Caught in the Act: TikTok’s Flawed Algorithm Promotes Outdated Stories

The popular social media platform TikTok has reportedly been delivering outdated and factually inaccurate content to its users presented in news-like alerts, according to the Financial Times.

The alerts included a warning labeled “BREAKING” about a weeks-old tsunami in Japan posted nearly a month after the actual tsunami occurred. Other alerts pushed inaccuracies, such as LA Dodger star Shohei Ohtani being banned from the MLB for five years for gambling or Taylor Swift canceling her tour in “Racist Florida.” In the case of the Ohtani alert, the platform appears to have pulled the content from an April Fools’ Day prank and then promulgated it again later on in the month, confusing viewers. Similarly, the Taylor Swift alert is nearly identical to a satirical post published by The Dunning-Kruger Times in May. She plays Miami in October.

The alerts would pop up on screen and present a personalized video recommendation, a format that has been shown to increase user engagement, but which some researchers warn could make viewers more susceptible to misinformation campaigns. The alerts appear to most often be derived from posts that the algorithm calculates will receive high levels of engagement.

“Notifications have this additional stamp of authority,” Laura Edelson, a researcher at Northeastern University, in Boston, told Ars Technica. “When you get a notification about something, it’s often assumed to be something that has been curated by the platform and not just a random thing from your feed.” TikTok removed the specific alerts discovered by Financial Times but declined to share how its notification-based alert system actually works.

This discovery comes amid a contentious U.S. election year that has seen rampant use of AI-generated misinformation and increased scrutiny of the social media platforms that are used to spread it. More than 50 nations around the world, in fact, will be holding elections of their own, including India, Mexico and South Africa. But while AI-driven misinformation efforts have exploded on social media in recent years, those platforms have repeatedly hamstrung, if not outright eliminated, efforts to reign it in. For example, in September of last year, Twitter/X CEO Elon Musk decimated the company’s disinformation and election integrity team. This past February, Meta cut funding to fact-checking efforts on WhatsApp.

For its part, TikTok has repeatedly assured governments, including the U.S. and the E.U., that it is working to counter misinformation campaigns on its billion-global-user platform. TikTok in May became the first social media company to automatically label AI-generated images, though only some of them .

Source: Financial Times via Ars Technica