TikTok launches new features to fight election disinformation

Short-form mobile video platform, TikTok, has launched new features to help Kenyan voters access factual information on the upcoming General Election.

The Monday launch was part of week-long engagements in Kenya aimed at curbing misinformation, disinformation, and hate speech in the lead up to the elections by ensuring correct information is disseminated responsibly.

By working with media experts, the electoral agency, fact-checking organisations such as Africa Check, AFP, Code for Africa, the firm seeks to promote safety and protect Kenyans against harm and misuse on its platform.

As part of the initiative, TikTok launched a three-part series titled #TikTokForPeace in partnership with Article19 including a series of live streams hosted at Strathmore University, Policy Innovation Centre and streamed on the Article19’s TikTok page. 

The series feature key opinion leaders in government, civil society and academia with a focus on tolerance, building bridges and active citizenship.

“We are engaging with diverse stakeholders in Kenya to discuss opportunities and challenges for dynamic solutions in a fast-paced digital world. Our resolve is to spread positivity in markets where we operate as we promote peaceful coexistence,” said Fortune Mgwili-Sibanda, TikTok’s sub-Saharan Africa government relations manager.

According to the guidelines launched in collaboration with the Independent Electoral and Boundaries Commission, Kenyan users will be directed to a portal that hosts basic information on elections including the countdown, candidates and voting requirements.

Although the video sharing platform with over a billion users is mainly an entertainment tool, politics has found its way on it and though it does not allow political advertising, anyone, including politicians and their supporters can be on the platform.

Such online platforms have in recent past been blamed for hosting negative content creators, or "keyboard warriors" who upload negative information meant to sway voters or harm political groupings.

“Apart from taking down offensive content, we collaborate with security agencies through our law enforcement teams that review particular threats posed by content creators. If the security agencies require specific personal details, then proper court orders need to be produced,” says Mgwili-Sibanda.

Misleading content on voting procedures was initially detected in the US during the last election where some people shared wrong election date and warned people not to show up because of Covid-19.

“People were told that due to Covid, they will get a pop-up message telling them when to go and vote. That was a lie. Then there were video manipulations where someone is shown as if he is talking though he is not the one actually talking,” said Mgwili-Sibanda. 

However, the official could not confirm whether they have received specific guideline infringements from Kenya “as we are in the process of addressing such emerging issues such as hate speech.” Some of the issues the company monitors include spikes in country-specific activities within the electioneering period.

“An account active for 24 hours will raise a red flag. Our team also identifies groups or individuals prone to post violent, graphic, or extreme activities, flagging such sensitive posts and phrases meant to harm people, such as ‘madoadoa’ where algorithms pull off such unsafe terms,” said Mgwili-Sibanda.

A recent case study on the word ‘madoadoa’ using 114,000 tweets showed that what began as an offline comment by a politician in a campaign rally had triggered an online debate with Kenyans associating the word with election violence as it brought out other results as "victim," "violence," and "survivors."

Mgwili-Sibanda said moderating content has human and technical approaches adding that while learning algorithms can be taught to detect blood or nudity, anything that slips through can be filtered further by experts.

Allan Cheboi, Code for Africa senior investigation manager says the media can use monitoring tools such as Civic Signal, Crowdtangle, and Meltwater that indicate the frequency and the platforms where such offending phrases have been used in the past and the context.

Kenyans have been encouraged to be a part of the campaign of reporting harmful content that appears untrue and fuels political intolerance.

Over 100 million videos, or 1 per cent of uploaded content was removed within a couple of months with more than 90 per cent of the content deleted before it interacted with other viewers.