Kleros Live Stream, 12 August 2026: can Kleros plug into a country's consumer courts?

Inês Bragança Gaspar is the only person to have completed the Kleros Fellowship twice, and the second time was not her idea. Kleros went to her. Her ongoing research asks a question every institutional conversation eventually reaches: could Kleros actually be integrated into a country's consumer dispute system, in her case Portugal's, and what would have to change first?

Her answer takes up the first half of the call: four integration models, one legislative precedent from Mexico, and a hard collision between anonymous jurors and a law that requires arbitrators to sign the award. The second half is the ten-year Kleros thesis on artificial intelligence, published on the blog this week, and what happens to a court when some of the jurors are agents.

📋 The call at a glance

Portugal's consumer arbitration system is growing. More than 12,000 new cases in 2025, up 17 percent, in a system where companies are obligated to arbitrate disputes under 5,000 euros.
Four models for integrating Kleros: a fact-finding layer for arbitrators, direct e-commerce integration, a digital chamber inside an existing center, or a technical partnership.
Mexico already wrote decentralized justice into law in 2024. That is the shape of the reform Portugal would need.
The anonymity objection has an answer in production: the Junín court restricts jurors to KYC-verified lawyers using soulbound tokens.
The decade-of-AI thesis is published, and the open question is whether agent jurors play fair or merely rational.
Scout submission costs are falling, which reads like agents already doing the work. The numbers are public on rewards.kleros.io.
The 10th Fellowship cohort is open. This kind of research, for your own jurisdiction, is exactly what it is for.

Chapters — jump to the moment

0:00Welcome 0:53Introducing Inês Bragança Gaspar 3:55Decentralized justice and consumer arbitration 9:11Where Kleros already resolves consumer disputes 11:25Four models for integrating Kleros 13:24Barriers and Mexico's 2024 precedent 16:42Why Portuguese ADR cases grew 17 percent 20:42Can arbitrators be foreign specialists? 23:31Anonymous jurors, KYC and soulbound tokens 28:34The 10th Fellowship cohort is open 31:17Confidential evidence (William George) 32:32A decade of Kleros and AI 34:47Agent to agent disputes and reputation 38:27The ultimatum game and the Machiguenga 44:01What this means for Kleros jurors 46:28Staking, Scout and Proof of Humanity rewards 50:18Scout skills and the ATQ registry 53:40Short-form channels and academic mentions

A consumer court the company cannot refuse

3:55 · Inês Bragança Gaspar

Inês is a Portuguese lawyer who found decentralized justice while researching her master's thesis in 2018, before there was much of it to find.

“I think that Federico coined that term, actually, so I was writing about it before it was even a reality.”
Inês Bragança Gaspar · 2:58

The system she studies has a property most countries lack. Since 2019, Portuguese law extends arbitragem necessária, mandatory arbitration, to consumer disputes worth up to 5,000 euros: once the consumer files with an authorized center, the company is obligated to take part. Around 40 arbitration centers exist, twelve of them authorized for consumer disputes, all non-profit, government-authorized and cheap or free to use. And they are being used more: over 12,000 new cases in 2025, a 17 percent jump, which Inês attributes to a government publicity push, television ads included, and to people simply buying more things.

That growth matters for the integration question, because a system processing more cases with the same budget is a system with a reason to listen.

Four ways to plug a decentralized court into a legal system

11:25

Kleros already resolves consumer disputes in the wild, and Inês built her models from those cases. Lemon integrated Kleros into its complaint process in 2024: if a user disagrees with the company's own resolution, they escalate to Kleros, jurors rule within five business days, and the ruling binds the company while leaving the consumer free to go to court. Maldo.uy, a Uruguayan freelance marketplace, is researching the same move, and an insurance pilot is underway in Mexico.

“The first model would be to use Kleros as a fact-finding or evidence assessment layer, so arbitrators could submit to Kleros jurors certain factual questions. For example, was the product delivered as described?”
Inês Bragança Gaspar · 11:45

Inês's four integration models, the barriers she identified, and the three-phase pilot she proposes.

“Mexico in 2024 was the first country to expressly recognize decentralized justice in its legal framework.”
Inês Bragança Gaspar · 13:47

Asked which path is likeliest in the short term, she did not pick the arbitration centers. Private companies first: they are comfortable with new technology, and food delivery platforms are resolving thousands of complaints a day with chatbots that do not understand what the consumer's problem is. The centers come later, with evidence in hand.

“If disputes are becoming more technological and digitalized, why are we still solving these disputes through the traditional routes, with lots of paper and going to court in person?”
Inês Bragança Gaspar · 22:42

Inês's full presentation, published as its own video · Kleros and Consumer Arbitration in Portugal

Who signs the award?

20:42 · Inês Bragança Gaspar & Facundo Trotz

The sharpest exchange of the call started with a detail of Portuguese law that cuts both ways. Any individual with full legal capacity can be an arbitrator, regardless of nationality, and parties can require specific qualifications. That opens the door to specialised chambers drawing on expert arbitrators from anywhere in the world. But the same law closes another door: jurors cannot be anonymous, because the arbitrators sign the arbitral decision.

“The jurors need to sign the arbitral decision.”
Inês Bragança Gaspar · 23:55

That objection, your jurors are anonymous so we cannot use you, is the most common institutional one there is. Facundo Trotz answered it with something already in production. The court used for the Junín pilot in Argentina is not open to anonymous stakers.

“In Junín cases there is a specific court in Kleros which is formed by jurors that are lawyers. And the way we do this is by using SBTs, soulbound tokens, that work as a kind of digital credential.”
Facundo Trotz · 25:08

The Kleros Cooperative runs a KYC process, the juror accepts terms and conditions, and the soulbound token gates the court: no credential, no ability to stake there. The protocol underneath is unchanged. Inês took the idea one step further: arbitration centers could issue soulbound tokens to their own existing arbitrators, with the credential encoding the expertise they already have on paper. The 2019 protocol she first wrote about is still there, Federico noted, Schelling point and all. What changed is how much can be configured around it.

