How TikTok's Algorithm Figures You Out | WSJ

- Okay, has anyone else noticed that your For You page has been a little too accurate lately? It hasn't been things that I'll google or I talk about. It's been thoughts. - TikTok knows everything about us. ♪ Hold up, don't scroll ♪ ♪ Lemme ask you something first ♪ ♪ Can someone please explain ♪ ♪ How this algorithm works ♪ - [Narrator] TikTok users often wonder how the world's fastest-growing social network seems to know them so well. - Is TikTok secretly listening to us while we're watching vi

deos? I don't know. (upbeat music) - [Narrator] The answer to how this app gets to know you so intimately is a highly secretive algorithm, long guarded by TikTok's China-based parent company, ByteDance. - TikTok has been so successful in terms of implementing their algorithms. - TikTok's algorithm could influence the thinking of US youth. (upbeat music) - [Narrator] To understand how it knows users so well, The Wall Street Journal created over 100 automated TikTok accounts or bots that watched h

undreds of thousands of videos on the app. We also spoke to current and former executives at the company. Officially, the company says that shares, likes, follows and what you watch all play a role in what TikTok shows you. We found that TikTok only needs one of these to figure you out. How long you linger over a piece of content. Every second you hesitate or rewatch, the app is tracking you. - I just wanna quiet the noise. - [Narrator] Through this one powerful signal, TikTok learns your most h

idden interests and emotions. And drives you deep into rabbit holes of content that are hard to escape. (noise of voices) ♪ I'm happy, happy guy ♪ ♪ Oh, just a happy, happy, happy ♪ - [Narrator] The TikTok experience starts the same way for everyone. Open the app and you'll immediately see an endless string of videos in your For You feed. Take this new user, a 24-year-old from Henry County, Kentucky. (jumbled music plays) TikTok starts by serving the account a selection of very popular videos ve

tted by app moderators. Is this person religious? - Because I still have a purpose. And you still hold a plan for my life. - [Narrator] Do they wanna participate in viral dances. (upbeat music) ♪ Hey, hey, who ♪ (upbeat music) - [Narrator] Are they feeling down lately? - [Narrator] Just remember, I loved you once. And that love goes for a friend, family, or any relationship. - [Narrator] What TikTok doesn't know is that the 24-year-old from Kentucky isn't a person at all. It's one of the bot acc

ounts programmed by The Wall Street Journal. Let's call it kentucky_96. We set up these accounts to understand how TikTok figures out your unexpressed interests. We assigned each bot a date of birth and an IP address, which told TikTok their location. None were given a gender. We gave each bot or user interests but those interests were never entered into the app. The only way our users expressed their interests was by rewatching or pausing on videos with related hashtags or images. Some were int

o extreme sports. Others were interested in forestry. Or dance. Or astrology. - I'm not the babysitter, I'm not the parent. - [Narrator] Or some other topic. - [TikToker] Keep scrolling if you hate animals. - [Narrator] For all our accounts, we found that TikTok draws users in at first by serving a wide variety of videos. Many with millions of views. Then as the algorithm sees what you respond to, the selection of videos and the view counts can get lower and lower with fewer of them vetted by mo

derators to see if they violate TikTok's terms of service. We reviewed our experiment and its results with a data scientist, an algorithm expert, Guillaume Chaslot, a former Google engineer who worked on YouTube's algorithm. He's now an advocate for algorithm transparency. The says TikTok is different from other social media platforms. - The algorithm on TikTok can get much more powerful and it can be able to learn your vulnerabilities much faster. - [Narrator] In fact, TikTok fully learned many

of our accounts' interests in less than two hours. Some it figured out in less than 40 minutes. - [Guillaume] On YouTube, more than 70% of the views come from the recommendation engine. So it's already huge. But on TikTok, it's even worse. It's probably like 90-95% of the content that is seen that comes from the recommendation engine. (dramatic music) - [Narrator] This is a visualization made from hashtags attached to the videos our bots watched. Think of it as a partial view of the universe of

TikTok content. Here's where we found dance videos, over here are the cooking videos. The spindly arm stretching out of the center represent niche content areas. This arm starts with general videos of cute animals but if we follow it out to the end, we find more specific videos for enthusiast of French bulldogs. As kentucky_96 starts its journey, it starts moving around within the mainstream where TikTok is trying to puzzle out what it wants. We programmed kentucky_96 to be interested in sadnes

s and depression. Let's see how long it takes TikTok to figure that out. - [TikToker] Life doesn't happen to you, life happens for you. So if life is taking people away from your life, and putting new ones in. - [Narrator] Less than three minutes into using TikTok, at its 15th video, kentucky_96 pauses on this. - [TikToker] And that love goes for a friend, family or any relationship. - [Narrator] Kentucky_96 watches the 35-second video twice. Here TikTok gets its first inkling that perhaps the n

ew user is feeling down lately. - [TikToker] Whoever comes, let them come. Whoever stays, let them stay. Whoever goes, let them go. - [Narrator] The information contained in this single video provided the app with important clues. The author of the video, the audio track, the video description, the hashtags. After kentucky_96's first sad video, TikTok serves another one 23 videos later. Or after about four more minutes of watching. - [TikToker] I'll leave you alone from now on if that's what you

