Thread by @NGrossman81

This @Quillette article claims to scientifically prove Twitter treats conservatives more harshly than liberals. But as a work of science it's fundamentally flawed.
Here's why, and what it says about debates about Quillette, IDW, science, academia.
(THREAD)
quillette.com/2019/02/12/it-…

Arguing that Twitter is out to get conservatives isn't novel. It's basically conventional wisdom on big parts of the right.
But @RichardHanania says he's doing something different: offering empirical, rather than anecdotal, support for the claim of anti-conservative bias.
2/x

To test it, Hanania created "a database of prominent, politically active users who are known to have been temporarily or permanently suspended" from Twitter.
He finds 22 cases, 21 of which supported Trump, and concludes that science shows Twitter is biased against the right.
3/x

But the study commits a fundamental error: selecting on the dependent variable.
To test if pro-/anti-Trump significantly explains Twitter suspensions, you can't just look at Twitter suspensions.
To explain the cause, you need cases where users didn't get suspended too.
4/x

To test empirically, take all of Political Twitter, code as pro- or anti-Trump, see if pro- got suspended at a higher rate and if it's statistically significant.
By selecting on the dependent variable, the article offers anecdotal evidence, despite aiming for the opposite.
5/x

Not the first @Quillette article with flawed science to come across my feed.
I previously wrote about one that made claims about immigration, PC, and race based on a study with fundamental errors (different ones, not selecting on dependent variable).
6/x
arcdigital.media/a-quillette-ar…

Why am I picking on Hanania and Quillette?
I'm not. I'm holding them to a scientific standard. As many academics could tell you, my criticisms are straightforward methodological points, and offered more politely than they sometimes are in academic settings.
7/x

This gets at larger debates about Quillette, IDW, and others who accuse opponents of arguing from emotion, bias, & ideology rather than science. And who sometimes go beyond critiquing PC/SJW ideology to denouncing academia as an institution, culture, & influence on society.
8/x

One of the most valuable things academia brings is scientific rigor. Research methods matter, especially the basics.
Peer review isn't just journal refereeing. It's workshops and conferences, where an error like selecting on the dependent variable would get caught early.
9/x

There are problems in academia, no doubt, and they deserve criticism. But a lot of academic work is rigorous science produced by researches who open themselves to considerable criticism and jump through a lot of hoops before publishing anything.
10/x

Quillette et. al. get a lot of criticism. Some thoughtful, like from @JeffreyASachs. Some ridiculous.
But one reason for the heightened scrutiny is the decision to present as scientific, rather than as opinion, and to critique others on scientific grounds, inviting the same
(END)