Botometer 101: Social Bot Practicum for Computational Social Scientists

Please don't take botometer seriously. It's a neat toy, and performs OK for some settings, but has severe problems that make it unfit for research and/or statements about reality.

See this excellent paper [1] by Adrian Rauchfleisch and Jonas Kaiser for an in-depth analysis, and this [2] blog post for a summary.

My personal opinion is also that bot detection is epistemologically doomed, and while inevitable and necessary for platforms, it is not something that can and should be used as an argument in public debates.

[1] https://ssrn.com/abstract=3565233

[2] https://cyber.harvard.edu/news/2020-03/false-positive-proble...