Did Twitter Keep Tabs On Journalists’ Political Leanings?

Did Twitter Keep Tabs On Journalists’ Political Leanings?

The question is not at all hypothetical.

Much of the coverage of Twitter CEO Jack Dorsey’s ongoing press tour has focused on splashier news, like him appearing on the podcast of antivaxxer Ben Greenfield to recount his alarming personal health practices.

But in guest spots on the podcasts of atheist philosopher Sam Harris and standup-turned-MMA-commentator Joe Rogan in recent months, Dorsey blurted out claims about the behaviour of journalists on Twitter that suggest something more troubling.

“One of the things that was very evident during the lead-up to the [2016] election was, just looking at our journalist constituency, which was—is one of the most important constituencies on Twitter,” Dorsey said during his February 5th appearance on Making Sense With Sam Harris, “the amount of journalists on the left who were following folks on the right end of the spectrum was very, very small. The amount of journalists on the right end of the spectrum following folks on the left was extremely high.”

That users ostensibly in the same profession would act so differently based on their political beliefs is surely an interesting piece of information. That is, if anyone could agree on where this claim came from.

“Jack was actually referencing a study done by MIT,” a Twitter spokesperson initially told Gizmodo nearly three weeks ago. “MIT and Cortico are probably better placed to answer your questions.” At that time, the spokesperson linked to a study carried out by a research firm called Graphika, which was published in, but is not directly associated with, the MIT Technology Review.

Gizmodo spoke extensively with one of that study’s co-authors, Camille Francois, who confirmed Graphika’s work had not examined the follow relationships between journalists of varying political appetites on Twitter. She could not recall any public academic study that had.

Of course, the Twitter spokesperson may have simply included the incorrect link. After all, Cortico is a real research group associated with MIT’s Media Lab, which has published research on Twitter’s conversational dynamics—research Twitter continued to insist was the source of Dorsey’s remarks. But when we reached out to Cortico, they, too, couldn’t source Dorsey’s claim.

“In our Media Lab research group—the Lab for Social Machines — we did do a study of the 2016 election […] which references journalists’ locations in the networks we mapped,” a Cortico representative told Gizmodo, “but has nothing to do with journalists following each other.” Vice News, which exclusively ran coverage on the study, noted that verified journalists were “mostly disconnected from Trump supporters” and Trump supporters “formed a particularly insular group” — a problem in its own right, but not at all the same claim made by Dorsey.

Given the lack of apparent sourcing for the remark, the Twitter spokesperson then suggested Dorsey may have “misspoke,” only to later write the following (emphasis ours):

What I have been able to find is that it was an internal research effort done by our product engineering team to help them better understand the concept of “filter bubbles.” We haven’t released anything publicly beyond Jack’s brief comments.

I’m looking into answers to your specific questions, but in general, no, we don’t track political affiliations.

But if Dorsey’s claim came from internal research, remarks he made on The Joe Rogan Experience a month after the Sam Harris interview suggest users’ political opinions are indeed something Twitter tracks internally. On episode 1258, co-guest Tim Pool and Dorsey shared the following exchange (again, emphasis ours):

Pool: I think you mention on Sam Harris that the left… these left liberal journalists only follow each other—

Dorsey: Yeah, in the run-up to the 2016 elections

Pool: Yeah, and so, I still believe that to be true and I’ve worked in these offices—

Dorsey: It has changed. They’ve done the study again, the visualisation, and now there is a lot more cross-pollination. But what we saw is folks who are [inaudible] on the left end of the spectrum mainly follow folks on the left, and folks on the right followed everyone.

Granted the “they” in Jack’s statement is vague—but verifiably neither MIT’s Media Lab or Graphika have repeated their respective filter bubble studies since the 2016 election. Prior to the publication of this story, a second spokesperson for Twitter pushed back on the contentions of the first, claiming to be unable to locate any such internal studies.

Both spokespeople repeatedly declined requests to speak with Dorsey or a member of the product engineering team.

So what accounts for Dorsey’s repeated claim that left-leaning journalists were considerably less open-minded than their conservative counterparts in 2016? (A claim that is practically tailored to appeal to conservative-leaning listeners of Rogan and Harris.) Let’s generously rule out the possibility Dorsey is knowingly lying or inventing figures out of whole cloth.

That leaves two options: Either Dorsey misunderstood academic work about his own platform and mischaracterised that data publicly, twice; or he shared only the most enticing piece of data from research his company carried out on users’ political leanings—research the company has refused to make public.

Whether justified or not, the political biases of social media companies are being fiercely debated right now. Public speculation has led to three separate congressional hearings—one of which Dorsey testified at personally.

Given this level of scrutiny, if Twitter has relevant internal studies on bias patterns, it’s strange it would keep them hidden. Just as strange would be Dorsey himself spreading misinformation, knowingly or unknowingly, about a topic so important to his company’s reputation.

Lacking any additional clarifying information from Twitter, we could really use your help. Are you a current or former Twitter employee with knowledge about internal studies like this? What other kinds of research is Twitter conducting that it hasn’t made public? Send us an email, or contact us confidentially on Keybase or Securedrop. We would, of course, welcome answers from Mr. Dorsey himself as well.


The Cheapest NBN 50 Plans

It’s the most popular NBN speed in Australia for a reason. Here are the cheapest plans available.

At Gizmodo, we independently select and write about stuff we love and think you'll like too. We have affiliate and advertising partnerships, which means we may collect a share of sales or other compensation from the links on this page. BTW – prices are accurate and items in stock at the time of posting.