Twitter says its research shows its algorithms amplify right-leaning content more than left-leaning content, but doesn't know what causes the amplification
Twitter is publicly sharing research findings today that show that the platform's algorithms amplify tweets from right-wing politicians … Source: Twitter .
ProtocolAnna Kramer
Context & Ripple Effects
Twitter’s finding follows evidence that its ranking systems can produce uneven outcomes: its earlier image-cropping bias investigation found different treatment by race and gender, while its timeline was also found to reduce exposure to external links versus chronological feeds. The new disclosure matters because it identifies a political-distribution disparity without identifying the ranking behavior behind it.
Twitter had already expanded recommendation beyond followed accounts, at times inserting posts from unfollowed accounts and amplifying extremist rhetoric. That makes unexplained partisan amplification a feed-design and governance issue, not merely a question of whom users choose to follow.
First-order effects
Right-leaning politicians receive greater algorithmic distribution on Twitter under the reported results, while Twitter lacks a defined cause to target with a corrective ranking change.
Twitter’s research team and policy leadership must distinguish whether the observed gap stems from engagement signals, candidate selection, or other ranking inputs before they can credibly evaluate a remedy.
Second-order effects
NYU’s subsequent ratio-based engagement explanation gives Twitter a concrete hypothesis to test: treating hostile reply patterns as engagement may reward politicians who attract them.
Political publishers and campaign accounts have stronger incentives to optimize for the engagement patterns Twitter’s feed rewards, while users relying on the algorithmic timeline face a distribution mix not explained by their follow graph alone.
Third-order effects
The episode strengthens the case that major feed-ranking systems need auditable measures of distributional effects, not just public statements of neutrality, because identical engagement objectives can yield political asymmetries.
If similar findings recur across recommendation features, platform governance will increasingly turn on the design of engagement metrics and the transparency of ranking inputs rather than on content moderation alone.
The trend: Social platforms are moving toward scrutiny of how engagement-based recommendation systems shape political reach, even when the operators cannot yet isolate the causal mechanism.
NEW: Twitter released its own research today that shows that the home feed amplifies right-wing political content more than left-leaning. I sat down with META team leader @ruchowdh to learn what Twitter will do with that information https://www.protocol.com/...
Twitter finds that right-wing news, politicians are disproportionately amplified on its news feed. While the bias is clear, the cause is not - Twitter is now working to study why this bias might exist: https://www.protocol.com/...
Wow. Twitter's algorithms amplify tweets from the political right more than from the political left in six (US, UK, Canada, France, Spain, Japan) out of seven countries studied according to Twitter's own research published today. The outlier: 🇩🇪Germany https://blog.twitter.com/..…
Twitter is a sociotechnical system - our algos are responsive to what's happening. What's next is a root cause analysis - is this unintended model bias? Or is this a function of what & how people tweet, and things that are happening in the world? Or both? 6/n
More of this please, from all the companies including my own. We can't limit ourselves to talking publicly only about AI problems that we've already figured out a solution to, or that are politically safe to talk about. These are hard questions and we need to treat them that way.…
One of the main reasons I joined Twitter was because of its approach to transparency and this is another great example of those principles. https://twitter.com/...
wait, so.... there's *not* a rule that says social media companies have to study this kind of thing only in private, keep the methods and findings under wraps, and then snippily refute them and trash-talk the researchers when they inevitably leak? https://twitter.com/...
For real. Go sign up for a Twitter account. Look at the accounts it recommends for you to follow. And then watch what happens when you don't follow anyone. Within a couple weeks, Twitter will just start pushing right-wing “highlights” to you.
Twitter amplifying right-wing politicians is not new, but hopefully with Facebook's reputation taking even faster, they'll actually do something about it https://www.protocol.com/... https://twitter.com/...
It's honestly amazing that people on the right keep pushing the “social media is biased against conservatives!!!” line even though every social media platform's algorithm is tilted *in their favor* https://twitter.com/...
@ruchowdh this is good stuff. I can tell you why, tho: there are multiple libidinal economies in play on the TL and the algo (capital, technoculture, and white supremacy) https://www.engadget.com/...
Twitter's own research shows that it's a megaphone for the right-leaning content — its algorithms amplify right-leaning political content more than left-leaning, but they aren't sure why..... ummmm, maybe BOTS??? https://www.protocol.com/...
Twitter's done well to study what its algorithms are doing and to tell the world rather than burying it. Next step is to figure out why it's happening. https://twitter.com/...
the thing about this is, if you are a company run by people who largely believe that the “political right” is harmful and you know your algorithms amplify content from the “political right,” wouldn't you disable algorithmic amplification? https://www.engadget.com/...
my research shows that someone has been getting into the Halloween candy while my wife is at work, but doesn't know who's doing it https://twitter.com/...