[Note: while I was writing this, TikTok pulled this really fun stunt where they started bringing the service back online and publicly thanking the incoming presidential administration, which didn't actually take power for another 18 hours or so, for their efforts. Cool. That's great. Nice one.]
As I'm writing this, it is January 19, 2025 and the Protecting Americans from Foreign Adversary Controlled Applications Act has taken effect in the United States. (This is the law that effectively bans TikTok.) Needless to say, I'm conflicted about this. Would you expect any different from me?
As I progress on my journey to have a healthier relationship with technology and social media, it's thrilling to see some sort of regulation on a service that sucks up the personal data of its users.
It doesn't apply to all of them.
I used TikTok for a while. I actually joined in mid-2021 when I started creating educational videos about my job in broadcast engineering. At some point, I started an account of my own for browsing, and even ended up posting some videos that weren't related to my job.
The very thing that makes scrolling on TikTok a great time sink is the thing that makes it potentially the worst thing humanity has ever invented. The almighty algorithm, that great black box that inspects your every action on the platform in gruesome detail, learns very quickly what you like and feeds you more ad nauseam.
A real and valid concern with most social media is the tendency for the included recommendation algorithm to push right-wing content — and any veil of neutrality has completely disappeared, first with Elon Musk's acquisition of Twitter[1] and its rapid decline into a neo-Nazi cesspool; and most recently with Mark Zuckerberg bending his knee to the incoming presidential administration. Believe it or not, I don't share the same concern about TikTok's algorithm... at least not yet. It can do its damage without having anyone's thumb on the scale.
I'll start with my bona fides: as of last week, the account where I posted broadcasting videos had about 51,000 followers and my most-watched video had approximately 135,000 views (not counting the one outlier that had 8 million views, but we'll talk about that later). I very much benefited from the platform. However, I started seeing the real problem as a user rather than a content creator.
I am by no means a programming or machine learning expert, and I don't have any special insight into how the TikTok algorithm works. This is just my perception...
The TikTok algorithm is eerily good at nailing the stuff you like. If you've ever seen (or posted!) a comment on a TikTok video saying something like “I built this timeline brick by brick”, then you know what I'm talking about. On average, you'll swipe away from a video you know you won't like after an extremely predictable short delay. If it takes you even a millisecond longer than average to scroll away from a particular video, that's a cue to the algorithm to start increasing content like that in your feed. The amount of time past that average it takes for you to swipe controls how much weight the algorithm assigns that content. Every interaction is tracked. It doesn't take a genius to recognize how good that is for engagement — it ends up commanding a lot of your attention.
The data based on your watch time is really only half the puzzle. The algorithm also has to know what to feed you, so TikTok runs a bonkers amount of machine learning analysis on every video uploaded to the platform. This is ostensibly to do good things like identify explicit content and create automatic captions, but it also gives deep insight into what sort of content is inside.
The very simple incentive structure of the TikTok algorithm (feed similar content based on watch time) exploits a natural human tendency to avoid cognitive dissonance. If you don't already know, cognitive dissonance is the discomfort you feel when you question your own beliefs and actions. This SciShow video on YouTube is a fantastic short overview of the subject.
What does an algorithm like this with truckloads of data on both the viewers and the content do? It amplifies. Sure, it can amplify the best in its users, but it frequently also amplifies the worst. I can tell you from experience that during the fall of 2024, up until I realized what was happening near the end of November, I did not like the person I was becoming. I started seeing things that kept pushing my worldview even further to the extreme, and it was at that point I realized I couldn't go any further.
Which brings me back to the one video that managed to get millions of views... that was only because it made many rounds within conspiracy theorist communities. I should explain:
In the United States, every broadcast TV or radio station, satellite radio provider, and multichannel video provider system must participate in the Emergency Alert System, which is jointly managed by the FCC and FEMA by way of the Integrated Public Alert and Warning System, or IPAWS. If you're old enough to remember Hurricane Katrina in 2005, you probably know FEMA doesn't have the best reputation among a large subset of the American population.[2][3] The IPAWS Modernization Act of 2015 requires FEMA and the FCC to run a nationwide test of the Emergency Alert System at least once every three years, which happened last on October 3, 2023. I made a video about the FEMA press release announcing the test with the intended audience of nerds and broadcast professionals like me, but thanks to the content of that video, I had people coming out of the woodwork spouting their wacko conspiracy theories about what it all meant.
TikTok helped me, but for the love of all that is holy, good riddance.