TikTok's algorithm exhibited pro-Republican bias during 2024 presidential race
psypost.org
psypost.org
Very similar methodology to an earlier study the government cited in their case against TikTok: https://networkcontagion.us/wp-content/uploads/NCRI-Report_-...
There are a number of issues with these studies, one being that the way the sock puppet bots interact with content is not exactly organic. Typically they search for content in a conditioning phase, followed by random scrolling during which the recommended videos are collected and classified by an LLM. Modern recommendation algorithms famously work by examining how long and how users engage with content, and there's none of that going on here. Still, the methodology itself and the use of LLMs to classify content is clever and probably about the best we can get.
Also, even if there _is_ a bias, it doesn't tell us why. Are the recommendations intentionally spiked, or is this simply the recommendation strategy that maximizes profit? (Or that the recommendation model thinks will maximize profit?) It's very difficult to tell, which is part of what makes these models dangerous and also part of what makes them difficult to regulate.
On a sidenote, TikTok (and presumably other content platforms) _really_ does not like these studies, as demonstrated by them nerfing search functionality after the second study above was released to prevent researchers using these techniques in the future. I haven't read the study in detail yet, but it will be interesting to see how the team at NYU Abu Dhabi adapted their methodology.
>Typically [...] followed by random scrolling during which the recommended videos are collected and classified [...] Modern recommendation algorithms famously work by examining how long and how users engage with content, and there's none of that going on here.
But they claim that videos are watched, not just collected from the recommendation page.
"The accounts watched 10 videos, followed by a one-hour pause, and repeated this process for six days"
When you say "how users engage with content, and there's none of that going on here", by "none of that", do you just mean likes/comments, that sort of thing?
I would usually consider watching as engaging with content, but if you mean additional engagement (as I would call it, anyways), that would make a lot more sense to me.
With the current methodology, the bot either watches the whole video, a fixed duration of it, or a random duration before swiping. The bot doesn't organically watch or swipe based on its interests like a human user would.
Because that is a political discussion that will inevitably derail the point.
If there is identified bias then the platform must address it. Or it should be labelled a national security threat.
If it's orgasmic then this is no different than any other form of organic popularity. Seeing as Trump won the popular vote and the electoral collage, people were interested in republican content. On the same token it's very easy to AstroTurf and claim it was organic. From my personal conversations in meat-space I lean organic.
Is there funny business going on? Absolutely, all the time everywhere in every way. Can we say this was funny business? Not without the code.
TLDR: Popularity algorithms push popular content algorithmically.
Not aware of any major platform that solely uses a popularity algorithm.
How could we regulate this? I can think of two ways:
- Results-based enforcement. i.e., companies are free to use whatever recommendation model they like, but have to recommend content within ideological bounds. i.e., you can't bias toward one partisanship more than X%. There's some precedent for this with the equal-time rule (https://en.wikipedia.org/wiki/Equal-time_rule) and FCC fairness doctrine (https://en.wikipedia.org/wiki/Fairness_doctrine).
- Algorithm-based enforcement. i.e., there are limits on the algorithm itself. Perhaps you have to present your algorithm to a government agency and provide a proof that it obeys certain properties. But the enforcement here is analytical rather than empirical.
Can you provide research for each of these.
Otherwise it's just muddying the waters to act like the bias is inherent to all platforms.
Here's a recent study on YouTube from the same author as this TikTok study finding left-leaning bias in US recommendations: https://academic.oup.com/pnasnexus/article/2/8/pgad264/72424....
Here's a somewhat older study from Twitter itself where they determined that their recommendations were biased toward right-leaning accounts: https://cdn.cms-twdigitalassets.com/content/dam/blog-twitter....
IMO the interesting question is not whether an individual platform is biased and what its biases are, but rather how we might regulate recommendations given that there is always a risk of bias.
For more anecdotal context, my other interests lean into a demographic/stereotype of an upper middle class pilates/health/cleangirl girliepop. Which from my experience they centrist/ambivalent maybe skew left but not inherently pro trans.
