If it cared, TikTok could stop itself being used to stir up tribal hatred in Kenya

<span>Photograph: Dai Kurokawa/EPA</span>
Photograph: Dai Kurokawa/EPA

Over the past year, I have submerged myself in propaganda, trying to study the information nerve-endings of Kenyan politics. What I have uncovered is how the production of disinformation became a cottage industry in Kenya, how disinformation can often be used as a tool to consolidate power, and how European far-right groups have tried to manipulate Kenyan platforms for their gain.

Something that struck me, however, as I waded through all this content was that I would always find several TikTok videos being distributed across platforms. So I decided to focus on TikTok to try to get a sense of the world where these videos were originating.

TikTok likes to position itself as a dancing and lip-syncing app, but my investigation unearthed a lot more than that. There were brazen calls for ethnic hatred and violence.

One video contained a clip of a political candidate giving a speech at a rally with a caption saying that he hates “Kikuyus [an ethnic tribe in Kenya] and will be seeking revenge in 2022.” Another took on the form of a detergent infomercial, saying that “UDA [a Kenyan political party] can be used to remove madoadoa [“stains”] such as Kikuyus, Luhyas, Luos, and even Kambas.” (All these are tribes in Kenya.)

I also encountered phoney content: slickly produced videos containing fake assertions about Kenyan candidates, styled as if they’re from authentic Netflix documentaries, local TV stations or even US president Joe Biden’s Twitter account.

Just as troubling as the content were the view counts – they got millions of views. All this, despite the content clearly violating TikTok’s community guidelines.

Platforms such as TikTok can seem universal: hundreds of millions of people on a single app, sharing and liking the same content.

In reality, TikTok and other platforms function very differently. Their algorithms herd users into bubbles that are unrecognisable to others. Kenyans following the August election are experiencing just that – they’re trapped in a bubble unseen by most of the world, one that’s divisive, dangerous and teeming with lies. This bubble is especially troubling given Kenya’s fragile relationship with democracy.

It has been shown how the influence of Cambridge Analytica overshadowed the elections of 2013 and 2017. In 2017 in particular, platforms actually made money from politicians’ attempts to sow discord in Kenya’s politics. Offline activity was matched by a plague of online vitriol, driving polarisation. Now, another election is upon us – and this time there’s a TikTok bubble, too.

As I browse TikTok content, I have an advantage that most Kenyan users don’t. I don’t just consume TikTok videos; I also study them. My most recent research at Mozilla, titled From Dance App to Political Mercenary, focuses on the platform. What I saw wasn’t genuine content – it was propaganda carefully crafted to hijack algorithms and misinform voters. (Since reviewing my research, TikTok has removed much of the content and many of the accounts I flagged.)

While my perspective offers me some immunity from the propaganda, it also makes me deeply cynical. Because I don’t just see hateful content – I see a billion-dollar platform that’s unable, or unwilling, to moderate that content. I also spoke to TikTok moderators, who described a boiler-room operation: one where quantity of moderation is valued over quality, and one where moderators work in languages and contexts they don’t understand.

Perhaps what’s most troubling about Kenya’s TikTok election bubble isn’t its graphic content or its moderation shortcomings. It’s that it doesn’t have to exist, yet still does. Big tech companies’ much-touted moderation systems are nowhere near as effective as they’re claimed to be. They consistently fail to identify and take down the most egregious content.

Kenya’s election is TikTok’s first real test in Africa, and the platform has had ample time to prepare and learn from older peers such as Facebook, YouTube and Twitter. Or maybe the Chinese platform is already learning from them. After all, Facebook and Twitter’s focus on expansion with complete disregard for civic obligation has seen the companies make billions of dollars. In this respect they’re clearly walking in the path of their predecessors.

When I took my findings to TikTok, I got a familiar template response: “We hear you, we’ve taken down the content, we’re investigating …”

It was however very clear to me that TikTok could have much deeper and meaningful engagement with civil society and factchecking organisations across the country. Even the people I’ve spoken to at AFP, TikTok’s factchecking partner, have little clarity about how factchecking for Kenya is carried out. We also have no idea about how much moderation resources are committed to Kenya or Africa at large. TikTok could do so much more.

Related: Kenya election: 'I'm tired of people asking about violence'

Platforms, more than anyone else, have the best opportunity to deal with problematic content in their ecosystems. Kenyan and African employees of TikTok also need to demand more from their overlords, and lobby internally for reform.

If none of these work, the message of a mass walk-out by African employees of a big tech company will send a clear message to the industry. But it’s something that would require extraordinary courage and collective action.

The information dystopia that Kenya has endured during elections could have a much larger effect beyond the country’s borders. Indeed, there’s a broader impact on trust of institutions and the media at large. The narrative that the world is full of fake news and that you can’t trust anything you see has much bigger consequences than the content these platforms are spreading.

It’s now less than two months until Kenya’s election. And while much time has been squandered and a great deal of damage has already been done, TikTok and the electoral process don’t have to be a doomed couple in Kenya. With the right steps and investments, it’s possible to open the app and be informed, not inflamed.

  • Odanga Madung is a Mozilla fellow, journalist and data scientist based in Nairobi, Kenya