Do this for your own country

28:34

The point of putting this research on the call was not Portugal. Applications for the 10th Fellowship cohort are open, and Federico's ask was to replicate exactly this shape of work, mapping how Kleros could attach to the legal system of whatever country you live in. Inês's endorsement carries some weight, being the only person to have done the Fellowship twice, and she was clear it is not a lawyers-only program: her cohort included mathematicians and economists. Her full report will be published soon.

Ten years of the AI thesis, written down

31:17 · William George & Federico Ast

William George's update was brief: the confidential evidence talk he gave in English two weeks ago and in Spanish on Monday's call is becoming a blog article, a longer version of the toolbox. The bigger publication is already out: the essay tracing ten years of Kleros research on artificial intelligence, from the Doges on Trial experiment and the 2019 thesis it produced to the Lemon experiment published by Kluwer, where different AI models gave different answers to the same real cases.

Justice in the Algorithmic Society: A Decade of Kleros and Artificial Intelligence

The thesis has three parts. High-complexity, high-nuance cases stay with humans, because they require moral context. Mid-complexity cases become a combination, perhaps an AI first instance with appeal to humans.

“You don't want an AI to make just a final decision. That could be dystopian.”
Federico Ast · 34:38

And the third part is a category of dispute that has not existed before: agent against agent, at volumes and speeds no human process can serve, settled through escrows and reputation systems. Federico sketched agents using reputation as loan collateral, working capital lent from one agent to another on the strength of a record. Announcements in this area, he said, are cooking now.

“We just needed in some way the world to catch up with some of these ideas, which seems to be happening now.”
Federico Ast · 36:17

Do agents play fair, or merely rational?

38:27

The most fun stretch of the call was a behavioural economics question with real protocol stakes. In the ultimatum game, one player offers a split of 100 dollars and the other accepts or rejects; rejection leaves both with nothing. Pure rationality says accept anything above zero. Humans do not: offers under roughly 30 percent get rejected out of something like spite, or fairness, depending on your school. The known exception is the Machiguenga, a tribe in the Peruvian Amazon with almost no market economy, who accept low offers exactly as the textbook predicts. The anthropologists' thesis: notions of fairness and reputation come from operating in markets, where lowballing strangers has a cost.

So which are the agents? The empirical answer so far, from an emerging research field, is that it depends on the model: sometimes they reject low offers like humans, sometimes they behave closer to homo economicus. For a court that works because jurors converge on what other jurors will think is right, the difference is not academic. An agent juror weighing what other agents will do, all of them having read the same game theory literature, is a different equilibrium from a human crowd.

“The system needs to be robust against hyper-rational actors. But you may have a bunch of semi-rational actors that are kind of the average honest user.”
William George · 44:59

This, too, folded back into the Fellowship pitch: experiments on how different models decide disputes, and how agent juries behave, are precisely the research William and Federico want proposals for.

The falling cost of a submission

46:28 · Jean

Jean closed with the rewards transparency page, which now shows all three programs in one place: monthly staking rewards for jurors on Ethereum mainnet and Gnosis, Scout rewards for registry submissions, and a PNK reward for registering on Proof of Humanity. The page, built by Marino, surfaces one number worth staring at: the cost per accepted registry entry is falling, several-fold.

“The submitter is not the manual worker anymore. It is the manager of the agents. Maybe one submitter in one hour can make 50 submissions, and before they needed eight hours to do the same thing.”
Jean · 49:48

The Scout skills released a few weeks ago point the same direction, and the effect cuts the other way too: the ATQ registry's entries can tag a million contracts each and used to require JavaScript knowledge to review, which meant few people could challenge them. With AI assistance, checking and challenging technical submissions stops being a specialist's job. Federico's summary was that this is what an AI economy at work looks like: the cost of finding, submitting and verifying information dropping across the board.

The papers nobody tells us about

53:40

The academic mentions digest is getting feedback from the universities and authors it features, and may become weekly, since the mentions keep arriving. The team is also reviving its ResearchGate and Academia.edu presence to catch the research that never enters the journal circuit: master's theses applying Kleros to problems nobody at Kleros thought of. The example that surfaced last week was a thesis on Kleros for dispute resolution in the horse industry.

“That's something we didn't see coming. I imagine how many people are doing master's theses that go under the radar, and maybe we want to know about that.”
Federico Ast · 56:07

Mentioned in this call

VideoThe full stream · 58 min, 18 chapters, every timestamp above deep-links into it
VideoKleros and Consumer Arbitration in Portugal · Inês's full presentation, published as its own cut
VideoPrivate Evidence, Public Court · William's talk, given in Spanish this Monday, blog article coming
ArticleJustice in the Algorithmic Society: A Decade of Kleros and AI · the ten-year thesis, published this week
ArticleDecentralized Justice and AI: Friends or Foes · the 2019 original, from the Doges on Trial era
ArticleThe New Arbitrator Selection Problem in the Age of AI · Kluwer Arbitration Blog, the experiment on real cases
ApplyKleros Fellowship of Justice, 10th generation · replicate Inês's research for your jurisdiction
Productrewards.kleros.io · Scout · Proof of Humanity
PartnerLemon · the escalation integration Inês's second model generalises · Maldo.uy · researching Kleros for freelancer disputes

Full transcript · August 12, 2026

Auto-generated transcript, lightly processed and pending a final human edit. Speaker labels are approximate. Every timestamp is a deep link into the recording.

0:30Hello, hello, hello, how are you?

0:33My beloved friends and citizens of the Kleros Republic, 12th of August 2026, how are you, Jean?

0:40Doing great and you?

0:42I'm doing great.

0:43Do you have a haircut or is it me?

0:46No.

0:47No, just uh another uh headphones.

0:50Oh, change of headphones.

0:52Okay.

0:53Okay, so we have a very special um day today at the community call.

0:58Um we have um one uh of the I would call oldest community members of Kleros uh and the only person that did the fellowship twice in uh the history of the fellowship.

1:13So um usually we don't accept people twice uh because uh we already have lots of applicants and uh we try to give room to everyone um But in this case, we actually had an exception because so this person uh she made her first research on fellowship about tokenization a few years ago, and then

1:36Uh last year we were interested in understanding a bit more about uh Portuguese uh arbitration ecosystem and how Kleros could uh you know, uh be used in consumer arbitration in in Portugal.

1:49So we actually this is uh we actually went to her to ask her to apply uh to to the fellowship again.

