Facebook’s AI is making some headway. Since 2017, the number of hate postings that could be automatically deleted has increased by a full 24 percent.
Thus, in the last quarter of 2020, 97 percent of hate speech was removed before a user could report it. The context in which they were written is particularly important when recognizing such contributions.
The technicians from Facebook have put a lot of work into ensuring that their AI also takes the various contexts into account. Another point is speech recognition.
Improvements for languages like Spanish, Portuguese, and Arabic have increased the number of remote hate postings from 22.1 million to 26.9 million within a quarter.
Thus, Facebook got 7-8 views of hate postings per 10,000 views in total. Also pornographic articles or images glorifying violence were a lot less to be found on the site.
Overall, Facebook removed in the last section of 2020:
- 6.3 million posts due to bullying
- 6.4 million posts of organized hatred
- 26.9 million hate speech
- 2.5 million suicidal contributions
For 2021, Facebook plans to make its various systems more transparent so that they are easier to understand. In addition, contexts in different languages are to be examined in more detail.