Read on Twitter
New: here& #39;s @DFRLab& #39;s full readout on online traffic manipulation in the #Azerbaijan - #Armenia Twitter feud. https://medium.com/dfrlab/patriotic-astroturfing-in-the-azerbaijan-armenia-twitter-war-9d234206cdd7
The">https://medium.com/dfrlab/pa... renewed tensions brought out Azerbaijan& #39;s patriotic web brigades on Twitter.
The">https://medium.com/dfrlab/pa... renewed tensions brought out Azerbaijan& #39;s patriotic web brigades on Twitter.
To follow up on this thread - https://twitter.com/zkharazian/status/1283941850840268801">https://twitter.com/zkharazia... - I used a measure called the Coefficient of Traffic Manipulation (developed by @benimmo/ @polbots) to determine the likelihood that the traffic to four pro-AZ hashtags was manipulated.
https://comprop.oii.ox.ac.uk/wp-content/uploads/sites/93/2019/01/Manipulating-Twitter-Traffic.pdf">https://comprop.oii.ox.ac.uk/wp-conten...
https://comprop.oii.ox.ac.uk/wp-content/uploads/sites/93/2019/01/Manipulating-Twitter-Traffic.pdf">https://comprop.oii.ox.ac.uk/wp-conten...
All four hashtags I studied - #StopArmenianAgression, #StopArmenianOccupation, #KarabakhIsAzerbaijan, and #ireli2020 scored within range of previous manipulated traffic flows we& #39;ve seen (although well short of the DFRLab record):
Breaking down the aggregate score showed that the hashtags all had a relatively high avg/posts per user - 12-15 posts vs 2-4 in previously tested organic control samples. (for reference #TavushStrong scored only 2.3)
The CTM is a relative, rather than absolute indicator of traffic manipulation, so we also looked at other indicators.
Across all of the top pro-Azerbaijan hashtags analyzed, dozens of accounts tweeted to the hashtags hundreds of times day b/w July 14 — July 18 (sometimes 1000s)
Across all of the top pro-Azerbaijan hashtags analyzed, dozens of accounts tweeted to the hashtags hundreds of times day b/w July 14 — July 18 (sometimes 1000s)
Among these high-volume posters, I came across three types of accounts.
1. Once dormant accounts (created months ago, never posted);
2. Retired accounts (created when tensions erupted, were hyperactive, but then calmed down);
3. And steady high-volume posters.
1. Once dormant accounts (created months ago, never posted);
2. Retired accounts (created when tensions erupted, were hyperactive, but then calmed down);
3. And steady high-volume posters.
The DFRLab aso found a small like/retweet ring around the account @guys_az, the account of the General Union to Youth for Support (GUYS), a youth organization created in 2019 to support Aliyev’s youth development policy.
see how the likes are in the same order?
see how the likes are in the same order?
A great proportion of these high-volume accounts were recently created... 117 accounts out of the top 483 accounts were created on just July 15, 2020
A couple of caveats:
1. There was no evidence that these accounts were fully automated “bots” — rather, they appeared highly dedicated human users, many of them identifying themselves as college students or associated w/ pro-regime youth groups.
1. There was no evidence that these accounts were fully automated “bots” — rather, they appeared highly dedicated human users, many of them identifying themselves as college students or associated w/ pro-regime youth groups.
2. This does not mean there were not authentic pro-Azerbaijan users employing pro-Azerbaijan hashtags as an expression of online civic action — there were. But these efforts were significantly boosted by a small group of highly dedicated accounts posting at a high volume.
Read the full report here: https://medium.com/dfrlab/patriotic-astroturfing-in-the-azerbaijan-armenia-twitter-war-9d234206cdd7">https://medium.com/dfrlab/pa...
