Russians boasted that just 1% of fake social profiles are caught, Discord leak shows. Unconfirmed, and it's the Russians claiming it, but could mean huge implications for the web

@JosephMenn: https://www.washingtonpost.com/technology/2023/04/16/russia-disinformation-discord-leaked-documents/

Russians boasted that just 1% of fake social profiles are caught, leak shows

An analysis of Russia’s effectiveness at boosting propaganda on social media platforms is part of a trove of documents recently leaked on the chat app Discord.

The Washington Post
@drewharwell @JosephMenn yeah though 1) adversary boasting and 2) they may be counting their inflater accounts - the stuff they use to boost. We see, for example, 30k accounts deployed to "Like" a new page. Most look like Indonesian click farm type stuff, not sophisticated personas. Grains of salt.
@Noupside @drewharwell @JosephMenn That's interesting. Could you say more about why you don't think it's a big deal? I thought what the article was saying was that the Russians have been extremely successful in manipulating search ranking, trending, and recommender algorithms -- effectively getting massive free advertising, flooding the zone with propaganda, and making information on the internet less accessible and useful -- using fake crowds faking popularity. That's a big deal, isn't it?
@glinden @drewharwell @JosephMenn maybe too much for a reply but: there are a couple things here. first is, do they get attention - mass boosting does not necessarily translate into showing up in recommenders, this varies a lot depending on platform. Twitter in its current trash state may no longer be using "low quality account" assessments, who knows, but prev we've seen networks where content got little authentic engagement - all of the Likes, etc, were from other members of the network.
@glinden @drewharwell @JosephMenn then there is the question of persuasion vs entrenchment vs distraction strategies - I would be curious to see assessments of how they are splitting among these now. Reinforcing true believers in Telegram channels vs actually convincing people of their POV. Different capability set.
@glinden @drewharwell @JosephMenn one thing that will change a bit in upcoming months is that generative outputs can theoretically lead to improved persuasion and entrenchment and potentially activation content. (That is true for any adversary w/access to generative text tools) But the need to still work personas into agents of influence/ trusted members of communities is still a quality not quantity game.
@glinden @drewharwell @JosephMenn IRA accounts - even in 2016 - had a handful of real winners, and then a lot that existed but weren't particularly effective at getting pickup from authentic ppl.
@glinden @drewharwell @JosephMenn i'm curious to hear more about your book! :) there are novel new vectors for getting things out into the info realm via search given the deployments of chatbots alongside traditional search now.
@Noupside @drewharwell @JosephMenn Thanks, Renee, we agree on all of this, I think. But just a note on this point, and I hope you agree, I think the effectiveness of the IRA efforts are unknown, as in, they may have been disturbingly effective, but it's nearly impossible to tell. We don't know for sure, for example, whether or not any election outcome was different. It would be great to know that, but we simply have not been able to determine yes or no conclusively. You agree?
@drewharwell @JosephMenn I can believe that. Before I left Facebook and Twitter I was constantly reporting verifiably false accounts almost daily and the response was 99 times out of 100 that the profiles did not “violate the standards.”
@drewharwell @JosephMenn
Just wait til the Putin, Trump, Tucker Carlson threesome hotel footage is released!!!
Lets see how this one is flagged...
@drewharwell @JosephMenn not at all surprising. If you've ever reported bot accounts on major social networks you know they barely ever do anything about it.