x
Social-media researchers overemphasized the platform now called X for years. But now, as it rapidly changes into something new and frightening, we risk paying too little attention.
Last month, my team at the University of Washington’s Center for an Informed Public in Seattle looked at data from X (formerly Twitter) to find the most influential voices in the discourse surrounding the Israel–Hamas war (see go.nature.com/3qwdi). X no longer offers researchers free access to the application programming interface (API), which allowed us to extract and process large amounts of data from the platform. Researchers now have to pay, and the cost is beyond the reach of most. This was one of our first major analyses without the API — and we found it difficult. It took more than a week to answer questions that once took only an afternoon. We had only partial data, collected from a set of content with high engagement that was accessible through the public search interface. In conversations among our team, sentences that began with, “If we still had access to the API ...” became a running theme.
What we found was extraordinary. A small group of seven accounts, many unknown a year ago, were racking up hundreds of millions of views each day, out-performing standard news accounts by an order of magnitude and exercising significant influence on the discourse around the war. X’s owner, Elon Musk, had interacted with or explicitly recommended six of those posters, potentially bringing them to the attention of his 162 million followers. Reporting that built on our work revealed some of the apparent identities behind these accounts: a London teenager who has posted antisemitic content, a US soldier in Georgia who seemed to have pulled at least some news from pro-Russian propaganda channels, and a right-wing news group in Poland.
Researchers scramble as Twitter plans to end free data access
Twitter was always a mix of credible and less credible sources — but our research supports the notion that X is changing dramatically, in ways that are not fully apparent even to researchers who have followed the platform for years. The influence of this new group of accounts, previously unknown to us, had skyrocketed shockingly quickly. In my more than ten years in this field, I’ve never seen an almost entirely new set of accounts come to dominate a major platform in less than a year.
For my team, our findings were a wake-up call. The end of specialized researcher access to X’s API had come with a crumb of comfort for us. We knew that we had focused on Twitter too much in the past, to the exclusion of other parts of the online information environment. It’s still true that misinformation researchers need to move the discipline to wider, cross-platform studies, with a more international focus. Although X serves a small portion of the population, it is still used by many who are deeply engaged in news, as both content producers and consumers. It sets the news agenda for many, particularly during fast-changing events. Although there are open questions about the resiliency of many platforms against misinformation, X is in some ways unique — an influential platform that is historically well-studied but about which less and less is known.
Although many of the warning lights that we once used to identify emerging misinformation are now out of service, the ones that remain are flashing red.
These shifts are occurring in an environment that has been stripped of many of the credibility signals that once helped users — blue ticks that indicated notability; fact-checks by the Twitter curation team, which was laid off last November; and labelling of deceptive content by the platform. X has even stripped headlines and article summaries from links, making it harder for users to identify credible sources and decide where to focus their attention. X has become a ‘natural experiment’ in what happens as a company reduces existing trust, moderation and safety teams. Tracking the way these sudden reversals are affecting discourse on the platform can give us important insights into what could happen if others follow suit.
Thousands of scientists are cutting back on Twitter, seeding angst and uncertainty
Such insights are desperately needed. Our work on last year’s US midterm elections showed that, even before the worrying changes at X, the platform was able to disseminate election conspiracy theories broadly and with remarkable efficiency (see go.nature.com/47gg02). We feel that X will play that part in next year’s US elections — including the presidential race — as well as in dozens of others around the world in what is shaping up to be a very important year.
Refocusing on studying X won’t be easy. It will require new methods and approaches that do not depend on the large amounts of data that were easily accessible before. It will require funders being willing to support important work that might be smaller in scale than earlier studies. Above all, it will require researcher communication and collaboration. My colleague, Kate Starbird, co-founder of the Center for an Informed Public, describes the shift to the new API-less environment as the difference between seeing a landscape through a picture window and looking at it through a series of portholes — now, several views must be pieced together to see the whole. Our recent research shows one piece of the new landscape; to comprehend it in its entirety will require the creativity and focus of many.
