Nairobi News

On Tuesday, April 9, TikTok’s head of public and government relations for Sub-Saharan Africa, Fortune Mgwili-Sibanda appeared before the National Assembly Public Petitions Committee to challenge a petition to have TikTok banned in Kenya.

Ben Ndolo, the petitioner, contended that the app is promoting violence, explicit sexual content, hate speech, vulgar language and offensive behaviour, posing a serious threat to the cultural and religious values in Kenya.

To give clarity and defend its position, Mr Sibanda appeared before the committee alongside Majorel Kenya Ltd CEO, Sven De Cauter.

Majorel Kenya is contracted by ByteDance, TikTok’s mother company, to provide content moderation services in Kenya. De Cauter, who had a difficult time convincing the committee of how his company goes about moderating Kenyan content, informed the parliamentarians that it had 250 local employees dedicated to the task.

But even then, De Cauter had a tough time explaining to the committee why pervasive and harmful content was still finding its way into the app.

MPs poke holes in TikTok’s content moderator

A majority of the concerns were on TikTok’s moderation formula, with the Committee accusing TikTok and its moderator of not doing enough.

Mbeere South MP, Nabert Muriuki argued the moderation formula wasn’t effective enough as it isn’t tailored to factor in the local dialects spoken in the country.

“What you are doing is simply not enough. Your AI and moderating formula is not built to work effectively in Kenya. How is your system going to filter content that is spoken in different dialect such as Ki-Meru?” Muriuki questioned.

De Cauter stated that whereas TikTok was a Majorel Kenya client, his company can only deal with what it has been given through artificial intelligence (AI) by the client, as it does not have direct access to TikTok’s users accounts.

He said Majorel Kenya operates on the TikTok software through a back end to moderate content as guided by its client.

“Clients determine what content is moderated and assign a content quest for our employees to review. All content is controlled by our clients,” Mr De Cauter stated. Mr De Cauter did, however, admit that his company was unable to deal with the raised concerns of moderating content in local dialects.

It’s a position shared by Sibanda who admitted that, whereas the app has increased its efforts in moderating content in Kenya, the system isn’t yet perfect as some content still does slip through the AI and human moderation formula.

“We are the first to admit that the system is not perfect despite our best intentions. There is still some content that is going to slip. If you are dealing with user generated content, you don’t know what the next person is going to say or do. Our effort is to see how best we minimise or mitigate the content that slips through the system,” noted Sibanda.

However, for a country that Reuters Institute Digital News Report 2023 ranked as leading the world in TikTok usage, with 54 percent, Mr Sibanda told the committee that there had been a rise in flagged content.

“In the last year, we have taken down over 296,000 videos from Kenya alone. Around the globe 96.7 percent of harmful content is taken down before it is reported while 77.1 percent of content is taken down before garnering any views,” the director noted.

How TikTok moderation works

Still poking holes in the moderation, John Bwire, MP for Taveta, observed that it was disappointing to hear Majorel Kenya admit that it cannot moderate some content.

“Some content in your area of jurisdiction is harmful, yet there is nothing Majorel Kenya Ltd can do about it. Have you taken the matter up with TikTok,” Mr Bwire raised concern.

However, De Cauter’s defended his company saying it was doing its best, maintaining its well-trained moderators are responsible for reviewing TikTok’s user’s profile and user-generated content on the platform which includes images, comments, direct messages, chats, live streams and videos to tell if the user-generated content violates the clients’ community guidelines.

In an interview with the press, Mr Sibanda noted that TikTok content moderation is primarily by use of automated means.

“Through AI, algorithms, and machine learning. 95 per cent of the content we take down for violating community guidelines is via machine learning. Then, after that, we have human beings coming in for context; the machine doesn’t know this is humour or parody and things like that. Human beings also come in for the local language. We also have fact checkers AFP and code checkers as partners. Externally, there is the issue of laws and policies in countries we operate that we have to adhere to.”

TikTok’s defense

Even as the app promised to keep improving on its moderation processes- the elephant in the room- it observed, a move to ban it on that basis would be catastrophic to the creative economy, especially considering that it has already started investing a lot in capacity building and empowering of content creators.

“What we have begun doing in Kenya is working to support creators. We want to see more of the likes of Dennis Ombachi, and Elsa Majimbo appealing to the global audience, not just being big in Kenya,” Fortune noted.

“In markets like Kenya, where we haven’t launched our full suite of monetisation features, what we do is try to build up creators, then pair them with brands. We take creators that are promising, to build them from let’s say 10,000 followers to a million, teach them what are the best times to post, how to engage with their audiences, and have a call to action. I mean the best practice from what we see,” Sibanda notes.

This article was first published on Buzz, Sunday Nation