I'm not sure this is accurate, speaking as someone whose social circle has a lot of people in that group and who also sometimes ends up on that side of the algo. For various reasons, it's only socially acceptable to be outspokenly on the left in that group (more so for women over the age of 25/30), but there's a sizable minority who either disagree with certain topics or are more aligned with the right, but they aren't going to jeopardize their social circles or (particularly in the case of content creators) their audience.
It's the girl version of the men who put 'apolitical' on dating apps to hide being a conservative because they don't want to limit their dating prospects.
There are just a lot more secret conservatives in that group than one might expect from outward appearances. The center of the Overton Window is as far 'right' as they can go without losing their audience. I can usually pick up on it pretty clearly, but I also worked in political communications for a while and became very sensitive to signaling and find it a fascinating topic. I wouldn't be shocked if the algorithm picks up on the association and assumes some similarity between down-low conservatives and forthright conservatives and starts serving the latter to viewers of the former.
How about Mastodon or Bluesky?
Consider 4chans fascination and fetishization of "femboys" and even this whole hilarious "saga": https://knowyourmeme.com/memes/favela-trans-girl-saga
This is the same "conservative gen Z" group that somehow helps get trump elected with the Q shit...
Many of them just hate non-passing trans people. Sad but true, and a lot of trans people also dislike other non passing trans people.
From my experience passing, how attractive you are, and your interests/hobbies are all attached to how well you a treated in the world right now.
Trans people call it passing privilege for a reason.
That's always been true for anyone in just about any context.
I only started dressing as a woman again after I couldn't pass as a guy anymore and learned how to be as attractive as I could be. Which is a common experience amongst trans women
I'm a cis lesbian and I only get pro-trans content, BUT I do get a lot of 'random bisexuals shitting on lesbians' content.
I've also noticed that when I scroll away from political posts from the left, it almost immediately tries to give me political posts from the right. The algorithm doesn't seem to pick up that my main concern with political content is whether or not it's FUNNY. The only political content I want to see on TikTok is shitposts, but the algo seems to operate under a pretty basic 'If scroll past left/right, then show right/left'.
Big brush, broad strokes.
I'm not sure but anti-trans content/socializing in general is primarily a driver for poor mental health with trans people. Not engaging and blocking a lot of the content is day 1 advice in the community. Also, a lot of anti-trans content is targeting non-passing trans people, which is often full of ways to trigger ones dysphoria and dysmorphia. Like there are keyboard warriors for sure but i don't know anyone that intentionally watches that stuff and most even avoid tiktok entirely because of it
And I'm skeptical it even identifies me as trans if i'm just watching a few trans people that aren't talking about trans issues or experiences really. Where i scroll past much more xyz issue that affects females in particular, when it's trying to diagnosis me (which i think is cause of my health/nutrition interest it trys to convince me xyz is cause i'm not balancing my hormones or its pocs or assumes my period is terrible lol)
Perhaps more succinctly, there was more anti-Democrat material from both sides (and the binary clarification system is reductive in a way that obscures what's going on)
I'd argue they ran one of the least controversial campaigns in comparison to the current global political climate. I think those in disagreement were successfully convinced by the opposing party.
Trump's first few weeks makes it clear he was the big mover and shaker.
- Martin v. Boise. This was overturned, but for 6 years, insanity ruled.
- The general situation with asylum at the border. It's insane that millions of people can just walk across the border, turn themselves in to an immigration officer, tell a sad story (this included domestic violence according to the Biden administration...), get a court date ~10 years in the future, and just live your life in the US until then.
- Seeing clearly dangerous/crazy people with rap sheets a mile long, including violent offenses, released with slaps on the wrist for the 30th time, only to kill someone.
- Less insane but still questionable and a potent issue: various affirmative action laws/rulings, very "Harrison Bergeron"-esque diversity motivated laws/rulings.
If you force people to choose between the rule of law, and... that, then the rule of law might lose. That's what is happening now.
And once that's done, please also show real-world examples of some of POTUS47's claims that "they're eating the pets, they're eating the cats, eating the dogs".