1:56So because that's how top of mind uh she is uh in the in the Portuguese um you know market.

2:02So well um if you don't know her, so she's Inês Bragança Gaspar and uh she's um uh well one of the oldest uh oldest community member of Kleros in Portugal I think.

2:13Um hello Inês, how are you?

2:16Hello, doing well, thank you.

2:19Tell us um before your presentation, tell us a bit about you.

2:22I mean, uh when how did you first listen heard about Kleros?

2:26Uh, how long have you been following project?

2:29Yeah.

2:29Tell us a bit about you.

2:32Um okay, so I I am a Portuguese lawyer.

2:35Um and when I was researching about topics from my master thesis back in uh 2018, 2019, uh I discovered about blockchain technology and a concept uh that was still very, very early at the time, which was the decentralized justice.

2:58I think that Federico coined that term actually, so I was writing about it before it was even a reality.

3:07And um yeah that's how I discovered uh Kleros that at the time was already um uh the more uh developed uh solution in the market, although it was still very, very early.

3:21And then yeah, I've been following Kleros ever since 2019.

3:27I participated in the 2023 fellowship and now Kleros invited me to participate a second time.

3:35And I'm very happy to to share my my ongoing research with everyone that is watching us That's great.

3:44Well, you can share your screen and then start the presentation and tell us about what your research was about so the community can know a bit more.

3:51Yeah, I mean the floor is yours.

3:53Okay, thank you.

3:55So on my ongoing research for for the Kleros Fellowship, uh I explored uh If and how Kleros could be integrated into Portugal's uh consumer arbitration system.

4:10Um I think you can see my second slide exactly Okay, um so the working title of my research is Decentralized Justice and Consumer Arbitration, Exploring the Integration of Kleros into Portugal's Alternative Dispute Resolution Framework

4:28I know the title is a bit long, but I think it captures very well the essence of the research.

4:35And I had five main questions driving my research So first, to what extent are consumer arbitration centers in Portugal used by citizens and how effective they are?

4:49To answer this question, I looked into statistics, publicly available statistics on the use of consumer arbitration centers and also different uh analysis and surveys such as um user satisfaction survey um then the second question was how Kleros could be integrated into the existing legal framework

5:16and if it could be integrated in the operations of consumer arbitration centers as an internal tool.

5:23So besides this option, I as I will show you, I also um came up with uh other alternatives for the integration of Kleros and then I think you all know how Kleros works, but how can Kleros' main characteristics be uh reconciled with the existing legal requirements?

5:44Um, and finally, I came up with um a strategy or like a step plan for engaging arbitration centers and regulators.

5:56So just to um start off with some brief context on the consumer Portuguese consumer arbitration framework.

6:05First, consumer protection is a constitutionally recognized right in Portugal.

6:12Um arbitration is not exactly new, so um it has already been incorporated into the Portuguese legal system.

6:21back in 1986.

6:23But there is an important uh distinction to make which is between uh voluntary arbitration where parties uh freely agree to submit their disputes to an arbitral tribunal uh versus a mandatory arbitration.

6:37In Portuguese it is called arbitragem necessária And it is a type of arbitration that is required by law and it allows consumers to initiate proceedings, uh arbitration proceedings without the other parties' consent.

6:54And uh in 2019, uh the scope of mandatory arbitration uh was expanded because before it was only uh allowed for uh certain areas um for example it's in essential services such as disputes related to

7:10uh water, um, electricity or uh gas supply and in 2019 this was uh significantly expanded to also cover uh reduced economic value consumer disputes So those um disputes that um have a value of up to uh 5,000 euros.

7:29And In these types of disputes, once the consumer uh submits their disputes to a legally authorized center, uh the company or the supplier of the product or service is obligated to take part in the in the arbitration uh proceeding.

7:47Uh moving on to consumer arbitration.

7:51Um Centers uh there are currently around 40 arbitration centers in Portugal uh and 12 of them are authorized to handle consumer disputes And um these centers are um private non-profit institutions.

8:09They are responsible with informing consumers of their rights and obligations and they are also able to uh resolve disputes to either mediation, conciliation, or arbitration.

8:24There in order to create a consumer arbitration center um it needs to have an authorization for the government and they are supervised by national entities and just a few um interesting statistics that I

8:41Further exploring on my research.

8:43Last year in 2025, around uh more than 12,000 new cases were filed.

8:49with consumer alternative dispute resolution entities and there was a growth of 17% of new cases in 2025 So these um upward trends are very relevant for for my research because it shows that uh consumer arbitration centers are indeed being uh used more and more

9:11And now just moving on to Kleros and before analyzing whether Kleros could fit into the Portuguese system, in my research, I also highlight um the cases where Kleros was already used uh to solve uh consumer disputes in the real world

9:30So as I'm sure all of you know, uh Kleros has um various subcourts.

9:36There is not yet a dedicated e-commerce subcourt But certain companies have already embedded Kleros in their own platforms to solve consumer disputes.

9:48One of the most known, uh well-known cases is Lemon Cash.

9:53So In 2024, Lemon Cash integrated Kleros into its dispute resolution process.

10:01So in practice, users they first submit a complaint through customer support like usually They file in a form describing the dispute, they can upload their evidence.

10:12And then if the user disagrees with what Lemon Cash decides, then they can escalate their disputes to Kleros.

10:22And then Kleros's jurors issue a decision within five business days.

10:28And if the user claim prevails, then Lemon Cash must comply with this decision But if Lemon Cash prevails, the user can still go to a small claims court.

10:41So using Kleros is just like an added uh layer And the results have been very interesting.

10:50I also explore a few of the statistics that were provided by Lemon Cash and there are also some other Pilot projects, for example, um in insurance, um a company called MetLife Mexico is also currently testing Kleros for insurance related disputes.

11:10And uh Maldo.

11:13Uy, which is a Uruguayan peer-to-peer services marketplace, is also researching how Kleros can be used to resolve disputes between freelancers and their clients.

11:25So based on my research, I have identified four possible models for integrating Kleros into the Portuguese consumer arbitration framework.

11:36These models are not mutually exclusive, so they can be uh implemented together or different arbitration centers can implement different solutions.

