ProudAntifaDonaldJMusk is running dozens of accounts
TropicalDingdongI have a more effective way of confirming things like this if interested…
Out with it then. If there’s 300, and they’re run by other people, we’ll need more than subjectivity.
I’ll dm you a link when I’m at my machine.
JackbyDevHere you go: codeberg.org/sillyhonu/lemmy-vote-analyzer
I needed a bit to pull it from the various scripts where I built up parts of it and organize it into a repo. But now its done.
I built it in part to do what I think you are looking for here, which is to detect “upvote” or “downvote” networks, and eventually plan to do a sentiment analysis project based on these networks.
However, I would strongly suggest not projecting onto these results that things are “absolutely” bots or brigaders, or whatever. Reality is that actually, many people upvoting or downvoting is just a part of social media. Many times people have accused Monk of sockpuppeting, but I haven’t seen concrete evidence of this beyond some cursory looks into the matter. I’m mostly interested in the more traditional organic networks of coordinated behavior (relative to the content of the post). But you should be able to use this to look for more malicious stuff.
Formatted
spoiler================================================================================ Community 5: 9 members Edges: 24, Density: 0.667 Avg down weight: 0.005994 Down/Up ratio: 0.06 Suspicion score: 39.96 Members: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected] ================================================================================ Community 3: 6 members Edges: 9, Density: 0.600 Avg down weight: 0.006017 Down/Up ratio: 0.92 Suspicion score: 36.10 Members: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected] ================================================================================ Community 1: 22 members Edges: 75, Density: 0.325 Avg down weight: 0.001563 Down/Up ratio: 0.08 Suspicion score: 5.07 Members: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected] ================================================================================ Community 4: 49 members Edges: 234, Density: 0.199 Avg down weight: 0.001628 Down/Up ratio: 0.17 Suspicion score: 3.24 Sample (first 25): [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected] ================================================================================ Community 2: 28 members Edges: 82, Density: 0.217 Avg down weight: 0.001291 Down/Up ratio: 0.10 Suspicion score: 2.80 Sample (first 25): [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]
Much nice, thanks.
Some explanation. Density is effectively a measure of how “compact” a given community within the system is. Its the average rate of “agreement” for any given vote direction (weighted or unweighted). You could think of it as being how “concerted” a group of voters are.
You have some options with how you build the networks. You can use just upvotes, just downvotes, or you can 'weight" the upvotes or downvotes. Weighting is important because some users just vote, way way more than others. In fact, there are users you can find in the wild who have practically never commented or posted, but vote on practically everything they see. Its not a safe assumption to make to suggest that everyone votes in the same manner or that voting as a tool is used the same way by all users.
In this network, we did “weighted” downvotes, so the weights are proportioned by the users total down-votes cast.
So the “Avg down weight” can be considered the “total average shared percent of user downvotes”. So in that first community, about half a percent of the communities total average downvotes are in concert.
You can of course use upvotes, or unweighted down or upvotes, and I’m just pushing a version now to do a k-partite version of the same thing with individual weights, so just… go nuts if you like.
Suspicion score is finally:
densityaverage_down_scaled10000
This means groups with high density and high downvote coordination have higher suspicion socres.