want. - [Narrator] This one is a breakup video with the hashtag #sad. - [TikToker] Do you why I leave you alone? 'Cause I care about your feelings more than mine. - [Narrator] TikTok's still trying to suss out this new user with more high view count videos. Does the new user wanna watch videos about friendship? (upbeat music) (playful music) Or to laugh at funny fail videos? ♪ Oh no, oh no, no, no, no, no ♪ - [TikToker] Nobody's gonna know. - [TikToker] They're gonna know. (dramatic music) - [N

arrator] Or do they like videos about home repairs? (dramatic music) (upbeat music) Other information from your phone, including location can impact the videos that are shown in a user's feed. - And as a Kentuckian, I never thought I'd lose my freedom over a virus. - [Narrator] For instance, kentucky_96 saw lots about Kentucky but whether or not it keeps showing you that type of video depends on your response to it. A TikTok spokeswoman said the app does not listen to your microphone or read tex

t messages to serve you personalized videos. (pensive music) At video 57, kentucky_96 keeps watching a video about heartbreak and hurt feelings. (pensive music) And then at video 60, watches one about emotional pain. (dramatic music) Based on the videos we watched so far, TikTok thinks that maybe this user wants to see more about love, breakups and dating. So at about 80 videos and 15 minutes in, the app starts serving more about relationships. But kentucky_96 isn't interested. - [TikToker] Your

voice, your smile, your eyes, your laugh. - [Narrator] The user instead pauses on one about mental health. Then quickly swipes past videos about missing an ex. - [TikToker] I miss you and I like having you around. - [Narrator] Advice about moving on and how to hold a lover's interest. - [TikToker] He spends more time on his phone when he's around you. - [Narrator] Kentucky_96 lingers over this video containing the hashtag #depression. - [TikToker] Something's wrong with me. (sad music) And thes

e videos about suffering from anxiety. - [TikToker] It's like a reward. - [Narrator] After 224 videos into the bot's overall journey or about 36 minutes of total watch time, TikTok's understanding of kentucky_96 takes shape. Videos about depression and mental health struggles outnumber those about relationships and breakups. From here on, kentucky_96's feed is a deluge of depressive content. 93% of videos shown to the account are about sadness or depression. - People been lookin' at me. I'm just

like what you looking at? - [Narrator] A TikTok spokeswoman said that some of the remaining 7% of videos are to help the user discover different content. But for kentucky_96, such videos were few and far between. The majority of videos it was shown outside of its depressive rabbit hole were ads. A TikTok spokeswoman said that the simulated activity generated by The Wall Street Journal's bots is not representative of real user behavior because humans have a diverse set of interests. But even som

e of our accounts with diverse interests rabbit holed. We showed our data and many of the videos seen by kentucky_96 to Chaslot. - What we see on TikTok is a bit the same that what we saw on YouTube. So basically, the algorithm is detecting that this depressing content is useful to create engagement and pushes depressing content. So the algorithm is pushing people towards more and more extreme content, so it can push them toward more and more watch time. - [Narrator] TikTok also says it allows y

ou to see less of something by selecting the Not Interested button but Chaslot says that's not enough. - The algorithm is able to find the piece of content that you're vulnerable to, that will make you click, that will make you watch but it doesn't mean you really like it and that it's the content that you enjoy the most. It's just the content that's the most likely to make you stay on the platform. - [Narrator] Our bots only escaped rabbit holes when we changed their viewing interests. When we

told one bot to stop watching videos about ADHD, the algorithm cut back on that content. Still, many of The Journal's bots were rapidly pushed deep into rabbit holes. TikTok learned our bots' most far-flung interests, like astrology, but even bots with general mainstream interests got pushed to the margin as the recommendations got more personalized and narrow. Bots with a interest in sexual content wound up way out here watching hashtag #kingtok videos about sexual power dynamics. And our bot w

ith a general interest in politics wound up being served videos about election conspiracies and QAnon. - Impeach Biden and Kamala. It goes from the House to the Senate. - [Narrator] Deep in the niche worlds of TikTok, users are more likely to encounter potentiality harmful content that is less vetted by moderators and violates the app's terms of service. - [TikToker] Make them angry and sad. They would be so much happier without you. - [Narrator] A TikTok spokeswoman said that the company catche

s a lot of banned content, which passes through both computer analysis and human moderators. It also reviews videos reported by users. - [TikToker] Don't tell them goodbye, just go. - [TikToker] It's not my time. Pew. (TikToker laughing) - [Narrator] There's a lot of fun, silly and life-affirming content on TikTok. (TikToker screaming) But while TikTok can draw out what makes you laugh. (TikToker humming) It can also make you wallow in your darkest box. - Turning the pain from mental to physical

works. - [Narrator] Without ever needing to eavesdrop on you or collect any personal information about you. - I've never attempted suicide or anything. I've never let it get that far. - Whether it's on TikTok, on Facebook, on YouTube, we're interacting with algorithms in our everyday life more and more. We are training them and they're training us. So we have to study this so we understand it better and we don't let it go in directions that are harmful to society or to certain groups of people.

(dramatic music)