Blockchain Startups Think Justice Can Be Decentralized, but the Jury Is Still Out

Image from iOS 4 1170x585 - Blockchain Startups Think Justice Can Be Decentralized, but the Jury Is Still Out

Parisian Marc Zeller works for a cryptocurrency brokerage firm by day. By night, and on weekends, he helps decide cases as a juror for the crowdsourced blockchain dispute resolution system Kleros.

The 39 year old has judged 35 disputes over the past six months and made 10.5 Ether and 229,000 Pinakion tokens in rewards — in total, worth around $3,200.

Most cases take no longer than five minutes to decide — though he has spent days diving into the technical details of a couple of particularly complex disputes.

Kleros has never confirmed his identity or checked if he has the technical knowledge required to understand the cases he’s adjudicating. He has no legal training or qualifications.

Instead, the platform relies on game theory and crypto economics, rewarding jurors with tokens in a way it hopes incentivizes them to vote fairly, in areas they believe they have expertise.

“Game theory means you participate honestly in the Kleros system — I believe more in that than in good will,” Zeller said. “If you behave in the wrong way, you lose more money on average.”

Decentralized justice

Welcome to the emerging crowdsourced blockchain legal system, a world of smart contracts and decentralized justice.

There are a number of cryptocurrency projects built on very similar concepts, including the smart contract creation and dispute resolution platform Jur and the “world’s first digital jurisdiction” Aragon Court. Both will start to judge cases in the next month or two.

The question is: Can these systems really provide a cheap and fair alternative to going to court or is the whole idea of decentralized justice just a utopian pipe dream?

Smart contracts need a dispute mechanism

Amy Schmitz, a law professor at the University of Missouri School of Law, co-chairs the American Bar Association’s task force on online dispute resolution.

She was hesitant about the entire concept at first, but said the increasing popularity of smart contracts requires a faster and cheaper way to settle disputes.

At their most basic, a smart contract is just computer code that runs on a blockchain and executes automatically — say, paying out a bet when a sporting match is over or sending a transport company money when a shipping container reaches its destination.

But smart contracts are ill-suited to existing legal mechanisms. In their raw form, they can only be understood by computer coders. And they have a questionable legal status in countries such as the United States, where civil law only recognizes contracts in written or documentary form.

Schmitz says “square” legal concepts — i.e., those of offer, acceptance and consideration — are difficult to apply to the “round hole” of smart contracts. And smart contracts are often between anonymous parties in different countries that have vastly different legal systems. Finally, the slow, traditional legal system is likely to prevent smart contracts from scaling.

Plenty of things can go wrong with smart contracts, including with the code itself. Schmitz cites a 2016 study revealing 100 errors per 1,000 lines of code.

“Computer coders make mistakes all the time,” she said. “The parties also might also have a different idea of what was in the code. There might be problems with the product that was delivered.”

“You can say that everything in a smart contract is ‘If/Then’ but a lot of times in life, it’s a ‘maybe’ — so they’re going to have to have some way to quickly resolve these disputes.”

Crowdsourced jurors and game theory

The idea to combine game theory with crypto economics can be traced back to a proposal by Ethereum co-founder Vitalik Buterin to utilize a concept called the Schelling Point back in 2014.

It’s named after economist Thomas Schelling, who famously asked a group of students in New York where and when they would choose to meet a stranger in New York City if they couldn’t communicate with them beforehand.

They had to try to predict what other people would do, and so the most common answer was to meet at noon at the information booth at Grand Central Station.

Kleros co-founder Federico Ast likes to perform a similar demonstration with audiences, writing three numbers on the screen: 4,555, 6,876 and 1,000.

He tells them they can win $20 if they correctly predict which number most of the other people will pick.

“They all choose 1,000 because they think the others will choose that,” he said. “Jurors are incentivized to vote with the majority because they win tokens. It’s the foundation on which the Kleros incentive system is based.”

Because individual jurors believe most of the other jurors will vote for the fairest outcome, voting for the fairest outcome becomes the best way to gain more tokens.

The concept is foundational for the Jur Open Layer and Aragon Court as well.

Kleros is on the case

The Kleros process works like this: Parties to the dispute pay Ether as fees to the jurors. The jurors who are selected stake a certain amount of Pinakion tokens (PNK) to participate, and they evaluate the evidence submitted by opposing parties. 

If they vote with the majority decision, they get to keep their tokens and receive a share of the tokens forfeited by jurors who voted with the minority.

As they can also lose tokens, they’re essentially gambling on their own ability to predict the “correct” outcome.

It’s still the early days — a total of around 70 jurors have judged matters worth about $300,000.

Most of the 130 cases so far required the jurors to decide whether an Ethereum token was suitable for listing on Bitfinex exchange offshoot Ethfinex, which has since morphed into the decentralised exchange DeversiFi (which still uses Kleros for listings). The list of approved Ethereum tokens is also available in custom front ends for Uniswap and DutchX.