11:45So the first model would be to use Kleros as a fact-finding or evidence assessment layer so arbitrators could um Submit to Kleros Jurors certain factual questions.

11:58For example, was the product delivered as described?

12:02Did it arrive without effects?

12:04Was it delivered on time?

12:06Um the second model would be integrating Kleros directly into e-commerce platforms.

12:13So I also identified that in Portugal uh clothing uh and footwear and home-delivered meals were among the most frequently purchased products online in 2025 and normally these platforms use

12:30um chatbots or automated systems that have very limited human oversight uh to solve uh consumer disputes.

12:38So maybe integrated Kleros and having uh real human jurors, um, the signing on the disputes just like on the Lemon Cash example Could be one of the solutions.

12:51Then the third model would be creating a specialized digital arbitration chamber within existing consumer arbitration centers And the fourth model would just be a technical partnership.

13:05So instead of using Kleros to um be embedded into the dispute resolution um system.

13:15Um other Kleros uh developed technologies uh could be used to support these uh consumer arbitration centers.

13:24And then I also identified a few barriers to uh to integration.

13:29Um so this is the topic where uh somewhat research is still needed, um but I uh based um this on also some international uh precedents that I identified.

13:45So Uh Mexico in 2024 was uh the first country to expressly recognize decentralized justice in its legal framework.

13:55So actually the Mexican law recognizes um in its general law on alternative dispute resolution mechanisms, certain methods such as the decentralized justice, smart contracts, open collaboration systems, and automated processes

14:13So my first uh barrier that I identified, which is that the Portuguese consumer ADR law does not recognize the centralized or blockchain-based dispute resolution uh could be solved uh through a legal reform such as the one that was um enacted in Mexico.

14:30Um then I also identified two other um uh constraints or barriers that are kind of linked between each other, which is financial uh constraints um and also lack of digital literacy.

14:45In my report I have a few statistics.

14:48on the digital literacy of Portuguese consumers and also of Portuguese companies.

14:56I also identify in my report um a test uh pilot that is being done also in in Argentina uh with Kleros um And then um so the conclusion that I reached is that well none of this obstacle none of the obstacles that I identified are insurmountable

15:22And I developed this roadmap for a pilot integration.

15:27So the first phase would be to approach consumer arbitration centers.

15:32uh talk with them, map their needs, understand if they would be willing to to integrate new technologies.

15:41Um then the second phase would be to design and to test with only a few arbitration centers, the ones that agree um uh to have like a very specific and well defined use case and then after that test pilot phase

15:59uh was done, uh we would present the evidence to the competent authorities and um open the discussion on what is necessary uh to make this work.

16:12So that is all I have uh to present to you today.

16:17My research paper will be published, I believe uh soon so you can read more about it and if you want you can scan this code to uh connect with me on linkedin thank you Thank you, Inês.

16:34That was great.

16:36So people will be able to read the report very soon as you as you mentioned.

16:42I have a question after after seeing your presentation.

16:44Maybe I don't know if you have an answer, but maybe um there was a seventeen percent increase in cases sent to arbitration centers in Portugal.

16:52Do we know why this is happening?

16:55Is people becoming more litigious or uh I think there's probably two main reasons.

17:02Um the one is uh the first one is that There's been a big push in the last years for people to go through these alternative dispute resolution methods.

17:15So the government itself has uh announced as one of their um in their like strategy roadmap for for the well for this year for 2026 It was use um getting the citizens to use more the alternative dispute resolution methods.

17:36We can see ads on uh television uh saying do you have a dispute you can submitted to the consumer arbitration centers.

17:46So there's been lots of more uh publicity because not everyone knows about this when people have uh a dispute with a company, they sometimes think that their only way is uh to go through the traditional courts and those take many, many years.

18:02Uh there they have a big um uh case backlog and arbitration centers uh these types of arbitration centers are very very low cost or even free Um so there's been a big push for that.

18:18That is one of the reasons.

18:19And the second reason I would say it's because of the increase of uh people just buying more things and using more services.

18:29And so it's not really people being more litigious, um, but it's just um the increase also in in consumption of goods and services, which is uh what consumer disputes are.

18:43Um and um so there are different ways in which Kleros can implement it and you you mentioned them very well, like one of them is a partnering with arbitration centers, another one is like partnering with

18:55consumer uh offices like what we do in in in Junín, for example, they use Kleros for um assisting their decision making cases that involve consumer claims, set of claims.

19:05Other cases are like partnering with private companies like what we do with Lemon.

19:10Um and you you mentioned what do you think is more likely to happen in the short term, like in Portugal?

19:15What of those paths do you think uh is uh More likely to I think probably a solution like uh lemon cash, uh so approaching the private companies first would probably work better because

19:29well this companies are uh more used to or more comfortable in implementing these types of technologies Whereas um governmental supported um arbitration centers are probably not um focused on technology, they already have their own issues, they also have the financial constraints that I mentioned.

19:57So I would say a company and I identify some of these companies on my on my report.

20:04Um for example like um food delivering apps where they have I don't know, probably thousands and thousands of user complaints every day and they are being solved by chatbot.

20:18that are not really uh understanding what the consumer uh actual issue is.

20:24So I think it would probably be more likely for a private company.

20:29uh to um adopt one of these solutions um than the other um uh approaches that i that I identified other solutions All right, maybe I can ask uh one question in this.

20:45I think one something that I found really interesting about the report is that the arbitration centers uh can be um formed by arbitrators that are not necessarily nationals from portugal

21:00So do you think that this could allow some kind of um setup in which arbitrators from different parts of the world that might have more you know detailed expertise in some kind of subject and you could have for example um arbitration centers that deal with specific uh disputes

21:25And they can tap on specific pool of arbitrators, um, and this becoming like a more specialized arbitration uh setup.

21:34Do you think this is something that could be Done, do you think it's um it could be useful?

21:41Mm-hmm.

21:42For sure.

21:43So um Uh any individual with full legal capacity can be appointed as an arbitrator regardless of of nationality.

21:53This is according to the law.

21:55But the parties to the dispute can agree that arbitrators must possess like specific uh qualifications or expertise.

22:06or um um experience.

