In September 2019, Kleros, an online dispute resolution platform that utilizes blockchain technology and crowdsourcing, released the latest version of its White Paper. The Paper sets out how they aim to revolutionize dispute resolution, by bringing justice cheaply and efficiently to the masses. A glimpse at Kleros’ Twitter feed reveals an ambitious vision beyond just dispute resolution in smart contracts, into decentralized courts that theoretically could solve any dispute (FN1), and even determine whether a persona on the internet is a human (‘Proof of Human’). The Paper also describes how the blockchain will be utilized, and the use of game theory to incentivize various stakeholders in a sui generis justice system. To use Kleros’ CEO Frederic Ast’s analogy to Henry Ford’s revolutionary Model T (FN2), it is not about asking people whether they would like their horses to go faster. As the Model T did for transport, so Kleros aims to achieve in dispute resolution. Eminent international arbitrator Professor Sophie Nappert recently gave a talk at Columbia Law School on “Arbitration in the Age of Blockchain and Augmented Reality.” Professor Nappert has previously discussed her views on Kleros elsewhere, and this article will touch on some of them (FN3, FN4), as we discuss how the platform challenges our notions of what having justice done by a jury or arbitration means.
Kleros was founded in 2017 (FN5), based on the concept of a klerotarion, a method of random selection of a jury (FN2), that was ostensibly used in a particular form of Athenian colony which were centered around agricultural plots (kleroi) allocated to citizens (FN6). The notions of allocation and democracy alluded to are important to how stakeholders may participate and are incentivized in the system.
The idea is that anyone can stipulate to have Kleros as their choice of dispute resolution system in their smart contract (based on the blockchain). What follows is a very brief summary. Once ‘arbitration fees’ are paid, the dispute and all relevant evidence is submitted to and arbitrated by ‘jurors’, who have claimed specializations relevant to the dispute. After a period of deliberation, they will decide for either party, or perhaps on a compromise. The final decision follows the consensus. This is where ‘arbitration’ and ‘jury’ are no longer used in a traditional sense. Firstly, arbitrators have to ‘stake’ an amount of tokens (called PNK, and which are a separate token system from the token/currency used in the contract), in order to prevent the possibility of any one actor taking control of the system and “dishonesty”. The greater the amount one stakes, the greater the probability of being selected. Secondly, mechanisms are put in place to prevent: i) bribery of jurors, ii) purchase of more than half the system’s tokens; iii) prevention of dishonesty; iv) premature release of decisions to prevent jurors’ deliberations from influencing each other; and v) attacks on the system by malicious actors. Thirdly, jurors are incentivized (i.e. paid) for decisions that accord with the majority, and penalized for minority decisions (for not being “coherent”).
Consensus and Game Theory
Importantly, Kleros is predicated on the concept of a Schelling Point. Invented by the game theorist Thomas Schelling, it refers to a solution that people fallback on in the absence of communication, because it is natural, special or relevant to them (FN7). Schelling uses the example of New York’s Grand Central Terminal at 12:00 PM as a natural meeting point for people who have to meet but cannot communicate.
Alternatively, as Schelling himself puts it, the Schelling Point is about “focal point(s) for each person’s expectation of what the other expects him to expect to be expected to do” (FN8) – the Paper equates this to “honesty”, plain and simple.
Intended use-cases of Kleros include escrow, micro-tasking (e.g. Mechanical Turk style work), oracles (methods of providing decentralized input about real world events or facts), curated lists (analogous to lists of legitimate and spam e-mail addresses), and safeguarding of social networks.
Analysis and Evaluation
Besides the often discussed and claimed benefits of efficiency, streamlining, automation, and decreased litigation that are often brought up in discussions about smart contracts and blockchains, Kleros begs even more questions. Kleros’ proposal is a radical reimagining of a judicial system: the system is neither truly a jury nor truly an arbitration panel. On the other hand, Kleros’ framing of the question of how to allocate scarce judicial resources and litigation funding is attractive because it is novel and compelling.
Reframing a question often provides an entirely new set of tools for solving a problem. Poincare, when considering the extremely difficult mathematical problem of how to predict accurately how three bodies would move in relation to each other (e.g. Sun-Earth-Moon), reframed the question, in simple terms as: “Will the moon ever fly away?” (plain reformulation by Columbia’s Professor Christos Papadimitriou). Complexity theory now owes a huge debt to that reframing. That does not mean all reframings are accurate or correct. This has several implications.
Crucially, the incentivization of consensus challenges several long-held and respected legal traditions and principles. The jury is seen as an almost sacrosanct representation of society. It continues in spite of modern criticism about whether it is truly representative or rational about the evidence presented to it. Jurors have a duty to fulfill, and a corresponding responsibility to be fair and just. That people need to be incentivized to carry out what is traditionally a duty and responsibility needs to be better explained. It offends the traditional notion of duty. In addition, efficiency does not speak to the moral power of a jury, even when efficiency is invoked as a path to increased justice and fairness. That people can and have been bad at their duties does not mean a market should be created to incentivize the best jurors or jurors with the most expertise. There are other important considerations besides efficiency, and the allure of more direct representation in the justice system via the adjudicator rather than an advocate. Just as we cannot all expect judges to be wise, we cannot expect all consensus to be wise.
As Professor Nappert and Paul Cohen have argued, this bears many similarities to social media dynamics: justice may become a matter of consensus, or a matter of who has more likes.
On the other hand, if the application of game theory means experts are efficiently incentivized by the resultant market, then perhaps experts on Kleros should be made to provide reasons. Arbitrators are engaged for their competence at adjudication, which always aims to provide reasons for a justice outcome. Reasons should provide a clear depiction of how an adjudicator has reached a conclusion. This holds the adjudicator accountable, and renders the outcome comprehensible to parties. On Kleros, while just outcomes are prioritized, the consensus does not become any more transparent. Understanding why one has been ruled against is an important component of justice.
Judges and arbitrators around the world have recognized that arbitration and other dispute resolution systems are increasingly seen as too expensive and sclerotic (FN9). In response, Kleros poses compelling solutions, even if they need to be more worked out. It will probably not be the last.