TWITTER’S PLAN TO CUT OFF FREE DATA ACCESS EVOKES ‘FAIR AMOUNT OF PANIC’ AMONG SCIENTISTS
When Twitter announced on 2 February that the social media platform would end free access to its application programming interface (API) in a week, meaning tomorrow, a clock began ticking for Jean-Philippe Cointet. Like other researchers interested in topics such as political polarization or how misinformation spreads, the social scientist at Sciences Po uses the API to freely gather data on the hundreds of millions of tweets sent daily. If Twitter tries to charge a significant price for such information, some of his projects will not be possible. “When we got the news last week, we started to rush on a few data collection projects,” he says.
Other scientists who rely on Twitter’s data are also anxious.
“A lot of people are worried about what this means for their research on political campaigning, disinformation, and other topics,” says Michael Zimmer, a social media researcher from Marquette University. Meanwhile, hundreds of scientists and organizations devoted to researching online information have signed an open letter calling on Twitter to “ensure that APIs for studying public content on the platform remain easily accessible for journalists, academics, and civil society.”
In the past, Twitter has given researchers much more access than other online platforms such as Facebook or YouTube, says Casey Fiesler, an information researcher at the University of Colorado, Boulder. Because of this, the platform has become the equivalent of a model organism in the world of those who study social media. “Twitter has been a hugely important source of data for researchers on basically any topic you can think of—from public health to political science to digital humanities to misinformation,” Fiesler tells ScienceInsider. “And it’s not necessarily because Twitter is the best data source, but rather that the data is so easy to get.” The platform “has considerably shaped a decade of social media research,” Morales adds.
As a result, there has been an overreliance on the platform, according to Deen Freelon, a political communications researcher at the University of North Carolina, Chapel Hill. “I think this [API policy change] is really going to force people to look at other platforms that have been understudied,” he says, pointing to Reddit as an example. “Reddit has more users than Twitter, the data are actually equally easy to access if not easier, but people haven’t been paying as much attention to it.”
Many researchers don’t believe Twitter’s new owner, billionaire Elon Musk, is targeting them. “Based on conversations I’ve had with a number of folks who have access to information inside the company, I think that Musk wasn’t considering academics whatsoever,” says Rebekah Tromble, a computational social scientist at George Washington University. Many companies and other nonacademic accounts use data they can freely access through Twitter’s API for their own products. Musk is seeking new revenue sources for the company and likely wants to make money off those users. “It’s an ill-considered decision to try to monetize the free API,” Tromble says.
What exactly will happen tomorrow is still unclear. Musk is mercurial, having already floated several new Twitter policies and then backtracked or revised them after outcries. The company, in a tweet, naturally, simply announced it would no longer support free access. “A paid basic tier will be available instead,” it said. How much this paid tier will cost is unclear.
Stephan Lewandowsky, a psychologist at Bristol University who mines Twitter data to study what type of misinformation people share, says a lot will depend on how much money Twitter asks for. “If researchers can buy access for a flat fee, e.g. $100/month, then many (but not all!) researchers may be able to absorb that,” Lewandowsky says. “If it’s based on volume [of data mined] and if it ends up being loads more than [$100], then it will disable some important research because many people couldn’t afford charges of multiple thousands of dollars.” (Twitter already has a “premium” tier for API use that includes monthly fees and charges based on numbers of requests for data.)
If researchers have to pay for access to Twitter’s data, that will advantage established researchers and wealthy institutions, Freelon says. “Who’s going to lose out are going to be grad students, people without institutional affiliations, people whose institutions are lower wealth.”
Even if researchers ultimately retain access to Twitter’s data, the uncertainty caused by the API announcement highlights larger issues, they say. Because the effects of social media influence society as a whole, it should not be up to social media companies to control research on this, argues Philipp Lorenz-Spreen of the Max Planck Institute for Human Development, who studies social media networks. “We do not depend on the oil industry to be able to measure CO2, but we are dependent on Facebook to measure polarization on Facebook,” he says. “That is a bad situation.”
In Europe, the Digital Services Act, which came into force in 2022 and whose rules will apply from early 2024 on, seeks to address this issue. Article 40 of the act allows national authorities to compel access to social media companies’ data for researchers studying “systemic risks.” But there is still a lot of uncertainty about how this will be handled, how onerous applications for such research will be and how fast companies will have to respond to the mandate.
For now, scientists such as Cointet are trying to finish projects as they await more information from Twitter. In an ongoing project, Cointet and colleagues aim to understand how political debates are structured on Twitter, in part by analyzing how likely people following one politician are to also follow other politicians. in France, they found a surprise. Instead of a simple left-right spectrum, the pattern that most accurately predicted who people followed was an axis from “global to local,” Cointet says: “Basically, people who are against Europe versus for Europe, people who are denouncing elites versus people who are not, people who distrust institutions versus people who don’t.”
They have been working to reproduce this work for 10 other countries, but may not be able to gather all the data before any cutoff, Cointet says. “It’s too bad we need to do that in such a rush.”
In the end, the debate highlights, once more, how much control Musk can exert over public goods, Lewandowsky says. “At an abstract overarching level, this reveals the fundamental problem of whether democratic societies can allow eccentric billionaires to control public spaces.”
