"The sharp rise in Violent/Aggressive Discourse in the Extreme Left is an artifact of the inclusion of Hexbear.net."
"The sharp rise in Violent/Aggressive Discourse in the Extreme Left is an artifact of the inclusion of Hexbear.net."
:yes-hahaha-yes-l:
:sicko-hexbear:
Otherwise gee fucking idiots I guess you realy needed a whole research department to figure things out like
•Alt-Right supergroup activity remains near its all-time high. This activity has been high since the FBI raid on Mar-a-Lago in August (Fig. 1).
•The forums with the greatest use of Violent/Aggressive discourse in November were in Health Misinformation, scoring higher than Incels/Femcels and the Extreme Right (Fig. 7).
•Reference to sex crimes was up 13% in December and was up 32% since August (Fig. 11).
:the-democrat: reading this : "Great, its all according to the plan. Carry on mrMcdoofus, keep us informed, this is very useful information, it will come in handy when we decide to do absolutely nothing for the next 2 years."
You're viewing a single thread.
Methodological Notes: Activity levels are defined as the % of posts with at least one key word, phrase, or stem identified with particular themes. Because the study of language and ideology is a key aspect of the larger project, results are limited exclusively to posts with at least 50 words. The complete list of current platforms from which forums were selected can be found toward the end of this document. A user’s manual and program code can be found at https://www.rogersperspectives.com. Any corrections for errors will appear in next month’s report.
I’ve been looking through the resources on his site but can’t find the list of keywords he uses for this site. Anyone more familiar with computational statistics wanna give it a go?
Yo, i think i found it?
https://www.rogersperspectives.com/tracking-of-extremism-report
Under documentation - https://84c851ae-e204-46aa-8461-295965921c85.filesusr.com/ugd/86bb54_fb085e4a47fe441d8c94a70b0186e680.xlsx?dn=frames_202211_framedictionary.xlsx
Also i like this screenshot
I don't understand, the framedictionary has tons of words that have no association with violence, like "being" and "scottish" and "cold".
the B column gives an idea as to what he's classifying each word as, and sorting by it makes it a bit more clear. i think violence is under the one named "pub_safety" but that has some weird ones like cia, dia, fbi, dhs, blm (lmao, really showing his ass here), pornhub (???) and a bunch of other weird words and phrases.
lmao "ainur"
ainur
Are Tolkien fan forums fasc?
Permanently Deleted
It's probably some incredibly dumb shit like looking for things like "die, dies, dying, kill, kills, killed," which would also explain why antivaxxers are so very elevated if it's reading shit like "fluoridated water KILLS!" and marking that as "violent" when it's just unhinged boomer brain shit.
That would also mean that a comment like "kicking people out of their homes is an act of violence, it is attempted murder and it kills and immiserates people," would read as more violent (because keywords) than someone explicitly calling for minorities to be genocided in short or even slightly obfuscated language. Discussing violence that has happened probably scans exactly the same as direct calls for violence.
Hell, this dispassionate explanation of their reasoning will probably have the algorithm in full :wojak-nooo: : :agony-shivering: :doomjak: mode because of all the keywords.
Chiming in as a silly little grad student applying natural language processing in my research... :
It’s probably some incredibly dumb shit like looking for things like “die, dies, dying, kill, kills, killed,”
Yup. The author references the Linguistic Inquiry Word Count (LIWC) program, which pretty much does exactly that. It's got hundreds of different "linguistic categories" that you can very easily and quickly scan for. Each linguistic category, or "dictionary", is a list of words that are said to reflect a different topic, idea, feeling, etc., and the program will tell you how many words in each text fit into different categories.
If you scroll down to Table 2 on Page 12 of this document, you can see some of their categories and a few example words for each one. You can also make your own "dictionaries", which I think this dude also did. They have a demo on the LIWC website that you can cry, and you will almost immediately see how limited this process actually is.
In research on natural language, like studies that scrape and analyze posts online, it's a hugely popular method for doing the laziest and easiest topic modelling or semantic analysis possible (i.e., looking for themes or concepts that people or talking about, or how they are expressing emotions in text). Like you point out, it's totally devoid of context, so it's validity doesn't extend much beyond "they are using XYZ words this often".
:biden-the-thing: Hmm, the leftists are talking about violence. Could they be discussing everyday violence directly or indirectly effected by the state? :biden-rember: No, it's the leftists who are violent!
I hadn't considered that we're "violent" partly because we talk about state and institutional violence a lot and that makes this even funnier.
It's like the liberal hand wringing of "talking about racism keeps racism alive" but computerized :policing-brain:
that you can cry
keep this typo because the methodological and philosophical faults in this study make me want to cry
LOL
So what you're saying is that now we're being watched we should make an effort to use as many violent words as possible, ie. instead of writing "I would really like a cookie" we should write "I could kill for a cookie"?
I think we owe it to Richard to be good little lab rats and give him some of that violent extremism he's looking for.
Boy howdy it's a scorcher out there, I would VIOLENTLY DISEMBOWEL and DECAPITATE someone for a lemonade today
Yeah, unless they explicitly look for fascist language, a fascist saying "day of the rope now" doesn't show up as violence on whatever algorithm they use
They do appear to have fascist dogwhistles on the list