22:10So I think that as um the digital economy evolves and uh people begin purchasing uh more goods and services online for example and even types of services that maybe are not so common today

22:28So as disputes become more digitalized, and this is like uh one of the questions that I've been researching ever since my my master's in 2019 is um if disputes are becoming more uh uh technological and digitalized, why are we still solving these disputes through the traditional routes with uh lots of paper and uh going to court in person?

22:54So it doesn't really make any sense.

22:56So yeah, I definitely see that specialized um for example specialized e-commerce chambers um could be created and this has already been um this already exists um in other countries.

23:12Right now in Portugal there is not a consumer arbitration center that is only specialized, for example, in um in uh in e-commerce for example.

23:24So yeah, I definitely think that uh might happen in in the future.

23:30Interesting.

23:31Inês what what what does the the law say in Portugal about an anonymous jurors for the cases So that is uh well that is not possible.

23:42I also talk um about um about debt um in my um in my in my report um because um The jurors need to like sign um the arbitral decision.

24:02Um and uh Yeah, but uh that is something that we could further um further explore.

24:12Um also seeing how this is being done uh in the already uh in the currently underway uh test pilots in in Argentina I think uh and to see how that uh requirement um was was approached with the with the competent authorities.

24:35Facu can explain how this works.

24:37Use a KYC project process for jurors or yeah explain Facu how this works in Argentina.

24:44She's frozen, I think.

24:46Okay, I can explain.

24:48So in our are you there, Facu?

24:51Yeah.

24:51I think he's back.

24:52I don't know, you're back.

24:56Explain, can you explain how uh juror selection process works for Junín, you know, with the KYC and the soulbound tokens?

25:02Because yeah, um it's not anonymous anymore if in case it's needed.

25:07Yeah, exactly.

25:08So in Junín cases, uh there is a specific court in Kleros Which is formed by um jurors that are lawyers.

25:20And the way we do this is by using SBTs, soulbound tokens.

25:25that it worked as a as a kind of digital credential and basically um this court um basically prevents from anyone who does not have this SBT to be able to stay in court and be able to participate to become a juror in this court

25:44Uh so the thing is, okay, how do we um KYC these jurors and how do we uh make sure that We provide this SBT to the wallet that we know is under control of a person that meets these conditions

26:02and that um we know who we who this person is.

26:06So basically with this can be done in in different ways.

26:09Now it's it's being done in a more centralized way by the Kleros Cooperative.

26:14So it's kind of a KYC process, and you um as a condition for granting the SBT, you also require that this uh juror agrees to certain uh terms and conditions Um yeah, so this also provides another layer of um of uh yes of confidentiality specific

26:38This can also be done in a more decentralized way as as we move forward, and that's the that's the the idea.

26:45But this is more or less how it works right now.

26:48And this allows us to have a more Yes, allows it allows us to have courts that we know that these jurors meet certain conditions.

27:00So we can we could do the same with um with Portugal.

27:04uh with the consumer arbitration centers.

27:06Yeah I also explore a little bit on the on the soulbound tokens, yeah, how how how arbitrator centers could also um issue these soulbound tokens to the already existing arbitrators and the soulbound token could have embedded

27:25the experience of the because normally the arbitrators are also uh lawyers or professors of law or judges.

27:34or uh jurists so normally it's already people um that have a certain experience and that are very recognized in in specific areas of the law There you go.

27:49So I mean the the early version of Kleros that I mean you knew in 2019 that when you first approached the project It's still there in some way because the Schelling point is still there.

28:00There still are course like based on anonymous token, uh you know, uh weighted voting and all that.

28:06But you know, also the protocol is way more flexible than what it was at the beginning.

28:10So this And this makes it uh able to address cases that were not possible in the beginning.

28:15Uh for example, consumer arbitration in in Portugal, so that is We hope that your report is going to be like the first step for more adoption in Portugal.

28:25So I'm super happy that we we got to to have you do another fellowship with us.

28:31So yeah, I mean and the the report is great.

28:34So I recommend everyone to read it if you are into consumer arbitration, not only in Portugal, because you can uh still apply because the batch The 10th cohort uh applications for the uh fellowship is open now.

28:46So if you are into this this the goal was this to motivate you to do this.

28:51same type of research in your country, in your jurisdiction to see how Kleros can be implemented in whatever country you happen to live in.

28:59Uh mean apply because this is the type of research we want to um foster for having higher higher adoption.

29:07Inês, I mean, no just I highly recommend the Kleros Fellowship as the only person that has participated twice.

29:17I highly recommend it to anyone that wants to explore about the possibilities of decentralized justice, just as Federico said, in your own country or in a specific area of law And uh you don't need to be lawyers like me to to do this research.

29:34Uh some of my other fellows um came from they were like uh mathematicians and economists.

29:43So yeah, it's open to everyone that is interested in this area.

29:48Inês, thank you very much for coming and your report will be very soon online for everyone to to read, okay Great seeing you.

29:58Thank you.

30:00Okay, well let's continue now with uh hi Jean.

30:04What do you think?

30:04That's amazing, right?

30:06And she's great.

30:08You're mute.

30:10Yeah.

30:12Maybe for the I was thinking that maybe for the next uh fellowship we need to Uh create a rule that Inês will not participate.

30:22Otherwise we will keep inviting her A cura a curated list.

30:27I mean I mean she she can definitely I mean hopefully we have more topics for her because she's she's great.

30:31So like uh yeah she's the the OG of the fellowship now.

30:36I mean the the only one that you know got like a two Uh and and we will bring her at some point to to speak a bit about the other topic that was um real world assets and tokenization, but uh so that's for another discussion.

30:50Uh there are so many use cases for Kleros that you know uh You can stop.

30:54Yeah.

30:55Yeah, I was thinking about that too because that area right now it's like uh really growing a lot.

31:03Almost I would say one of the main areas with blockchain right now.

31:06So I think I think Inês also ha have a really really good uh taste in topics because it's uh really good uh research.

31:17So okay, we have William here.

31:21Hey William.

31:22Hey buddy.

31:23Hello William.

31:24How are you?

31:26Good.

31:28Nice talk.

31:29For those who are not here, I mean Monday he delivered a talk about privacy in Spanish, in the community call in Spanish.

31:36So that was an awesome performance, William.

