Can mods do something? Decentralized content moderation for Web 3

Transcript

hello everyone uh all right i'll get started about now so my my name is shotaro granzier nakajima and i'm a developer at claros claros is a dispute resolution protocol it's a sort of on-chain court so i'm very happy to be here at ecc i attended last year as a student and now as a speaker so i'd like to thank each cc for supporting students that's how i am where i am now and yeah just supporting ethereum ecosystem and beyond so

i will talk about content moderation so does anyone recognize this photo this photo is of the infamous january 6th u.s capital insurrection these are trump supporters climbing the walls of the capitol there are still ongoing congressional investigations into the exact circumstances under which this uh insurrection and incitement of violence occurred but one thing is clear that social media played a very important role in the insurrection in particular many people cite donald j trump the u.s president at the time and his twitter activity as a

contributing factor or one of the main causes to organize and rally his supporters for the capital insurrection and as a result many web 2 companies social media companies including twitter soon followed by facebook and others decided to permanently suspend uh or ban trump from social media platforms so this begs the question should king jack or king mark have the power as unelected officials to censor politicians from their platforms as you know political discourse is moving to the digital town square should these unelected officials have

the power to censor anyone they choose on the one hand private companies well they can choose which which customers they provide services to but on on the other hand well censorship is bad and it's bad for democracy so who in short who moderates the moderators so they're very interesting very interesting efforts that have been happening at facebook they've introduced a sort of supreme court which has some independence to attempt to put more transparency into their content moderation by having very contentious cases being solved by this allegedly independent uh

supreme court there are challenges though um this this court has only solved a handful of cases in the past year um and what if there were a natively transparent credibly neutral web3 native system to moderate contentious content at scale well meet claros so claros is a disputed resolution system it's sort of like an online court so a typical example let's say alice is a client in france and bob is a graphic designer in guatemala they agree on a contract they agree on a payment instead of paying bob directly alice is

going to put the payment into a smart contract that acts as escrow where in case they have a dispute claros will be the arbitrator when a dispute is created jurors are randomly selected these jurors deliberate on the evidence and they vote if alice wins then alice will receive a refund so how how can we use claros for content moderation well the idea is that if there's very contentious uh if there's a very contentious case over constant moderation well a case can be created in the claros court and jurors can decide so how how can we do this specifically there are many ideas to unbundle

uh user-generated content platforms from content moderation so the idea is that claros can offer content moderation services to many existing platforms and the platforms don't have to nationalize the content moderation themselves content moderation is very difficult to get right so i mean lots of startup founders jack dorsey mark zuckerberg they don't want to be the ones in charge making these very high-stakes decisions when they would rather be building and that that applies for a lot of crypto and tech startups as well they don't want to spend all of their time moderating communities and so there's there's a few types of content moderation and in particular one effort is to uh address fake news and

misinformation and so it seems kind of unclear how to do this properly but at least one thing is clear that whatever system to moderate fake news and misinformation should be open source and easily verifiable so meet uh pithia ptr is a curated list of fake news on twitter its focus is crypto twitter influencers um and it was started by matthias uh former developer at claros so how does it work any user can submit a tweet including a url to the tweet and a screenshot the screenshot is uploaded to ipfs in case the tweet is deleted

and the submission must be accompanied by a deposit about 40 die the tweet can be challenged during a challenge period of about four days the tweet is challenged then a dispute is created in the claros court if the tweet is not challenged then it's accepted into the curated list of fake news so here's a few examples um this is from this is a tweet from jimmy ragosa of cismo from one year ago at etc and it reads breaking news eip 1559 has been rescheduled to 2022 as most ethereum core developers report quote massive rob a hangover unquote a champion of the eip was heard this

morning at etc saying the burn i just need water right now so someone added this tweet to pithia and by show of hands how many people think that this tweet was accepted to this fake news curated list does anyone think it was accepted okay who thinks it wasn't accepted all right so it seems a little bit split but i think some in the audience already know the answer well so in the end the tweet was rejected and the tweet was rejected because the policy states that the misinformation and the tweet containing fake news should not be humorous it shouldn't be a

joke and in context we can read this tweet and see okay jim jimmy is clearly making a joke he he doesn't mean this seriously and most reasonable people reading the tweet wouldn't think he's being serious so here's here's another example the tweet reads cardano is completely useless technology and charles hodgkinson is basically a cult leader so someone added this tweet to pithia and yeah again just just by show of hands how many people think that this tweet was accepted to the list of fake news oh okay just just a couple hands interesting how many people think this tweet was rejected from a curated list of fake news