Because we all know, Donald Trump is known for pushing bullshit, and having lackeys that will ignore reality to believe him. This is not a new trend, it's one with 50+ years of examples.
Do you not remember "DEFUND THE POLICE!"? That is actually insane policy that the entire party got taken over and was forced to follow along with... even Kamala got slammed for it in the 2024 election! One of many things.
So yes, Trump and the far-right have some off-the-wall crazy stuff, but the left does too. Remember your bias. What seems crazy (i.e. abortion restrictions) does not seem crazy to an entire other segment of the country that you don't share the same exact values to.
And your values and things you don't consider crazy may be considered insane by others.
There are positions which I think are dumb, or potentially harmful, or counter to what I believe, etc. But there's rarely been a position I consider "insane" (from all parties!).
If you're immediately jumping to calling every position you disagree with "insane", you're just being hyperbolic.
Even when dissenting political content slips into the feed it's usually still about Elon Musk, the algorithm surfaces more mentions of him than the actual president.
The previous incarnation had its own biases and such, but it did try to run a little more like a real company with official policies. The current incarnation is whatever Elon's mood this second happens to be.
[1] https://cdn.cms-twdigitalassets.com/content/dam/blog-twitter...
This is why Trump wants to save it. Back in his first term he tried to destroy it[0]. Flash forward to now, he openly acknowledges that it helped him win[1]. He doesn't want to lose that influence if it gets dismantled, so of course he's now talking about introducing a sovereign wealth fund and use it to buy tik tok[2].
- [0] https://www.nytimes.com/2024/12/28/us/politics/trump-tik-tok...
- [1] https://thehill.com/policy/technology/5042241-president-elec...
- [2] https://www.whitehouse.gov/fact-sheets/2025/02/fact-sheet-pr...
This, or China supports the Dems, and everything else becomes logical. But let’s start the narrative that Trump supports China by banning their apps and products.
Videos coming out of the Harris camps were milquetoast in comparison to the Trump camp. Of course those videos get more attention, from people that like him and hate him, and therefore get pushed more by the algorithm.
From the article:
"These differences were consistent across all three states and could not be explained by differences in engagement metrics like likes, views, shares, comments, or followers."
The amount of deceit per dollar you can generate today is profoundly inexpensive when compared to the entirety of human history. Sadly, our cowardly tech giants are the antithesis of the morals espoused in comic books they grew up on, seeking to abdicate all responsibility as they achieve unprecedented power.
No-one gets more outraged than left leaning viewers watching Trump videos.
Which CGP Grey also described a decade ago: https://youtu.be/rE3j_RHkqJc?t=200 "This Video Will Make You Angry" [2015-03-10]
Part of it might have also been the (remarkably cynical and dishonest) Arabs for Trump movement, directly linking Trump campaign stuff to pro-Palestinian viewers.
This is an issue heavily pushed by the Republican Party
[1] https://www.i24news.tv/en/news/international/americas/artc-t...
Liberals love consuming Trump content, maybe even more than conservatives. Sort of like the Howard Stern effect. Because after they do, they can go post on Twitter and reddit about how stupid Trump is.
Conservatives didn't have that with Harris. Maybe because all of her content was so scripted and her entire personality that we were allowed to see was designed by committee, and frankly there was just much less content. Trump would go on a 4 hour tirade about some random thing like Greenland. Meanwhile, Harris was tightly controlled by her handlers, only giving "safe" interviews, probably thinking that playing it safe and not fumbling somewhere would be the path to the white house.
Conservatives spent days nitpicking every single thing they could about Harris and would just invent things about Walz. Tim Walz's son cried and conservative media and social networks ran that into the ground for weeks. Acting like MAGA didn't say anything about Harris/Walz is pure delusion.
I never said MAGA didn't say anything about Harris/walz. Is that what you took away from my post?
> Conservatives didn't have that with Harris.
Yes that is what I took from this.
> Sequestering your candidate away from any kind of challenge certainly is a strategy.
Interviews with Fox News and a public debate does not read as running away from challenge. She is not the one who ran from the debates.
https://www.wsj.com/politics/elections/trump-harris-tv-podca...