A new study that specifically looked at political speech suggests Twitter's content moderation is biased against misinformation, not conservatives

from the if-you-don't-want-to-get-banned-stop- sharing-bullshit dept — Behold! An actually interesting academic study exploring whether … Source: PsyArXiv Preprints .

Techdirt 2022-04-19 Mike Masnick

Context & Ripple Effects

The preprint adds a focused empirical result to a dispute already challenged by an NYU assessment of conservative censorship claims: it examines political speech rather than treating all enforcement actions as equivalent.

It also separates moderation from distribution. Twitter's own research had found greater algorithmic amplification of right-leaning content, meaning claims about political treatment cannot be resolved by account removals alone.

First-order effects

Conservative-censorship claims lose support from a study designed around political speech, while Twitter can point to misinformation enforcement rather than ideology as the pattern identified by the preprint.
Researchers and public-policy participants assessing Twitter's rules gain a distinction between moderation outcomes and political viewpoint discrimination.

Second-order effects

Twitter's critics and defenders face pressure to evaluate ranking, amplification, and removals separately, since an enforcement finding does not explain the platform's distribution effects.
Political-ad restrictions remain a poor substitute for broader misinformation policy: related research found such bans had limited effect while burdening smaller advocacy groups.

Third-order effects

Platform-governance debates are moving toward layer-by-layer accountability, with moderation rules and recommendation systems requiring separate evidence and scrutiny.
If that distinction becomes standard, broad claims of partisan censorship will increasingly be tested against content-level enforcement data rather than inferred from visibility or virality.

The trend: Social-platform governance is shifting from blanket political-bias narratives toward evidence that distinguishes misinformation enforcement from algorithmic distribution.

Discussion

@dg_rand David G. Rand on x
🚨New WP🚨 Many people - from Trump to @elonmusk - have accused Twitter of anti-conservative bias Is this accusation accurate? We test for evidence of such a bias empirically - and turns out it's more complicated than you might think... https://psyarxiv.com/ay9q5 1/ https://twitter…
@tomcoates Tom Coates on x
Interesting thread that says that Republicans are more likely to be suspended than Democrats ... because they are *far* more likely to spread dangerous and inaccurate misinformation. https://twitter.com/...
@abbyayers1 Abby Ayers on x
TLDR: conservatives are more like to spread misinformation so their tweets and accounts are flagged for spreading misinformation more frequently. There's no empirical evidence of “anti-conservative bias”. A truly shocking revelation, I know. https://twitter.com/...
@chrismegerian Chris Megerian on x
Instead of political bias in content moderation, “the observed asymmetry could be explained entirely by the tendency of Republicans to share more misinformation.” https://twitter.com/...
@bardicvoice @bardicvoice on x
And the answer is — self-identified “conservatives” are much more likely than others to tweet misinformation banned by terms of service & thus get suspended for such action. (Reality has a liberal bias ...) https://twitter.com/...
@justinhendrix Justin Hendrix on x
Yet more evidence that social media doesn't have a ‘bias’ problem- it's more that there is a dramatic asymmetry in behaviors that produces more actions against right-leaning accounts. https://twitter.com/...
@ajbauer A.J. Bauer on x
Of course it isn't. Liberal media bias is fake too. But the *idea* of both is politically productive because it dovetails with a sense of victimhood that us been baked into conservative identity since the 1930s and '40s. In this book I will... https://twitter.com/...

Chronicles