okay more hands more hands so yeah this this tweet uh was challenged and a dispute was created in the claros court and parties submitted evidence that well cardano does have some use cases and it's not completely useless technology so it was rejected so i'd i'd like to highlight i'd like to highlight um what these tweets demonstrate you know with content moderation there are many use cases that are better solved by ai for example filtering images that contain gore or sexually explicit materials but there are other content moderation use cases where there's more nuance and subjectivity involved

the tweet needs to be read in context who is charles hodkinson what is cardono what what is an eip what is rabe so this demonstrates a use case where you need humans involved making subjective decisions so this this is where claros shines and if if you're interested and want to follow this uh experiment in curating fake news you can follow pythia check this is the twitter handle where pythia submissions will be tweeted out and any challenges will also be tweeted out so so you can see which tweets have been accused of being fake news and so there's there's another use case

for content moderation on the one hand you have fake news curation but you also have toxic behavior on large groups so most of twitter community or crypto communities are organized on telegram and discord channels and many of these groups have thousands and thousands of members and as as i stated founders of crypto projects they don't want to spend their time all day moderating toxic behavior on their community channels and also community managers don't want to be the ones making these high-stakes decisions about which members should be accepted or rejected and it's also becomes very contentious

when you're talking about governance and dow governance and if if users have some toxic behavior on a governance forum or a governance channel does it become more political about which users are able to participate and how do you balance free speech from preventing toxic behavior from taking over your community so we're introducing something called claros moderate and clearos moderate is a way to moderate social media channels initially working with telegram but also working on discord to have fair and transparent moderation so yeah so how does it work um so cleros moderate uses reality.ethe and the

claros court on nosyschain so let's say there's a user who has some toxic behavior and either a permission set of moderators set by the group configuration or anyone in the group can report the toxic behavior when the behavior is reported a question is created on reality so this is reality and anyone can answer the question about the toxicity of the user's behavior and whether or not it broke rules of the group they have to when they answer the question they leave a deposit of some die in this case 10.

any user can change the answer by escalating and doubling the bond they can change the answer from yes to no yes the user did break the code of conduct no they did not and the preliminary answer is enforced so the preliminary answer the user has some penalty maybe a 24-hour mute or it could escalate if the user has already been warned or or been muted in the past the penalty could escalate at the same time anyone can request arbitration anyone can dispute the answer and they can do so by creating a dispute in the claros court when a dispute is created again jurors are randomly selected and then they deliberate on evidence and they vote

anyone can also directly add evidence from the telegram group or discord channel and when they add evidence the message in question is uploaded to ipfs in case the message is deleted to preserve the evidence and this evidence is also automatically displayed in the claros court so and when the when the case concludes then the penalty is enforced either the user has their reputation cleared or the user is uh will have some penalty for their toxic behavior so uh we're going to start rolling out claros moderate to select partners depending on their community needs so if

if you've had a community um a crypto community or otherwise that struggled with moderation and how to moderate at scale please get get in contact with us and we can try to have you know see what your needs are and what your community needs are and how clearos moderate could help so there's this qr code will take you to an interest form and if you want to learn more um go to claros.io backslash moderate and get get in contact with us we'd love to help help you with your communities and community management yeah thank you [Applause] how much time do i have

five minutes for questions hi thanks uh i'm just wondering i can see the logo of the european commission behind you do you do some projects or you have to receive some funding or so as as far as i know um claros the cooperative received a grant from from the european union horizon funds for um i don't know breakthrough technologies so uh this was like a one million euro grant by the european union federico in the audience can answer i'm the founder of clairol and so it was a prize called horizons 2020 blockchain for good and we presented up our project and we

split the price with the winner of the other winner and it's 500k at each euros it was for yeah for for the project not to build something specific it was for the project that was already built basically hi thanks for the talk um is there the idea that um if for example in the channel in discord on telegram uh i insult someone and and the clearer case uh punishes me is there the idea

that this uh punishment can propagate also to other channels if i i am in other the discord server or groups or stuff like this and could impact then my reputation and ban me from other groups or stuff like this yeah that's that's right it is possible to have um even cross-platform and cross-group persistent reputation um but it's it's up to the groups about which which rules and which groups they want to uh also um um respect their moderation platform so different groups can have different rules um but yeah it's it's possible that your reputation can follow you yeah

these are experiments with reputation systems negative reputation systems hi uh do you maintain some um some kind of uh irreparation or is it a one-shot judgment every time sorry could could you repeat the first um do you maintain an e-repetition or is it a one-shot judgment every time um so so there's no explicit persistent uh reputation um so each each time the uh the user behavior is judged on the evidence alone and past history of the

user is not that doesn't affect their on ongoing case all right thank you [Applause]