NYU research suggests Twitter's algorithms promote conservative politicians because they are more likely to be “ratioed”, which Twitter may count as engagement
Our research suggests conservative politicians are ‘ratioed’ more often. That may explain why they're in your timeline.
Context & Ripple Effects
Days after Twitter's own researchers admitted its algorithms amplify right-leaning content without knowing why (Twitter's internal research found right-leaning amplification but no cause), an NYU study supplies a candidate mechanism: conservative politicians draw more hostile replies, and if 'ratioed' replies are logged as engagement, the ranking system rewards them. The finding lands on top of Twitter's earlier practice of inserting tweets from unfollowed accounts into feeds, which already showed out-of-network distribution can surface extreme material.
The study also extends a research arc that began with evidence that heavy Facebook users consume increasingly polarized news, with the effect far stronger for conservatives (the ~200K-user Facebook polarization study) — making this less about one platform's bias and more about how engagement signals behave across social media.
First-order effects
- Twitter now faces a specific, testable explanation for the amplification its own team documented, putting pressure on how it defines and counts engagement in its ranking system.
- Conservative politicians gain visibility through hostile reply volume they did not seek, while critics of those politicians effectively boost the content they are denouncing.
Second-order effects
- Rival platforms running similar reply-and-engagement ranking — Facebook foremost among them — face demands to audit whether their metrics produce the same asymmetry.
- Advertisers and regulators gain a concrete artifact to interrogate: if hostile replies are treated as positive engagement signals, platform claims of political neutrality become measurable rather than rhetorical.
Third-order effects
- If the pattern holds, engagement-optimized feeds structurally reward conflict regardless of intent, pushing platforms to redesign what counts as engagement or inviting regulation of recommendation systems.
- The eventual remedy likely runs through transparency: opening ranking code and metric definitions to outside audit, a path the industry only began down years later when Twitter partially open sourced its algorithm in 2023.
The trend: Social platforms are moving from denying algorithmic political skew toward explaining it as a byproduct of engagement metrics that reward conflict — a shift driven by outside researchers auditing systems the companies themselves could not explain.