I was curious what a TikTok user might see if they made a fresh account. I provided no preferred content categories, and did not interact with any video in any way except the skip them.
In my first fifteen minutes, I received:
9 manosphere “inspiration” videos
The official accounts of Tucker Carlson and Vivek
2 clips from each Fox News, and tabloids New York Post and Metro
1 “America first” rando
1 Copaganda vid of “cool” new police robots
Elon’s nazi salute without context or commentary
I also got some questionable AI slop; an edited perp walk for Luigi Mangione that presented him as especially dangerous, and one robotic voice that mused something nonsensical about the “stereotypical appearance” of Mexican people.
Did I receive any real news? Yeah, a little bit!
Exactly one clip for each of the following publications: ABC, CNBC, BBC, MSNBC, SBS, NBC, CBS, and People Magazine.
I received one interview clip with Bernie Sanders, Kendrick at the Super Bowl, and someone having their vulva tattooed. (lol)
Everything else was dance videos, pet compilations, and recipes.
I would like to repeat this experiment with a more structured methodology, instead of me just screwing around on my couch. I’m not 100% sure how to do that with any amount of scientific rigor so that’ll take some research and prep.
When I concluded my experiment, I reset the feed and logged into my own actual account (which has never been used). I provided it with queer content and legit news follows, and specific preferences for content types.
It proceeded to not give me any of those things, and instead provide either the specific things I asked not to see, or recipes and dance videos in completely random language.
I do not think TikTok appreciated my experiment. lol
Oh, but it DID suggest I follow a family member, so it knows exactly who I am now. (I did not provide access to my contacts, so presumably that family member did)
@Haste my iPhone last night suggested that I name one of the Luigi photos I have in my camera roll. And they’re right it’s easier to find Luigi memes if I name his face so I did.
Then it immediately wanted me to connect that name and that photo to one of my contacts. I don’t have Luigi in my phonebook so I opted not to do that, and I will never ever do that to anyone I care about.
But I’m quite sure people who have iPhones with photos of me and my phone number in their phone have done this. And there’s no way to stop them .
Oh also I had Apple sell my information to Carter’s baby clothes when I stopped logging my period because I’m 51 years old. Carter’s claimed I signed up for their email list right around the time I would have been four months along if I was actually pregnant. I’m still mad about it.
@Haste Would be interesting to experiment from different countries as well
@philfr agreed; on at least some items, TikTok does disclose that its recommendation is passed on location.
@Haste The phrase "thanks, I hate it" comes to mind.
@bstacey it was almost omnipresent in my own mind as well
@Haste
That sounds consistent with this article about political bias in TikTok recommendations:
(Personally, I've never used TikTok - so I wouldn't know.)
@karadoc thanks for the link! I’ll give it a read
@Haste I had much the same experience browsing YouTube anonymously in a private window. It's lots of propaganda from the local Nazi party and manosphere garbage.
@zomgwtfbbqkewl Ah, I’ve never even tried YouTube, good idea
@zomgwtfbbqkewl @Haste yep when I did that I got AI animals, so many AI animals. And the rest was half dressed women being harassed on the street in a way that wasn’t calling out the harassment at all it was encouraging and normalizing it. It was gross.
Google receives about 50 signals from your interaction with their platforms.
I don't know how many TikTok uses, but I'd guess they use your IP and Geolocation on top of things like your device footprint and others.
For your test to be conclusive, you need to throw in a VPN and factory default platform.
I think.
But even then they likely monitor VPN exit nodes ranges.
It's interesting nevertheless.
@n_dimension that’s a good point, thanks for the deets. I think I’d need to partner with a Real Nerd 
to do this in a satisfactory way.
@Haste they are really #dotcons the question is why people don't lift the heads more https://hamishcampbell.com/why/
@Haste I clicked on YouTube the other day when I wasn’t logged in and almost every short they suggested for me was a woman who was almost naked being sexually harassed in a way that acted like it was normal behavior.
Like the videos weren’t calling out sexual harassment, it was to show women’s boobs bouncing around or a man slapping their naked butt. It was so weird I don’t watch content like that at all, I’m a middle-aged white lady, so I assumed that was just what they show people When they look at YouTube without logging in. Or maybe it knows it’s me even though I’m not logged in and since I watch a lot of feminist content they thought I wanted to see some gross red pill stuff. Idk it was literally all AI videos about animals and women being sexually harassed like it was normal.
@maggiejk Im really sorry you had to wade through those harassment videos. :(
@neatlittlepilesofchaos thanks for reading, fam
@Haste the crap on socials is pretty wild, feels similar to how it was pre Brexit, maybe is the same operators although now it is rehashed using AI added incite.
@Haste
I wonder if there are any other pre-install factors that TikTok might use to feed its recommendation algo. Eg, Location, ISP, phone version, et al
Not that you'd be seeing something wildly different.
@kwame That’s a good question. As a rule I keep most a close watch on data sharing and privacy settings on my device; my friends tell me I’m insane for having it so locked down. But there’s definitely device fingerprinting things I could not eliminate.
If I repeat this experiment, I’ll need to do it on new devices somehow. That sounds expensive but not impossible
@Haste This sounds like it could be a good old-fashioned crowdsource activity.
Lmk if you're looking to principal a bigger group study on this topic. This is definitely an interesting topic that could probably be usefully applied.
@kwame I have a couple of interested pals in the infosec / general nerdery space that may help me organize it.
@Haste I mean if your MAC address has ever made it into any one of the data broking databases out there (legitimate- or darknet-wise) the chances of your new account not in some way being associated to an older one is probably next to nil.
@Haste Do TikTok accounts include gender info? Some guy said he made a new account, IIRC said nothing of preferences, and received a lot of scantly dressed women and teenagers. (I was years ago, and I've never tried TikTok myself.)
@lffontenelle They do solicit that information, but I did not give it to them in this instance. If I had, I presumably wouldn’t have gotten manosphere posts.
@Haste Did you use a new computer and location for the test? There's a technique called "fingerprinting" used to track people. They try to find screen resolution, OS type, etc. in order to identify a user without needing to use cookies.
Some of what they can use may surprise you. For example,, they can use the TCP timestamp option to measure clock skew (each computer's clock runs at a slightly different rate) to help identify a computer.
@bzdev I am broadly aware of fingerprinting techniques. I didn’t control for them in this test, and would definitely need to do so in a formal one.
torstein@Haste i did something similar a while back. The only datapoints i provided was age (44) and sex (m). Pretty sure at least half of the initial content was thinly disguised ads for Only Fans accounts. The rest was a mix of Ben Shapiro, copoganda, road rage videos, and sports.
People are concerned about what these platforms do to kids, but middle aged people are getting the full hose, but without the funny dance videos as relief between.
@Haste yes, it’s so brutal.