A European Union-funded initiative called Eunomia has contracted various thinktanks, tech groups, and Fediverse evangelists to help with a new social credit scoring system. The system will be applied to social media and the federated destinations using blockchain technology. It will be governed by a crowdsourced system where groups rank up and rank down the trust levels of information. In simple terms, it’s an adaptation of China’s social credit scoring system for the Fediverse. Well, someone working on the project leaked some chat logs, and it reveals a rather antagonistic nature from the people working on the project toward the users who are skeptical of this EU funded project.
For those of you unfamiliar with this project, it’s explained over on the Trilateral Research page – which is one of the groups being paid to help develop the program, €280,437 to be exact – where it explains…
“The primary objective of the EUNOMIA project is to assist users in determining the trustworthiness of social media information using an intermediary-free open-source approach. EUNOMIA empowers users to take ownership of the social media disinformation problem by providing a technical platform and software tools enabling users to vote on the trustworthiness of each social media post. The project will address the challenge of disinformation in social media from three different angles: by establishing the original source, identifying how information has changed in an information cascade, and helping the user evaluate the trustworthiness of the information.”
According to a Purdue University report, the system will help journalists identify sources of news and curate information for aggregation services like Blasting News to prevent the spread of “fake news”.
The project will rely on machine learning and other systems that will extract data from participating social media APIs, which will include services like Twitter and Mastadon. It states…
“ Trilateral’s approach to participant observation will involve applying network analysis, natural language processing, and machine learning to gain insights from data extracted via social media APIs (e.g., Twitter, Mastodon) on the social (e.g., age, gender, nationality) and political factors.”
This obviously didn’t seem like an innocuous project from the EU especially given the Union’s recent anti-freedom of speech initiatives like Article 11 and Article 13. In fact, leaks from someone close to the project has revealed a little bit more about how the project will develop and some of the insights into how it’s coming along.
It was originally commissioned by the EU back in December of 2018 for €2,942,780, but it’s not expected to be finished until November 30th, 2021, according to the fact sheet over on the European Commission website.
The fact sheet also reveals that Mastadon creator, Eugen “Gargron” Rochko, a mixed Jewish and Russian programmer from Germany, is currently being paid €63,290 to contribute to the project.
So what is Eugen’s role in the Eunomia project? Well, according to the leaks, Rochko is working with engineers to build this social credit system that will attempt to weed out untrustworthy individuals and suppress “fake news”.
The leaks were posted up over on the GNU Jihad website, where they break down some of the conversations that were snapped and saved from the chat log for the Eunomia project. One of the initial concerns they discuss is how people might feel about the data collection and user info gathering that will take place, to which Rochko warns that telling people that they haven’t started gathering information “yet” won’t go down well with the general public.
If you’re unable to read the image, Ryan states…
“[…] Generally there is a lot of paranoia that is difficult to address. All we can do as a project is be absolutely transparent about everything we do.
“even in the case of datasets we generate and how our algorithms work etc. An interesting challenge to say the least, which of course will be open source as well.
“Replaying to mark is that we intend the contrary
“We do not want to shape trustworthiness by singular individuals but as a democratic process that is resilient to adversarial conditions eg collusion
“I’m a way though [sic] the responses we will get from the community will be a bit like MPs voting in brexit options … there will never be a consensus on approach. Hence the need to demonstrate that our definition on trustworthy is not a definition per se but an approach where an individual can take objective metrics to make their own decision. For example where sentiment analysis provides an objective enrichment of language used in posts.”
So basically, they’re talking about using the system as a collective platform to determine what is and isn’t “trustworthy”.
In the actual chat logs we see that Rochko is also managing the chat room for the people invited into the room, which includes Ryan, pictured in the image above, along with Christoph Bauer, Nelscon Escravana, and someone from the SYNYO GmbH group, which is one of the groups being funded by the EU to work on Eunomia… €191,231, to be exact.
A large portion of the chat leak consists of them just trying to organize which rooms they’ll be using to organize information and catalog work. However, near the bottom they begin discussing what sort of criteria they will be using to determine what is and isn’t “fake news”.
They also use Social Science One as a reference for a repository on gathering a criteria for fake news.
The project has been receiving regular criticisms for a number of things, especially for some of the organizations involved in the project, such as INOV, a Portuguese company that received €353,125 to develop the machine learning systems for the project. One user was critical of INOV for developing systems to help prevent illegal immigration, which they had an issue with.
Other users who were aware of the project began asking about more important topics, though, such as data “scraping”, with a post on the Scholar.Social verse from the administration account asking about scraping user data from their instance.
Ryan responds to Rochko in the private Eunomia chat by saying that they would not be “scraping” user data from the instances, but then later states that it would be “difficult to say” if they would be scraping user data for “sentiment analysis”, or determining if posts contain content or structure that might be negative, mean, empathic, etc., etc.
Ryan criticizes people for not being on board with the project and not understanding the importance of the project. One of the links they criticize from Pleroma points out that the Eunomia project is basically like a Western shell that will open up to reveal the same sort of draconian social credit system similar to the one utilized by China.
For those of you who don’t know, if your social credit score in China gets too low based on bad behavior, crimes against the State, or for any other number of acts deemed socially disruptive by the Ministry of Justice, you could be penalized from participating in certain publicly available services, or worse yet you could be whisked away to a re-education camp, as detailed by Vox.com.
Ryan recoiled at the comparisons and paranoia, claiming that the people who are criticizing the system – which is still far from complete – are mistaking it for a social credit system, when in reality it’s a crowdsourcing system.
The main issue obviously is how do you filter fake news when there are people who actively peddle it while being regarded as a source of authority? A perfect example is CNN or the New York Times.
How often have they been caught pushing fake news?
The Covington story is a perfect example, which was misreported by a number of mainstream outlets and now they’re being sued over it. This is despite the fact that sites like YouTube tried to help bury certain facts by terminating YouTube accounts that helped expose the false information that mainstream news sites used to harass and target the high school students.
In fact, this very point was put under contention by YouTuber Japan Anon, who pointed out that plenty of the stories pushed by mainstream news have turned out to be fake, yet they were regarded by platform holders (i.e., such as Google) as being legitimate.
What happens when the people are led to believe fake news and utilize the system to keep real news suppressed on the Fediverse?
Ryan claims that the system will be designed to prevent “collusion and bias”, but what happens when the collusion and bias is orchestrated off-platform and then pushed through the Fediverse? A perfect example of that is the falsehood that #GamerGate was a harassment campaign.
According to Ryan’s logic, #GamerGate as a harassment campaign would be a truism because it’s a widely agreed upon stance by all major news outlets. But if you ask anyone for facts or evidence of #GamerGate’s organized harassment, no one has any facts or evidence to back up the claims. But that’s because the “truth” that people believe about #GamerGate originated in closed spaces like the GameJournoPros, where journalists colluded on the narrative and then pushed it out onto the public space, which then got picked up by mainstream media. That’s precisely how fake news is considered “true” in the public space, and it took taxpayer money and a year long investigation from the FBI to finally reveal in a 2016 report that the headlines claiming #GamerGate was a harassment campaign were false, and there was no actionable evidence based on what the media reports claimed.
Nevertheless, Rochko has been continually running damage control over on the Mastodon instance, stating that the trustworthiness of content will be based on the people and not a central authority, and so people should withhold judgments.
But all of this completely jumps the shark because it ignores the most basic questions at the center of the matter: Why is a social scoring system even needed for social media if users didn’t ask for it?
(Thanks for the news tip Japan Anon)