31:38Good job.

31:41Um well tell us what what um this week's topics, uh yeah, research.

31:46Yeah, go ahead.

31:47Yeah, so the the talk Line Monday was about confidential data and like how can a system like Kleros handle data that's private, evidence is private.

31:55I gave the same talk basically last week here in in English in this community call.

32:00And uh very soon there will be a blog article version of these ideas, uh which is even like a slightly longer version of like the toolbox and different ideas, what can take.

32:10and goes a bit more into depth so you'll be able to read this soon.

32:13I I look forward to sharing this with you.

32:15Uh so I mean though that's that's the main thing.

32:17You know, we're we're still, you know, like thinking about our various like AI research.

32:21There are AI aspects in the privacy things as well.

32:25Um, but uh, you know, that's that's sort of like the main um theme for this past week We just published today, uh G maybe you can share the note here on the on the screen uh an article where William also contributed a bit about AI and and Kleros and our thesis of uh how AI and Kleros will interact.

32:46are interacting already.

32:48And the thesis that we have developed for like 10 years already, you know, the first time we wrote about this was uh in the Doges on Trial Um well um uh experiment we did and then after we did that we um ran some tests with like AI.

33:05face rec or I don't want to say face recognition, but dog recognition algorithm to see if they could identify the dog or not dog.

33:12And then that was twenty nineteen, and this is when we developed our our first thesis about how Kleros and AI will interact and so lots of things happen after that.

33:23Harmony as using AI as a mediator to onboard cases.

33:27Uh d and then AI models as judges, for example, like uh uh you have different AI models that make different decisions.

33:35about this, about cases and we ran an experiment with uh with Lemon that uh was published a few weeks ago by Wolters Kluwer um depends on what model you use, you will get different answers.

33:46So this is a really, really long blog post, you know, where we summarized a lot of our research and how I we imagine the the mechanism to work in the future and we have like a three main things I think I would say our thesis about AI has three main topics.

34:01Um one of them is that for high complexity cases and high nuance where there is subjectivity, there is um the need of understanding the moral context of some decision.

34:13we we will not have uh I mean this is still going to do but be done by humans because they are the th those who only can solve these cases.

34:22Then for like more like lower value, middle complexity, we will have a combination of uh AI and humans and uh we will have like uh this AI models making decisions and we still need to figure out

34:34what is the right way to do it because maybe you have like one first decision made by AIs but then you have to appeal to a human because uh you don't want an AI to make just a final decision uh that could be dystopian.

34:47Um and then um for the the third main thing of the thesis is that we are about to enter a part of the economy uh where we are going to see um disputes uh of a nature the humanity has never seen before like uh agent to agent disputes

35:06uh agent to human disputes and disputes that will need uh extremely high uh like uh velocity of resolution um and this is um something we are exploring now um and so Dispute between an agent, suing another agent, going to a panel of agents, making a decision about that case through an escrow or through a reputation system.

35:30We are Very much exploring how escrows work in these type of settings and how reputation of agents will work um in the sense that uh agents could use the reputation as a collateral for asking a loan.

35:42Imagine like a very high velocity and high frequency loans from one agent to another or like a Uh working capital loans.

35:49I mean I will give you this, but you pay me later, but hey, how do I know you will pay me later?

35:54So you could put your collateral as a reputation in some way.

35:59I mean there's tons of stuff we are exploring and you will see very soon uh announcements about this and we there are some things that are being uh uh cooking now uh as we speak.

36:11Uh so yeah this is this is I mean something we have been working on for a very very long time.

36:17We just needed in some way the world to catch up with some of these ideas which seems to be happening now.

36:24I don't know William, any any comment?

36:27From the point of view of mechanism between like agent to agent dispute compared to agent to humans disputes.

36:33I mean what what why should this be different I mean the sort of the big question is if there is a difference.

36:38Uh like on like the uh on the behavioral economics, do do the agents act differently than humans?

36:44I mean they're trying to on a human text.

36:46Uh so um, you know, maybe that all their sort of like weird behavioral quirks, um, you know, partial rationality, like sort of like social effects are the same, you know, and that they're like behavioral economics very similar to humans.

37:00Or maybe they're more homo economicus and they kind of, you know, act in like kind of profit maximizing super rational ways, somewhere in between.

37:08I like I think that the like the behavioral economics research on AIs that are you know this is an emerging field uh has um like it depends a lot on the model and like you have sort of results that are like you know sometimes they act more you know like very similar to humans, sometimes they're more a little more homo economicus as a

37:25You know, we've thought about what experiments we can do in the context of Kleros.

37:29There are sometimes like very specific things that like you, you know like the AI that are very special to AIs, uh that like, for example, in the talk I gave about privacy you know having like an AI that has some kind of like material component so that it's just incapable of of sharing information except in like certain ways.

37:50That's you know like you you can't you can't do everything here.

37:55You know, like to what degree are those like very limited but you know important differences important?

38:02Um, you know, what how like how are we going to see those evolve?

38:05Lots of open questions.

38:07I mean, like, do you think AIs, for example, will have uh biases in decisions like loss aversion bias and sort of things we see in like behavioral economic stuff?

38:19I mean it depends on how they're trained.

38:20Uh I I think that the research on on behavioral economics of value is it seems to indicate that yes, it looks just something's fun.

38:27One of the one of the things that we discuss with William usually, I mean, uh is like for me one fascinating topic is um you know this ultimatum game where you have like two parties one of the parties uh has to uh make an offer to the other one they have they have to split 100 dollars and then the receiving party needs to accept or reject if it accepts

38:48It gets whatever is offered by the first one, uh, and if it rejects, both get zero.

38:53So the rational theory, you know, like uh um Rationalization making theory, econom basic economics say, okay, if the guy offers more than one dollar, I mean the second party I mean should accept because this is the rational thing is one is more than zero.

39:09So yeah, okay.

39:11Take it.

39:12But what what people see empirically is that in these experiments about the Ultimatum game, they usually don't accept I I think the the average is like they don't accept anywhere like be under 30%.

39:25uh offer uh under thirty dollars they just go and screw you you're going to give you like that so little uh you got zero now.

