calibration, and exposure to malicious intrusions, as well as physical or medical errors. Over a thousand recalls were issued on software-based medical devices from 1999 to 2005. Hundreds of deaths have been attributed to software failure in medical devices.” Long Comment Regarding a Proposed Exemption Comment of a Coalition of Medical Device Researchers in Support of Proposed Class 27: Software — Networked Medical Devices, at 2, U.S. Copyright Off. Libr. Congress.
(13) “Appropriation of existing material is a venerable and necessary practice. As the economists Romer and Arthur remind us, recombination is really the only source of innovation — and wealth. I suggest we follow the question, “Has it been transformed by the borrower?” Did the remixing, the mashup, the sampling, the appropriation, the borrowing — did it transform the original rather than just copy it? Did Andy Warhol transform the Campbell’s soup can? If yes, then the derivative is not really a “copy”; it’s been transformed, mutated, improved, evolved. The answer each time is still a judgment call, but the question of whether it has been transformed is the right question” (Kelly 2016: p. 209).
References
Black, Duncan (1948). On the Rationale of Group Decision-Making. Journal of Political Economy, 56. The University of Chicago Press.
Blt, Manuel (1981). Coin Flipping by Telephone: a Protocol for Solving Impossible Problems.
Boegehold, A. L. (1995). The Lawcourts at Athens: Sites, Buildings, Equipment, Procedure, and Testimonia. American School of Classical Studies at Athens.
Brabham, Daren (2013). Crowdsourcing. The MIT Press Essential Knowledge series.
Brassard, Gilles; Chaum, David and Crepeau, Claude (1988). Minimum Disclosure Proofs of Knowledge. Journal of Computer and System Sciences, 37, 2, 156–189. Academic Press, Inc.
Buterin, Vitalik (2016). Decentralized Court.
Buterin, Vitalik (2014). SchellingCoin: A Minimal-Trust Universal Data Feed.
Buterin, Vitalik (2015). The P + epsilon Attack.
Buterin, Vitalik (2014). Ethereum, A Next-Generation Smart Contract and Decentralized Application Platform.
Buterin, Vitalik (2017). Introduction to Cryptoeconomics.
Certicom Research (2010). Standards for Efficient Cryptography 2.
Clarkson, William (2013). Time-Lock Cryptography, Sending Messages to the Future.
Douceur, John (2002). The Sybil Attack. Springer-Verlag.
Ford, Bryan (2002). Delegative Democracy.
Friedman (2005). The World Is Flat: A Brief History of the Twenty-first Century. Farrar, Straus and Giroux.
Friedman, David (1994). A Positive Account of Property Rights. Social Philosophy & Policy, 11. Cambridge University Press.
Hadfield, Gillian (2016). Rules for a Flat World: Why Humans Invented Law and how to Reinvent it for a Complex Global Economy. Oxford University Press.
Katsh, Ethan and Rabinovich-Einy, Orna (2017). Digital Justice: Technology and the Internet of Disputes. Oxford University Press.
Laudan, Larry (2006). Truth, Error, and Criminal Law. An Essay in Legal Epistemology. Cambridge Studies in Philosophy and Law.
Martic, Dusko (2014). Blind Arbitration. Sintelnet WG5 Workshop on Crowd Intelligence: Foundations, Methods and Practices.
Nakamoto, Satoshi (2008). Bitcoin: A Peer-to-Peer Electronic Cash System.
Peterson, Jack and Krug, Joseph (2015). Augur: a Decentralized, Open-Source Platform for Prediction Markets.
Reitwiessner, Christian (2016). From Smart Contracts to Courts with not so Smart Judges.
Schelling, Thomas (1980). The Strategy of Conflict. Harvard University Press.
Sztork, Paul (2015). Truthcoin, Peer-to-Peer Oracle System and Prediction Marketplace.
Teusch, Jason and Reitwiessner, Christian (2017). A Scalable Verification Solution for Blockchains.
van den Herik, Jaap and Dimov, Daniel (2012). Towards Crowdsourced Online Dispute Resolution. Journal of International Commercial Law and Technology.
van den Herik, Jaap and Dimov, Daniel (2011). Can the eBay’s Community Review Forum Fairly Resolve Disputes? Proceedings of the 23rd Benelux Conference on Artificial Intelligence.
