The Deprogram

I know of Michael Sivirianos. From what I hear from relatives, his family is pretty right-wing. I know people who have had him as a colleague. He's a second rate professor of computers who managed to carve out a career out of studying social media, which nets him significant grants from the EU, as his conclusions usually lead to "it's Russian bots!". He has been obsessed with "extreme left wing spaces" since a few years ago. He co-wrote a similar study on the GenZedong subreddit.

Here's a PDF of this study: https://arxiv.org/pdf/2507.23699

Here's an interesting quote highlighting the academic level of these hacks (emphasis mine):

Takeaways: We find left-wing extremists have substantial activity on Lemmygrad.ml. Our analysis shows that Lemmygrad.ml’s top two communities, c/genzedong and c/gen-zhouarchive, are associated with tankie subreddits, r/GenZe-dong and r/GenZhou. Furthermore, our findings suggest that tankies’ platform migration resulted in an increase in user activity and toxicity on Lemmygrad.ml, highlighting the importance of considering the impact of platform migration to decentralized networks on online communities

They use the word "tankie" multiple times throughout the paper by the way.

A short summary of what they did (further demonstrating the ridiculous academic rigor of these characters):

They go on to describe how they used various LLMs to scrape the site of posts, shortly before and after the GenZedong subreddit was quarantined by reddit (probably what was causing a lot of outages back then) and had the AIs grade the "toxicity" levels of each post (they fail to mention what exactly where their criteria for their grading system, which indicates it was probably based on their own perspectives).

Then isolated keywords in posts to identify the top 20 topics of discussion and manually checked 100 posts per topic to ensure the LLMs did their work correctly. Out of 2000 posts they checked, they present 3 as examples that highlight:

• How we support "authoritarian" AES countries (capitalized in the paper, so I assume it was capitalized in the OP):

“DPRK IS THE ONLY KOREA, IMPERIALISTS CONTINUE TO OCCUPY THE SOUTHERN REGION OF DPRK.”

• How we support and justify the "Russian invasion of Ukraine":

“No one’s asking you to be ok with all the horrors of war. But if we don’t understand what started it, we won’t be able to finish it. Ukraine’s ethnic cleansing of Russians, suppression of the LPR and DPR, and flirtation with NATO must stop if ever this war will end.”

• How we not only condemn Israel's genocide but are also antisemites:

“1. Jews arent a nation 2. Isael is a fake country. 3. The only thing which could be more fake than Isael and the “jewish” national identity would be for the fans of start trek in to create a language of their own similar to english, then enter the bourgeoisie and amass wealth and power, and then invade some imperialized nation 200 years later, settle in it and claim their own stark-treakish fake nation state. And they will justifyng this by quoting some science fiction book written in the 1900s which tells how are they the chosen people to colonize this new land. Everyone identifing as a jew is a zionist and an enemy of the communist and anti imperialist movement.”

Interestingly, they actually point out that many posts condemning Israel are careful to also condemn antisemitism. However, they still come up with a horse-shoe theory conclusion that we are just as antisemitic as right-wing extremist spaces

Essentially, they did the bare minimum in analyzing their data, presented highly subjective criteria to categorize their data, counted words without context to present pre-determined conclusions, and only gave 3 random posts as examples to support their conclusions (one of which is a pretty bad example of a typical post and all 3 fail to highlight their points). Not to mention there's barely any discussion on the validity of their methods, minimal discussion on previous work, and their conclusions take up 2 short paragraphs. Their statistical analysis seems to be non-existent, and there's barely any comparisons. They have 4 graphs and a table summarizing their data, and wrote a 6-page long paper (of which half is the reference list). They reference frequently 2 particular communities (c/genzhouarchive and c/genzedong), but they are very unclear whether the posts they collected are just from there, or they are coming from other places as well, which I attribute to hasty writing.

I have a suspicion that the only reason this paper exists is because they had some sort of deadline on publishing something or else they risked losing their grant. I have no idea how this paper passed peer-review based on my criticisms above, which is probably why it's hosted by a variety of low-reputation online "journals".

(I made this comment in the other post, and I thought it'd be a good idea to post it here too)