39:32Uh this is what they I think that this the big exception to this um is in some particular tribe in the Peruvian Amazon called the Machiguenga, uh about which Vargas Llosa writes a book called El Hablador.

39:47Uh it's a tribe where they apparently is the only one that behaves um in the predicted way of rational like uh neoclassical economics where they do accept the hey one is more than zero so why not it's free money okay

40:03So I mean but this is the only one and because there is the thesis that they do, they they have these researchers and this this comes from a uh book on behavioral game theory.

40:12Um uh that they they made this all of this uh research of different tribes in different countries, different cultures, and then what they find is that um apparently this tribe they don't have

40:26uh like any kind of market integration.

40:28They don't have like a market economy.

40:30They are like only like families that just have reciprocity ties between themselves.

40:35So any uh interaction with the outside is uh almost non-existent.

40:42So they um the so what and the thesis they have is and this these are anthropologists that make this type of research is that they Like when you have a maca when you operate in a marketplace and you have to deal with people on that, so you start to develop some like uh

40:59understanding or some ideas about uh reputation, about reciprocity ideas between uh business partners And if you offer one uh and you and this results results on ideas on fairness, if you offer one of one hundred, you will be seen as a cheap guy.

41:15I mean uh like uh yeah low reputation guy.

41:18So you would fear people will not want to interact with you, right?

41:23Um so like you have to start to so this is the thesis.

41:26I don't know if it's true, but the the question is I don't know, like uh will uh AI agents also reject Um, you know, an offer that seems to be higher than zero, but less than what they would think is fair, how much would they think is uh is fair?

41:42Uh I don't know William, but what do you think?

41:44But what what what are we going to see here?

41:47Yeah, I mean I I agree with all that.

41:49Like I I think to the people who've done this research, like in some cases the AIs reject you know the very low offers.

41:54I mean ultimately this comes to like Notions of justice.

41:58And um, you know, like what notion of justice the AI has sort of simulated.

42:04Yeah, I mean so now we yeah, go ahead.

42:07Well one interesting thing is uh is that probably the AI has access to the uh entire uh Like uh all all the the books about game theory, all the experiments, like all the research about the game theory, and they shouldn't expect the other AIs to also have the access.

42:27So Maybe they would that will be the equilibrium.

42:36Because that's not how people think.

42:40You know, this is I don't I'm I'm I'm reading now, you know, this uh Daniel Kahneman book about uh actually bugging William with lots of you know uh common common knowledge situations.

42:51It's about like uh how coordination works with lot of and they they do lots of like uh experiments I mean Steven Pinker, um not that it's a Steven Pinker book.

43:01Um so and he does like um all of these experiments with okay, what happens if You know, you now you have these people these people want to go to a restaurant.

43:09They prefer to go with each other, but they also have a preference for the restaurant.

43:12So they don't these uh they have lots of these little games.

43:15of game theory of how they coordinate and the and what one thinks the other thinks that I think that he thinks.

43:21Uh so all it's about all of those um things and the chicken game and all that and how common knowledge Is helpful for well society to work basically in a in a more or less efficient way.

43:33So yeah, so uh these are the kind of things that are in a but uh the question is okay, those these are general topics about um well game theory and agents uh of which we used to think at the beginning when we were starting our research that

43:50uh uh AIs would like behave like as a perfect like John von Neumann, you know, Oskar Morgenstern uh assumptions of perfect rationality.

43:59We don't think it's the case anymore.

44:01How will this affect decision making in Kleros.

44:04I mean when an agent thinks how other agents are going to decide a case, even if they um think that the Schelling point might be A, but maybe the fair thing to do would be like B in their opinion.

44:18So like uh this is a the type of topics that we we have uh very much in our head now.

44:25Um I don't know any any thoughts about you know this?

44:28I mean would a both vote for a decision do you think is unfair Just because it will make more money or maybe they will like prefer to not.

44:35I don't know.

44:36What do you think?

44:43Like I still think if you have an adversary that like thinks that their perfectly rational tool uh is like you know the best way to exploit some vulnerability in the system by being like super rational and finding some like very small thing to zoom in on, uh they will be able to do that.

44:57So I mean the system needs to be robust.

44:59against hyper-rational actors.

45:02But you may have a bunch of like semi-rational actors that are kind of the average, average, honest user.

45:10AI is very good.

45:11Okay, so so guys, if you want to research this type of topics with William and with me, I mean do apply to the fellowship, AI, decision making, Kleros.

45:20You can decide do cool experiments about like okay what if I use cloud to decide this uh dispute about consumer about sending someone to jail I mean I mean theoretically you could use it if you want and it's fine

45:34Um ChatGPT, how it compares with like uh Gemini, how it compares with so this is the type of topics that we are trying to research now.

45:42Uh so the fellowship is open.

45:45So you can go from researching consumer dispute resolution in your country to like a very like uh cutting edge uh dispute between agents in like a virtual reality world that is going to be resolved by a panel made of agents and uh yeah from different uh you know models and whatever.

46:04Anything else William you want to comment today?

46:10Thank you, William.

46:11Okay, Jean, what do we have now?

46:13Um anything else?

46:16You have like news uh about the project?

46:19I don't think Uh I don't have anything special besides stuff about the um since we mentioned about agentic economy.

46:27Yeah.

46:28Yeah, let me share uh the page about the rewards that we we published a video about this uh this week So here it is uh loading.

46:39Basically we have a couple of uh reward programs.

46:58We have the Scout Rewards program that uh rewards uh submissions to the Scout rewards so for uh people get rewarded for tagging new contracts and new uh tokens new pairs of domain and contracts

47:15And this helps us like uh help block explorers uh have more information about what people are uh using And um we also have rewards for people register in Proof of Humanity.

47:30I think there's uh limit of people, but the if the first I think a thousand people get uh uh like an amount in PNK.

47:40And one interesting thing about the current rewards specifically is that uh Marino Marino uh created a page that shows the information in a way that uh makes it easy to like understand how these rewards are uh becoming uh entries into the registry

48:01And these entries of course correspond to contracts and and tokens registered and basically the cost that we have for getting submissions.

48:11So if you can see The cost is going uh down.

48:17Uh and there's another uh like uh thing in the gym.

48:22That could be people using more AI for this, right?

48:25So in the case that they Can uh do the sourcing of information through AI and this brings the cost down, right?

