Hate speech and disinformation are circulating widely on TikTok in the lead-up to Kenya’s August general elections, according to a new report from the Mozilla Foundation. Even videos that appear to be in clear violation of TikTok policies seem to have been amplified by the algorithms that underpin the platform’s main For You page (recommendations feed), the report said.
An analysis of around 130 videos, which have collectively been viewed over 4 million times, indicate that TikTok has become a significant avenue for election-related disinformation, including attempts to stoke ethnic tensions, according to the report. In 2007, more than 1,000 Kenyans died in post-election violence, which was caused in part by widespread hate speech targeting various ethnic groups. Some of the content on TikTok examined by the Mozilla Foundation includes references to that violence and attempts to capitalize on lingering fears and prejudice toward ethnic groups, with explicit threats against the same communities targeted in 2007, according to the report’s author, Odanga Madung, a Kenya-based researcher and Mozilla Fellow, who called the videos “ghosts of the past.”
In response to the report, TikTok has removed videos flagged by the Mozilla Foundation that the social media company deemed to be in violation of its community guidelines.
Facebook, WhatsApp, and Twitter were each heavily scrutinized for spreading disinformation in Kenya’s 2017 general elections — which also ended in violence, with police officers killing at least 33 people. Since then, TikTok has risen to prominence in the country, creating a new medium for influencing Kenyan politics. Over 60% of Kenyan users aged 16–64 surveyed in Hootsuite’s Digital 2022 report used TikTok each month.
TikTok currently bans political ads on its platform, but, in Kenya, it has still become a popular venue for discussing politics and outspoken takes on the current pool of presidential candidates. The hashtags #siasa and #siasazakenya (the Swahili phrases for #politics and #kenyanpolitics) currently have nearly 38 million combined views on TikTok. Unlike Facebook’s community guidelines, TikTok’s currently do not explicitly deem calculated references to historical violence as a potential content violation.
Regardless of whether they violate TikTok’s current community guidelines, the videos that appeared to be leveraging the trauma of 2007 are considered by Madung to be dangerous. “It’s like adding salt to an existing open wound,” he said. “It is a way to erode trust in the system in a way that directs someone to vote for one candidate or not to vote at all.”
TikTok did not immediately respond to Rest of World’s request for comment on the contents of the report. In a statement to the Mozilla Foundation, reviewed by Rest of World, TikTok said it has partnered with the accredited news organization Agence France-Presse (AFP), to fact-check posts in Kenya. The company also stated it currently has moderation support in Swahili but did not provide details on the number of moderators it had dedicated to Kenya leading up to the election. No details were provided on moderation of other widely spoken languages in Kenya, such as Kikuyu and Dholuo.
Without more transparency from TikTok on the number of moderators dedicated to widely spoken languages in Kenya, Madung questions whether the company will have the cultural knowledge and resources to properly flag and remove disinformation that presses on the country’s historical nerves.
One former moderator Madung interviewed for the report –– who largely worked with content circulating in the Middle East and North Africa, not East Africa –– said that the moderators were assigned daily moderation targets and asked to review 1,000 videos each day. As a result, they often watched videos at three times the normal speed, some of which were not in languages that they spoke.
Madung said that the view counts on videos flagged in the report, relative to the number of account subscribers, indicate that they’ve been “algorithmically boosted” by TikTok’s recommendation feed. “The For You page is not only able to give [these videos] wings, but it’s able to give them a rocket booster,” he said.
Echoing the criticism aimed at other platforms for amplifying disinformation via their recommendation algorithms –– including Facebook’s News Feed and YouTube’s recommended videos –– Madung points to the For You page as a future battleground in tackling disinformation in African elections. “We’re beginning to see a sort of a similar pattern,” he said. “We are calling it out as early as we can, but it’s likely the [For You page] will be key in terms of the broader disinformation threats that are going to spread through the rest of the continent.”