A new #Shadowban-Tester for Twitter can be found on https://t.co/9zxUiMAdvX. It also checks for Quality-Filter-Discrimination #QFD, which hides your posts in searches.
What Twitter says vs. what science says.
Check out the paper called "Setting the Record Straighter on Shadow Banning", which is a reference to Twitter's blog post.
https://t.co/XtBKqZFJH2
Are you or your contacts shadow banned? Check it out at https://t.co/yvEGSVYVRM and at the same time help academic researchers working on collecting information on shadow banning.
Purple nodes in the graph are subject to at least one type of ban.
The authors of the paper have also developed an own tester which is available at https://t.co/yvEGSWgwJk. It supports the detection of one more ban type (inverse ghost ban) and also checks neighboring nodes. We highly recommend you to check it out.
A scientific paper about Twitter's shadow banning practice is going to be presented on one of the top computer science conferences. We are cited in the Acknowledgements section.
It's very good to finally see some scientific research on shadow banning.
https://t.co/xliRrFvxjF
The change of the retweet function is an attempt of US election interference. Twitter has tested this feature before. There needs to be an investigation on Twitter's internal insights on how this feature affects the reach of political content.
https://t.co/S1CN9DEZRz
In some cases, the reply deboosting test will output "Something unexpected went wrong". This is the result of a recent Twitter change (The Twitter Web API is returning less tweets than requested.) and it means that we were unable to perform the test.
We are working on a fix.
If you want to grill Dorsey, scientific accuracy is required. You need to first clarify what your definition of shadowbanning is, so he cannot pick a different one. Otherwise, valuable questions are wasted during congressional hearings.
Dorsey's definition of shadowbanning is different to yours. According to Twitter, shadowbanning means to make content completely undiscoverable to others. Notably, only "limiting distribution or visibility of any Content on the service" is not covered by this definition.
Also, algorithms may limit the visibility of content based on follower relationships without taking into account actual political beliefs. For example, a left-leaning person can easily become part of a conservative network. While effects are similar, it is technically different.
We are currently seeing a lot of tests causing our server to exceed the Twitter rate limits. For this reason, tests may currently fail and you will see a server failure message.
We are trying to work around these limits but it is unclear whether our attempts will be successful.
Screenshots from the #TwitterHacked incident showing a moderation panel have surfaced. One screenshot shows a "Trends Blacklist" button, which is a feature we could so far only suspect to exist. It appears to prevent users from influencing the trends.
https://t.co/SfSwTmEuMk
A new #Shadowban-Tester for Twitter can be found on https://t.co/9zxUiMAdvX. It also checks for Quality-Filter-Discrimination #QFD, which hides your posts in searches.
.@Twitter users currently have three tools to address toxic or unproductive replies: Block the user, mute them, or report the account. Now, Twitter is adding a third feature that would allow users to selectively hide replies to tweets from public view. https://t.co/Rb4sqJbFAu