48:33Yeah, I'm pretty sure that there is something like that involved.

48:36Maybe for finding the contracts or uh making like pre-checking Because right now it's a lot easier.

48:43Uh remember that at at some point the it was very easy, for example, to get the token uh image on the right format.

48:52And this is something that maybe you had to use like Photoshop or something like that and right now it can basically executing Photo what?

49:00I don't know.

49:04You know, imagine in the back in the day you used to send the submissions with a pigeon, you know, for the sum you know the challenge and all that.

49:10You know, like now look at this, everything is so fully automated.

49:13I don't know what people are going to do for a job, but you know, like uh this seems like uh this This looks to me like a AI economy at work, you know, like uh uh lowering lowering costs of finding information, submitting information.

49:28uh by a lot you know uh 3.

49:317x four four x down above up and this is i mean for the for the registry is great because you can source information and data way cheaper But yeah, I don't know.

49:41So I mean some other day we can do like a special call about uh agents and employment, you know.

49:47I don't know Yes, and and I think uh one way of thinking about this also is that the submitter uh is not like the uh the manual worker anymore it's like uh the the manager of the agents so maybe one submitter in one hour can make uh I don't know uh

50:0850 submissions and before they maybe needed like eight hours to do the same thing or or or or more.

50:17And And one thing, this uh we released the scout skills I think two weeks ago, three weeks ago, so we will see the effects of that.

50:27Um on the on the coming weeks and and months.

50:33I'm starting to check the submissions myself with a bot, so maybe other people are doing the same thing.

50:40So I think this this will be pretty interesting to to see.

50:45And another uh comment about this is that uh on the ATQ registry we actually tag like lots of contracts in one single entry.

50:54So there are for example a couple of entries that tagged I don't know um one million contracts or something.

51:00So um And there are these are more complex to check because it they require like special knowledge of JavaScript and uh like some technical knowledge and this is something that maybe uh someone that couldn't check before uh now with the help of the a the AI they can they can do

51:21Uh so yeah you you could even try to make a submission without being super technical.

51:27So what you're saying is that if I'm not If I'm not a technical person, I can still make money like through submissions to the registry because it barrier of entry is like lower, right?

51:40Definitely, definitely.

51:41You can make money by making these submissions, by challenging, by uh reviewing, for example, these ATQ submissions that are technical, but maybe your bot can find an a an error.

51:53tell you to challenge it.

51:55So yeah, definitely.

51:58I will tell you know my aunt Rosita from Curuzú Cuatiá that she can make uh some extra money like through making some agents, uh making submissions and and claiming rewards uh i'm sure that she will say oh yeah my friends are all all already doing that you know so uh but yeah okay the the the the I think the generation

52:21Yeah.

52:22Imagine someone that for example has a like a a ChatGPT account, uh ChatGPT submission, and then I don't know for a specific month if you didn't use it a lot.

52:32and you have like extra usage that is going uh you're going to lose because you're not going to use it, you could basically tell uh GPT to try to do this.

52:43So there's a chance that you can find like solutions.

52:47Or what if I what if I rent my unused GPT capacity to you uh through some like tokenized uh derivative product and then we split you know whatever earnings you make with my and use computing power you know like uh but that's more a more interesting financial product right

53:04Okay.

53:05Just put DeFi into into the pro into the process, right?

53:09Yes.

53:12And then you know tokenize and tokenize, you know, like a CDO based on that, you know, and the unused capacity.

53:18I'm sure someone has uh it's already doing that, you know, like uh yeah, some options, you know, markets, you know, derivative.

53:25Okay.

53:25Yes, there is there there is actually one at least one project doing that.

53:30Oh really?

53:31Oh I it never fails.

53:34It never fails.

53:35Okay.

53:36Um what do we have?

53:37Anything else, Jean?

53:40I think these are the main things, but we do have like more materials coming out on our on our channels.

53:50So as we mentioned on a couple of calls, we have uh Instagram, TikTok, uh YouTube Shorts for short videos.

53:58And the the accounts are growing, but everything always helps.

54:03So if you can follow us and give some likes and ask your friends to follow us, everything helps.

54:08And we are also and everything.

54:13Comments, yeah, everything, everything.

54:16And one thing uh that we started doing uh recently, we started posting some kind of digests of um of mentions of Kleros, quotations of Kleros on academic papers, articles, and all of that.

54:33And we are starting to receive She received like some good feedback from universities and the people that wrote the papers.

54:42So this is pretty interesting too.

54:44Uh so we'll probably post them weekly, uh but lately we have been receiving lots and lots of mentions.

54:52So maybe we we we need to uh publish also about the individual articles but but yeah really happy with with that.

55:02Yeah and also we are going to to revive uh our account at R ResearchGate, you know, we had an account at ResearchGate is like social media for academics and we we I opened that a long time ago but we didn't really use it that much but now our academia.

55:17org we all also have one there Uh we all start, you know, because there's so many people researching about Kleros and and we don't even know.

55:24I mean we we we only catch those that are on the uh Google Scholar network.

55:30But the what what there is lots of like master theses and lots of people who are to research that doesn't get tucked into the like a big journal, you know, circuit.

55:40Uh but it's going to be cool, you know, because then Lots of people doing their master thesis are doing like a Kleros for XYZ, you know, like um things that we maybe we never thought about.

55:52Like for example, Kleros in the horse industry could you use Kleros for the split resolution between buyer and seller of a horse?

56:01Well guess something you can and we actually found out like uh last week so uh that's something we didn't see coming i can you i imagine how many people are doing like uh master thesis llm thesis or whatever other thesis that like uh since it's not tagged into the

56:20journal you know traditional tagging system and go under the radar and maybe we want to know about that.

56:26So this is going to hopefully we will catch a few more uh in academia.

56:32org Um I don't know anything else, Jean.

56:37No, I think I think that's it for today Thank you guys for coming and see you next uh week at 6 p.

56:47m.

56:47UTC for the community call.

56:49Bye bye.

56:50Bye everyone.

Want to be in the room? The Kleros Live Stream runs every Wednesday at 6PM UTC, with the Spanish call on Mondays. Join the next one, or subscribe to get these recaps in your inbox.