Is Facebook election project a model for research on big tech?

Academics who were granted unique access to the “black box of algorithms” that shaped use of Facebook and Instagram during the 2020 US presidential election campaign said the project showed the huge value of making such data available for research.

But those involved in the long-awaited project have also warned that future industry collaborations should not be based on its “independence by permission” model.

Social scientists, working in collaboration with researchers at Meta, have published in Science and Nature the first in a series of pivotal papers examining the role of social media in American democracy.

Campus collection: Higher education’s role in upholding democracy

The results reveal significant ideological segregation in political news exposure and demonstrate that social media algorithms are extremely influential on people’s on-platform experiences. However, although algorithm adjustments significantly changed what people saw and altered their levels of engagement, experimental modifications were found to not shift political attitudes.

Michael Wagner, a professor in the University of Wisconsin-Madison’s School of Journalism and Mass Communication, who served as the project’s independent rapporteur, said Meta might view this as evidence to resist future demands by outside academics to “look under the hood”.

But he told Times Higher Education that debate should instead focus on how particular structures provide the context that makes certain outcomes more or less likely.

“I don’t think the bar is, did Meta cause horrible things to happen? I think the bar is, how does the way social media platforms interact with humans influence a variety of outcomes we care about? And there I think the results are less wholly positive,” he said.

Professor Wagner said researchers on the project had been concerned and sceptical about how Meta might “spin” the results, but they recognised that it would be a unique opportunity to get a comprehensive understanding of how a democracy works in the 21st century.

The work involved accessing data on a scale – involving hundreds of thousands of participants and costing tens of billions of dollars – that they would likely never see again, he added.

“Social media has always been a black box, and Facebook has always been the blackest of boxes,” he said.

Professor Wagner said it was remarkable that the group of 17 outside academics were able to pull such rigorous analysis together in time – particularly as they were forced into an “arranged marriage” with dozens of Meta employees.

“But because they did it so quickly, they made many decisions that I think you wouldn’t want to replicate going forwards. The work is good, but I don’t think it’s a model for future industry-academy collaboration,” he said.

Professor Wagner said the outside academics had rejected the opportunity to become temporary Meta employees so they could remain wholly independent, but that choice meant that they did not have access to the raw data.

Future projects should offer greater protections to industry workers so they can reveal more about their platforms and should have longer lead-up times to avoid researchers feeling that they are “building the plane while we are flying it”, he added.

In a commentary accompanying the research articles, Professor Wagner concludes that “independence by permission is not independent at all. Rather, it is a sign of things to come in the academy: incredible data and research opportunities offered to a select few researchers at the expense of true independence.”

Social media data have typically been difficult for researchers to access, with concerns that technology giants are becoming increasingly opaque.

Joshua Tucker, professor of politics at New York University, who helped to initiate the partnership with Mark Zuckerberg’s firm, said the findings were a “huge illustration of the value of making sure that platforms make data available to external researchers”.

patrick.jack@timeshighereducation.com