TikTok’s Muscle Power: How Children are Drawn Into a World of Extreme Exercise

Entry type: Single project

Country/area: Norway

Publishing organisation: NRK

Organisation size: Big

Publication date: 2022-12-03

Language: English, Norwegian

Authors: Christian Nicolai Bjørke, Henrik Bøe, Caroline Utti


Christian Nicolai Bjørke, Henrik Bøe, Caroline Utti are investigative journalists in Norwegian Broadcasting Corporation (NRK).

Project description:

TikTok is the platform that is growing the most among children and young people in Norway. Half a million children between the ages of 9 and 18 have a user account. At the same time, the Chinese-owned company is reluctant to share how the algorithms affect our children. With the article “TikTok’s muscle power”, NRK was able to show for the first time how and how quickly a Norwegian boy is drawn into the world of extreme training on TikTok. NRK was also able to document that TikTok had failed to prevent the algorithms from serving one-sided content.

Impact reached:

• “This can destroy lives”, said Anette Trettebergstuen, the Norwegian Minister of Culture, after reading NRK’s article. The minister brought it up in a meeting with TikTok on 9 December and asked for greater transparency around the algorithms and better moderation. TikTok replied that for the first time they will establish a contact point in Norway to have a better dialogue with the Norwegian authorities. The minister also announced a joint Nordic set of regulations to regulate TikTok and other large technology companies. For the time being, a committee has been set up to come up with proposals for measures in April.
• Three weeks after NRK put the spotlight on TikTok’s algorithms and how little transparency there is around them, TikTok sent out a press release stating that they are introducing a function where you can find out the reason why a video appears in your feed.
• The article has 330.000 page views. We have received a lot of feedback from other journalists, both domestic and abroad, who wants to know more about our methods and work. We have also heard from parents all over Norway. Several have also used the opportunity to have a chat with their teenagers about what they see on social media.

Techniques/technologies used:

Our TikTok-robot is built up by a program controlled by a computer, and a mobile phone equipped with the TikTok-app and a new user account. The user is set to be 13 years old. The robot is then tasked with using TikTok like a human boy would; by opening the TikTok-app and start watching videos appearing on the For You-page. We want to give TikTok’s algorithms a hint about what our 13-year-old user is interested in. Therefore, we give the robot a list of hashtags and user accounts related to building muscle. When he encounters videos related to the keywords on our list, he will take his time watching them. Sometimes he will also hit «Like».

For the next eight days, the robot swipes TikTok without interference.Swipe by swipe, the amount of muscle-related content rises. Three effective hours in, half of the videos watched evolves around getting bigger, stronger, and slimmer. After less than five hours, the count is 90 per cent; his feed is nearly packed with muscle-related content. During the last four days of our experiment, the count fluctuates between 75 and 90 per cent. To make sure the result isn’t just a one-off, we wipe the phone and create a new TikTok-account. We then retrace our steps twice and get the same results.

Most of our methods in this article used the Python programming language. To control an Android phone, we have used the automation tool Appium. In addition, we have used Microsoft’s Azure Computer Vision API for OCR recognition. To store the data, we have mainly used the database ElasticSearch and the visualization tool Kibana. In addition, we have used SQL for queries in a database, Excel to keep information in order and Jupyter Notebook for documentation.

Context about the project:

The algorithms of the large technology companies have long been an under-covered area among journalists in Norway, presumably because it places completely new demands on the journalists’ knowledge of technology, coding and presentation. At the same time, half a million Norwegian children are at the mercy of an unknown algorithm for several hours every single day – and that’s just on TikTok. In addition, other social media are working hard to imitate the addictive pattern of the TikTok algorithm. “TikTok’s muscle power” is our contribution to putting the topic on the agenda and urging other journalists to join in and hold international technology companies accountable.

What can other journalists learn from this project?

We want to be transparent and share our methods with other journalists. Therefore, we have just written down a method report (in Norwegian) in which we go into detail about how we have worked and which methods we have used. It is about, among other things, how to pretend to be a TikTok user in cities in other countries, how to systematically investigate TikTok’s algorithms via the mobile app (a method that can be reused in new fields with other keywords, subject tags and user accounts) and how one finds and collects TikTok usernames in a given age group. We are happy to share on request (and hope that Google Translate can help make it understandable).

Project links: