World Athletics, the worldwide governing body for track and field, has published a second Online Abuse Study.

The first was conducted during the Tokyo Olympics (in 2020), focused on Twitter, and inspired World Athletics to adopt what it calls a safeguarding policy in November 2021. The body revised the policy this past July, to require member federations to adopt it by 2023.

This time around the study was conducted during the World Athletics Championships Oregon22 and broadened to also include Instagram. It tracked comments and posts on the accounts of 461 of the 487 athletes with active accounts. This is about three times as many as the first study.

From July 10 to Aug. 1, 2022, researchers collected 427,764 posts and examined them for “slurs, offensive images and emojis or other phrases that could indicate abuse.” They used Natural Language Processing – a form of machine learning that IBM describes as “giving computers the ability to understand text and spoken words in much the same way human beings can” – to detect threats, such as “I’ll kill you,” and distinguish them from praise, such as “You killed it.”

Here are some findings:

  • There were 59 “targeted discriminatory posts” from 57 authors aimed at 27 athletes

By comparison, the first study – examining fewer posts (250,000) from fewer athlete accounts on a single social media site – found more than twice as many abusive posts: 132 from 199 authors in Tokyo.

In the latest study:

  • 40 percent of all posts consisted of “sexualised and sexist abuse” and these were “overwhelmingly” aimed at female athletes
  • 60 percent of the “detected abuse” was on Twitter
  • The abuse was usually driven by “events outside of competition” and “not necessarily […] by results in the stadium”
  • 60 percent was sexual or racial (the latter included use of the N-word and of monkey emojis in reference to black athletes)
  • Half of the abuse aimed at female athletes was sexual
  • The abuse aimed at male athletes usually included “general slurs,” and 29 percent of it was racial
  • Two athletes (one male, one female) received almost 40 percent of the abuse

59 percent of the posts were “deemed to warrant intervention from the social platforms,” and World Athletics is considering sending “evidence and reports to national law enforcement agencies” for the most “egregious” 5 percent of them.

The study found in addition that some of the abuse occurred not on athletes’ social media accounts but “in the comments sections of media outlets and other organizations covering the event.”

The “biggest shift” by World Athletics’ analysis was in doping accusations. These were prevalent during the first study, when the ban on Russian athletes was the talk of the Tokyo Olympics, but were few in Oregon.

The study also found little abuse on Instagram, which, as described above, fell outside the scope of the first study. World Athletics cites as a possible explanation Instagram’s “recent progress” in moderation, but it has itself been “regularly educating athletes on social media platforms and available tools.”

Also, one “persistent source of abusive content” was “removed in time for Oregon.”