NIX Solutions’ report: AI Overdone With Censorship

Facebook anti-spam systems automatically blocked all user links to publications about coronavirus, quarantine, school closures, etc for several days. The social network figured out the situation, calling it a mistake of automatic systems for checking publications for compliance with the rules of the platform. Similar problems arose with YouTube and Twitter.

With the tightening of measures against coronavirus, many companies sent their employees to work remotely. Large technology companies and social networks, including Facebook, Amazon, Google, Twitter, Slack, Microsoft, were no exception. Facebook content moderators were also sent home among other employees. However, only full-time employees can control content from home, and those who work on a contractual basis cannot for security reasons. As a result, the number of content reviewers has plummeted. The deficit was compensated by increasing the share of automated moderation systems in this work.

But the artificial intelligence that remained “on duty” turned out to be unprepared for independent moderation of news content, notes NIX Solutions.

In recent days, Facebook users who tried to share in their feed links to articles about coronavirus, protection measures, school closures, etc., and received notifications that these publications violate the rules of the platform, are spam, and therefore are blocked. Among the blocked were links to articles in such media as USA Today, NBC, New York Post, BuzzFeed and others. Users launched a discussion of this problem, suspecting the social network in attempts to prevent the appearance of negative publications about the actions of the US Presidential Administration Donald Trump in the fight against coronavirus.

Facebook representatives rushed to clarify the situation. Guy Rosen, vice president of the company, explained on his Twitter account that the posts were deleted by mistake. “The problem was in an automated system that removes links to sites with offensive information. This system mistakenly deleted many other publications,” he said and added that all publications deleted by mistake have already been restored.

Twitter has already promised not to block permanently those accounts that will be judged by AI systems as violating the rules.

And YouTube warned users that the site will respond more slowly to user requests in the near future, and the number of blocked and deleted videos will increase. This number may even be content that does not violate the rules, for the same reason: lack of employees.