Facebook is teaming up with academics across the country to determine once and for all whether Facebook is in fact influencing the 2020 election. The only catch: They won't know the answer until well after it's over.
The new research project, which Facebook announced Monday, will study how the 2020 election is playing out on the world's largest social network, and how the platform affects things like political polarization, voter participation, trust in democracy and the spread of misinformation. A 17-person research team, which includes leading academics in the fields of media and politics, will work with some two dozen Facebook staffers to design the experiments.
Once users opt in to be part of the study, the research team will deidentify their data, split them into groups and begin tinkering with their News Feeds, switch up their ad experiences and, in some cases, even ask them to stop using Facebook temporarily, all while surveying participants to see how their experiences and viewpoints evolve and stack up against control groups. The findings, which Facebook will have no veto power over, will be published for free to the public beginning next summer.
In some ways, the undertaking demonstrates how far Facebook has come since 2016, when it eagerly courted political clients with the promise of influence, then, following President Trump's victory, just as eagerly denied that it had any influence at all. Mark Zuckerberg himself famously called it a "pretty crazy idea." But the fact that this sort of research is only now getting underway also demonstrates just how little we actually know four years later. So much of what regulators and the public have come to believe about how social media affects elections boils down to anecdotes and assumption, not data. This research could change that.
"We're in these really uncharted waters where you have two or three enormous companies that control the vast majority of the data we need to advance science," said Joshua Tucker, a professor of politics at New York University, who is leading the outside research team along with Talia Stroud, a professor of communications at the University of Texas at Austin. "This has been just an amazing opportunity to work with a platform, and so we have to try that."
Facebook has a spotty record when it comes to social science research, particularly as it pertains to elections. In 2010, the company performed a randomized controlled trial on 61 million users to see whether serving different iterations of its "I Voted" sticker to different groups of users would impact voter turnout. (It did, sparking outrage.) And of course there was the Cambridge Analytica scandal, where a Cambridge University professor used a Facebook app to scrape data from tens of millions of Facebook users, which he then sold to the now defunct and disgraced data analytics firm that powered the Trump campaign.
This time around, Facebook is being significantly more cautious. Facebook employees will be the only ones with access to the raw data, for one thing, and users will have to explicitly opt-in to participate. That said, Stroud, Tucker and the 15 other researchers they've selected will have maximum input on the research questions and the experiment design. They will preregister their study plans so anyone interested can see exactly what the researchers are setting out to find before they even begin and can compare those goals to whatever they ultimately deliver. Facebook won't have any say over what the researchers do or don't publish, and the researchers won't be paid by Facebook, either.
"The academic team really thought about what the key questions might be, and then in collaboration with Facebook, we've been thinking through what the designs would be, with us bringing the expertise as far as methodology and the statistical techniques involved, and the Facebook team really bring a lot of expertise about what the platform is able to do," Stroud said.
The researchers plan to recruit about 200,000 to 400,000 people across the U.S. to participate. Those who opt in might see fewer political ads or fewer news stories about certain topics in their News Feeds. They might be asked to download apps that monitor their online behavior, though Facebook said researchers won't have access to private messages. They might even be asked to stop using Facebook altogether. Some who will act as a control group won't see any changes at all.
Both the researchers and the Facebook executives are confident the experiment group is so small that these tweaks won't have a meaningful outcome on the election. Of course, that might not stop political operatives from claiming that it will; some, like President Trump's digital director, Gary Coby, have already speculated without evidence that Facebook will intentionally register more Democrats than Republicans as part of its voter registration push.
But the potential insights may be well worth the reputational risks. Researchers and regulators have been pushing Facebook and other tech companies to be more transparent for years. In 2018, Facebook helped launch an organization called Social Science One, through which it has been working to open up data to third-party researchers. In February, as part of that work, Facebook released 38 million URLs that were shared on Facebook between 2017 and 2019. That tranche of data was the result of a painstaking process that took years to come to fruition, as Facebook and the researchers sparred over the best way to share data without sacrificing user privacy.
With this design, the researchers are taking a less intensive route, in order to ensure the research could actually get underway before November. "It's not the ideal system, but it is going to be unprecedented in the degree of research we will be able to accomplish," said Nate Persily, a professor at Stanford University School of Law and one of the co-chairs of Social Science One. "If we didn't do something related to the 2020 election, then that would have been a problem. This is the thing we can do."
Persily as well as Tucker and Stroud, who co-chair committees within Social Science One, hope that if this project is successful, it can serve as a model for other social media giants like YouTube and Twitter to open up their black boxes, too.
Of course, whatever the researchers do find about Facebook's impact on the election, it won't come in time for Facebook to act on it before November. But Facebook's head of research and transparency, Chaya Nayak, said such a huge undertaking wouldn't have been possible two years ago, because the company needed time to figure out how to make it work. "There's going to be other elections, and it's really important for us, as a company, to understand the impact of our platform on both the election coming up as well as elections moving forward," she said. "It's never too late."