A good article detailing some of the difficulties Twitter has been having with their new automated labeling system. It does seem to go for a superficial analysis that using an automated system is bad rather than discussing how soon a new system like this, automated or manual, should be fully functional. At the time this article was written the new labeling system had been in place for two weeks. Should a labeling system like this be fully functional after two weeks? Is there any research on what an acceptable percentage of false positives and false negatives would be? The article is also pretty selective with the results of the research they do cite. Although articles that weren't labeled in the MIT study were more readily shared than the baseline, the ones that were labeled were much less likely to be shared. Of course, if Twitter didn't want negative coverage of their roll out of this system they should be more readily transparent with the timeline of the roll out and what they are doing.

This is a very well-sourced and balanced look at Twitter's handling of COVID-19 and 5G related information on its platform and how it is now deploying labels. This piece is concise and gives readers plenty of real-life examples of this new policy at work.

Giv8ng Twitter excuse to use automated editing tools that are also used to silence political speech.