Internet academics are scratching their heads at the closed box of secrets that is the TikTok algorithm. AI experts are prodding at the corners of it. And the triple-threat TikTok stars are acting, dancing and singing their way around it. In fact the TikTok algorithm is such a mystery it’s now a TikTok meme in itself.
Gaming the ‘For You’ page (FYP) – the default feed we land on when we open up the app – is the holy grail of TikTok. The FYP is the feed which brings you new users and content, as opposed to the ‘Followers’ page, which will keep you safe in your bubble. And if you look at it from an ambitious up-and-coming TikTok’er’s point of view, the FYP is the hose which will feed you to millions. Crack its mystery and TikTok royalty awaits.
Smiling is frowned upon
In a trending set of videos, TikTok user @Franstad frowns and dances, in the next she smiles and dances – all in the name of science. She’s scratching at the TikTok algorithm. Results? Frowning while dancing attracts 2.9m views, while smiling and dancing gets a mere 104k views. Does this mean TikTok prefers sad people to happy people?
Hoodie or blouse
Meanwhile, TikTok’er @becca_cam is asking people to like the first video of hers they see. Upon investigation of her profile, she’s dressed-down in a casual hoodie, whilst in the other she’s in a pretty, pink blouse. The first video that came up for us was the hoodie. Are we building up an algorithmic preference for frowning, hoodie-wearing dancers? Probably not. But that’s not stopping half of TikTok dancing to this current meme. Maybe it’s just another good reason to dance.
This TikTok’er is exploring the algorithm through dance and clothing (2/3) pic.twitter.com/ymTbJ5Ceel
— Interhacktives (@Interhacktives) March 4, 2020
TikTok points to behaviour
TikTok itself recently told Buzzfeed: “Our recommendation of accounts to follow is based on user behavior: users who follow account A also follow account B, so if you follow A you are likely to also want to follow B.”
But this doesn’t explain Marc Faddoul‘s experience on TikTok where he created new accounts and was already spotting recommendation patterns. Faddoul, an AI Researcher at UC Berkeley School of Information in California, researches computational propaganda, algorithmic fairness and geopolitics. He knows what to look for.
Feature or bug?
From Faddoul’s brief foray into TikTok, he found reason to believe that TikTok’s recommendations, to him at least, are based on physical appearance.
Faddoul says: “TikTok’s assumption is that, if I follow a blond teen, it is because I like blond teens, not for her humour or the music she features. The AI-bias techlash seems to have had no impact on newer platforms. Follow a random profile, and TikTok will only recommend people who look almost the same.”
On his Twitter account, Faddoul shared images of the recommendations that popped up after he followed other TikTok users. The evidence seems compelling.
Mystery spikes
Canadian TikTok’er @madlynobrien made an otherwise unremarkable video, six days before this article was published, about how easy things blow up on TikTok. The video has 7m views and yet Madlyn has only 66k followers . No obvious reason can be found for this spike. Maybe the algorithm rewards audacity?
And this TikTok’er’s video about how easy it is to go viral on TikTok…went viral (3/3) pic.twitter.com/LmkWjjSJCM
— Interhacktives (@Interhacktives) March 4, 2020
Collaborations
Some spikes can be explained, however, such as The Telegraph’s TikTok video on coronavirus. As The Telegraph’s senior social media editor Francesco Zaffarano says:
“We posted on the telegraph so this video was published a few weeks ago but last Thursday it spiked, jumping from 300k views to 12 million in just a few days . TikTok created a hashtag page dedicated to coronavirus where our video featured as one of the first ones. And this happened in the same days as TikTok partnered with WHO to launch their account and started pushing their content, which is purely informative.”
Whatever the algorithm is, TikTok is still dancing.