The system was first tested with “Doges On Trial,” where users were encouraged to try to slip a picture that wasn’t a dog past the jurors — using bribery, manipulation or fake evidence. While it highlighted areas that needed refinement, the system held up well to 51% attacks and showed that it disincentivises bribery due to its appeals process. 

More recently, Kleros has been running test “fake news” cases to see if jurors can reliably identify which media reports are accurate and which contain misinformation. Other use cases include adjudicating crowdsourced answers on Realitio and, soon, translations on a decentralized app called Linguo.

Ast said Kleros will test more use cases over the next two months, including getting jurors involved with social media content moderation and deciding on insurance claims.

Theoretically, the process could be used to decide anything from disputes over property to freelancer agreements — with specialized sub-courts set up for each area and expert jurors in each.

Legal expert takes Kleros for a spin

James Metzger, a lecturer at the University of New South Wales Faculty of Law, wrote a paper on the emerging blockchain justice field and put Kleros to the test as a juror.

“I think it’s really interesting but potentially problematic for a lot of reasons,” he said. “In a court, you have procedural fairness — but here, it’s the Wild West with no regulation.”

There are no rules of evidence, no judge to ensure both parties receive a fair hearing, he said.

“The assumption the platform has is that everybody is going to consider evidence carefully and vote what they believe to be a fair and just outcome because they’ll assume that’s what other jurors are doing.

“But there’s no proof of that yet; it’s just an assumption based in game theory.”

Game theory isn’t perfect

Game theory is applied to situations where decision-makers interact strategically and follow rules based behavior — when they act rationally and purposefully to maximize their own advantages.

One criticism is that people often behave in a way game theorists might view as irrational or against their own interests. Think of a religious person refusing medical treatment or someone harboring such a grudge that they’ll put up with being hurt themselves as long as they can cause even more pain to the person they hate.

Critics argue it’s difficult to account for all the different variables that can influence strategies and outcomes. There could always be a wild card that is not factored in.

The real world just isn’t as neat as some of these theories, as the New Yorker pointed out when it outlined the benefits and shortcomings of Nobel Prize winner John Nash’s concept of the Nash equilibrium.

In real-world game theory experiments, subjects often have a personal stake in the outcome. Metzger is concerned these concepts may not translate perfectly to a decentralized online space, where individual jurors don’t have a personal stake in the outcome of the dispute.

He believes the danger is that voters are simply incentivized to vote in a way that increases their stack of tokens — not to vote fairly.

“I’m not convinced that the incentives these platforms think exist, actually do exist,” he said.

“The jurors don’t have a stake in the dispute but they do have a stake in trying to keep their tokens.”

Expertise? We have a theory for that

Both Metzger and Schmitz raise concerns that jurors without appropriate expertise can “self-select” the sub-courts they wish to work on.

However, there are disincentives for jurors selecting the wrong courts.

For example, in a sub-court devoted to disputes involving web design, jurors who know nothing about web design are more likely to vote differently than the others who do. So, they’re more likely to lose tokens.

“If you don’t know about website design, you will vote randomly, but the others will know more and vote coherently,” Ast said. “So, Kleros jurors have the incentive to self-select into courts where they have the experience to arbitrate.”

However, Kleros expects that most users will chose to have cases heard initially by just three jurors, as it is cheaper. And with such few participants, it’s possible that none of them will know anything about the subject, which would render the supposed disincentive moot.

Schmitz also worries that those who control more tokens could end up with too much power — as of last week four wallets had more than 51% of the PNK supply — and that incentivizing people to vote with the majority decisions could become a form of mob rule.

“That’s a huge problem we have in today’s society in general with mob justice going on all over the place,” she said.

“How do you incentivize people to go with the right answer? I don’t know. I feel more comfortable having a single arbitrator who is an expert in the field or a panel of three [experts].”

Real-world testing and use hasn’t borne the fears… yet

During the “Doges on Trial” live testing, Zeller — the moonlighting online juror — said he attempted to bribe the other jurors as well as gain 51% of the tokens — all in the name of science, naturally.

“I managed to win a few cases but I ended up losing all my tokens,” he said. “The system is pretty ruthless and the incentives are really on the side of good decisions.”

He believes the system encourages jurors to vote for the minority if they believe it’s the most fair outcome that will prevail in the end.

That’s because appeals escalate to be heard by twice as many jurors — and if a decision is overturned, it’s quite lucrative for original minority jurors.

“You are grabbing all the stakes from the ‘wrong’ majority in the first round, which is crazy good financially for you,” he said.

Do-it-yourself smart contracts

Rival platform Jur hopes to have its competing Open Layer system up and running in the next month or two.

The Jur Beta Platform went live in November and already allows ordinary people to create their own smart legal contracts. Legal Zoom and Rocket Layer are experimenting with similar blockchain-based smart legal contracts.

Jur will offer dispute resolution in three ways. The crowdsourced Open Layer is aimed at small value disputes. A Community Layer has experts in the field judge disputes — so, conveyancers might judge property disputers, for example. And then there’s the Court Layer, where legal practitioners provide arbitration that is recognized by 161 countries. It’s designed to comply with the requirement of two international agreements: the New York Convention and the UNCITRAL Model Law on International Commercial Arbitration.

“The end goal is to render arbitral awards that are recognized and enforceable almost everywhere,” Jur’s legal engineer, attorney Luigi Cantisani, explained.

The first genuine contract was executed on the platform in early November, when Vienna-based graphic designer Daniela Rieser bought a second-hand car from her former partner Stefan.

The pair reported it only took a few minutes to fill out some forms, agree to conditions around the payment of $10,000 worth of JUR tokens, and they had a legally enforceable smart contract with a $1.50 fee for accessing the VeChain blockchain.

The money was held in escrow until both parties were either happy to sign off — or raise a dispute.

Stefan, a 32-year-old lab technician, said they had to consider how a dispute might play out when setting up the contracts.

“We had to formulate very clear resolution proofs — KPIs where I had to hand over the keys and documents in seven days and she had to give me the money,” he explained.

“We did documentation of the handover of the keys and in case of a dispute, I could have uploaded the photos and then the Jur community would have decided if the handover was OK or not.”

No disputes allowed, yet

Fortunately, neither raised a dispute, as the Open Layer is still being tested by 30 law students at the Lab For New Justice at Radboud University in the Netherlands. This means that no disputes are able to be adjudicated by the community for the time being.

Apart from user-testing the interface and identifying bugs, a panel of legal academics led by Andre Janssen, a professor at Radboud, will evaluate the decisions to see if the “crowd” gets it right.

Janssen said he’s cautiously optimistic about the technology.

“Our justice systems right now are time consuming and cost a lot of money. Jur and Kleros are trying to fill this gap with something better. There are some key issues that need to be resolved but it might [provide] cheap and immediate access to justice in the future.”

He said the legal profession is paying close attention to developments.

“The increasing interest in these smart dispute resolution systems is very obvious when you read legal journals,” he said. “I talk to a lot of practitioners and they are very interested in it.”

“I think it will get market share.”

Jur’s Open Layer really is crowdsourced

Jur’s Open Layer relies more heavily on the wisdom of crowds than Kleros. Anyone with JUR tokens can vote — potentially meaning hundreds could weigh in on a case.

Voters can vote as often as they wish on a case, as long as they stake more JUR tokens.

While this seems as if it could create problems, Jur’s creators have built safeguards into the system.

To stop “whales” from spending lots of tokens to open up an unassailable lead in the voting, there’s a mechanism called the “100% rule” that stops the counting of votes on the majority side once it doubles the minority’s vote. The theory behind this is that the minority side would be able to make up the gap if it’s not too big.

Jur’s game theory expert, Michele D’Asaro, suggested this rule is not so much based on game theory but is a way of safeguarding against possible defects in its application.

He said that as every juror has a personal stake in the platform succeeding, they theoretically shouldn’t undermine it’s chances of success by cheating.

“On the other hand, in real life you could have actors that try to steal as much as they could playing unfairly and then disappear,” he said.

“The 100% limit is a way of incorporating these kinds of concerns into the Open Layer dynamics.”

Similarly, only the votes necessary to get the majority vote over the line are rewarded with tokens.

So, if 50 votes are cast for the minority and 1,000 votes are cast for the majority, only the first 51 people who voted for the majority win tokens.

“If you come along later and try and game the system by just voting with the majority to get tokens, you’ll get no rewards,” legal engineer Cantisani explained.

These “rules” and incentives will be tweaked and refined during the testing process. 

Cantisani is pretty confident in the system. He said it’s “based on math” and that all the underlying formulas will be released soon.

 “We know this is going to work because it’s based on game theory — it’s not based on speculation but on science, so it’s reliable,” he said.

“Second, we know it will work because even if it doesn’t work, we will fix it.”

Utopian token holders

Schmitz said she had rejected the whole concept of decentralized justice initially, but after she learned more about it, her initial fears were not realized.

“When I first met with the Kleros guys, I was very against it. I called it pure gamification. I thought it was just like the lottery, I thought this should never be permitted because it is essentially betting on determinations,” she said.

“But when I saw the use cases, and how it’s working it seemed less problematic to me.”

Part of the reason it works, she said, is because token holders understand blockchain and have “utopian” ideals, “so it would almost be like blasphemy within their world to cheat the system.”

Cantisani pointed out that there’s another less utopian and more pragmatic reason that token holders don’t behave unethically:

“If you game the system, then it becomes unreliable. And then your tokens become worthless.”