is The Verge’s senior AI reporter. An AI beat reporter for more than five years, her work has also appeared in CNBC, MIT Technology Review, Wired UK, and other outlets.
Can AI help “smooth over” discussion on abortion, racism, immigration, or Israel-Palestine? Columbia University sure hopes so.
The Verge has learned that the university recently began testing Sway, an AI debate program currently in beta. Developed by two researchers with backgrounds in philosophy and psychology, Sway matches up students with opposing views to chat one-on-one about hot-button issues and “facilitates better discussions between them,” according to the tool’s website. Nicholas DiBella, a postdoctoral scholar at Carnegie Mellon University who helped develop Sway, told The Verge that about 3,000 students from more than 30 colleges and universities have used the tool.
One of those may soon be Columbia.
News of the potential partnership comes after more than two years of escalating tensions at Columbia between students, administrators, and the federal government. The university has spent years at the center of controversy after controversy: expulsions of pro-Palestinian student protesters, a string of police raids, and demands from the federal government.
People at Columbia’s Teachers College are testing Sway in order to potentially integrate it into the conflict resolution curriculum and “bridge-building initiatives at Columbia,” DiBella said. He said there’s also been interest from other teams at Columbia in using Sway for the fall 2026 semester and onward. Simon Cullen, the other developer behind Sway and a visiting research professor of civil discourse and AI at UNC-Chapel Hill, told The Verge that the company is also in touch with Columbia University Life.
Sway places an “AI Guide” in every chat that “asks tough questions to improve student reasoning.” The tool also “suggests a rephrasing” for language it deems disrespectful. One example debate topic laid out in Sway’s intro video: whether or not the US “should prioritize Palestinian rights and stop sending weapons to Israel.”
Columbia did not provide a comment by publication time, even after being granted a request for more time to respond.
‘A pattern that Columbia repeats’
Columbia is making sweeping changes as part of a $200 million settlement with the Trump administration that’s ostensibly meant to combat antisemitism on campus — a settlement that restores Columbia’s access to up to $1.3 billion in federal funding. The school has to provide staggering amounts of data to the Trump administration, enforce “strict rules against disruptive protests,” and “strengthen oversight of international students.” And on top of that, Columbia said it would commit to work with organizations to “to create constructive dialogue” on campus. It’s likely that the potential partnership with Sway AI falls under that category for Columbia.
It’s one instance of many when it comes to the university throwing money at student disagreements in an attempt to solve problems without friction, according to Columbia sources who spoke with The Verge.
“This is a pattern that Columbia repeats, where our conversations are evacuated of politics and history and context,” the Columbia source, who requested anonymity for fear of retaliation, told The Verge. They added: “Columbia, as a place of scholarship and study, excels at framing the nuance and politics of these issues. What the administration is trying to do is frame these as ‘difficult conversations,’ evacuated of their rooted context, both at the university and globally.”
“It’s pretty much the trustees trying to put out fires in my opinion.”
One place for such conversations is a group called Student Leadership Engagement Initiative (SLEI), billed as a way to facilitate dialogue between students and senior administrators, which had seven meetings last fall and seven last spring. It involves more than 70 students that are hand-picked by deans of students (and paid thousands of dollars each, per the Columbia source) to come together and “explore differences in points of view,” according to the website.
“It’s pretty much the trustees trying to put out fires in my opinion,” the Columbia source said, adding, “[Otherwise], you would never see trustees drop that much money per student to come to meetings to meet with senior administrators … It just seems like they were trying to throw money at it.”
“One of the constants of the last two years at universities like mine has been a crisis response-style approach to political controversy, dissent, and protest,” Joseph Howley, associate professor in Columbia’s Classics department, told The Verge. “What we have are approaches from the world of corporate crisis response, policing, and law enforcement being directed at disagreement and dissent as if they are problems to be solved rather than fundamental values to be cherished.”
‘Looking for magic bullets’
Sway’s Cullen, who is also a faculty research fellow at The Segal Center for Academic Pluralism in New York City, has said publicly that the tool is tied to the US intelligence community when it comes to part of its funding and research. Sway also received recent funding from the Arthur Vining Davis Foundations, the Philadelphia-based Snider Foundation, the Omidyar Network, the Tools Competition, and Carnegie Mellon University itself, DiBella said.
DiBella said that Sway will share anonymized data with the public and the intelligence community, but not transcripts or specifics. “All the data that we share is public, so there’s not any specific data-sharing pipeline with the intelligence community,” he said.
The reason for the intelligence community’s involvement is that they fund DiBella’s own postdoctoral work, he said. They “have a bunch of these postdocs that they fund each year to carry out basic scientific research that might be of interest to the intelligence community,” DiBella said. “They fund basic research that could be of interest to them, but it’s entirely unclassified, unconfidential research and there’s no specific data that’s shared with them.”
He also said that though the company does not share student transcripts nor answers with instructors, it does share with them each student’s score on a five-question “understanding quiz” they take after participating in a discussion, which gauges how well they understood the logic of the discussion.
When asked about Sway, the Columbia source said, “I don’t trust Sway would approach this with any understanding of international politics, of power, and it would just be about making people feel better. That really frustrates me, as it’s a common move at Columbia.”
In Sway’s early empirical studies, the team tested users on the topic of whether the 2020 election was “stolen.” But debates like this spur the question: Is it really productive for dialogue, in cases where one side is proven to be wrong, to merge closer together or to “sway” one person closer to a view based on misinformation? In which cases is moderation between two opinions decidedly not good — and who decides that?
“We are in a political moment where everyone is looking for magic bullets.”
Sway’s “understanding quiz” measures success based on a rotation of survey questions given to students in groups of about five questions. They include whether the student found the discussion valuable, whether they now had a better opinion of someone on the opposing side, whether they think the arguments presented by the other side are better than they did before the discussion, and, crucially, whether the discussion caused them to change their mind about the topic of discussion.
“Close to 50 percent actually say they changed their mind about something in the discussion,” DiBella told The Verge. Though he said that alone isn’t a measure of success because “it might be that they changed their mind in the direction of falsity rather than the direction of truth.” Ultimately, he said the Sway team isn’t trying to get students to change their opinions but is looking to get them to be open to arguments from the other side, with less hate involved.
“After having these discussions, students do become less confident in their own views,” he said, adding, “They’re getting closer to each other. They’re becoming more malleable. That’s actually why we used the word ‘sway’ … We want their opinions to be more malleable to allow for the capability of changing your mind.”
The potential Sway partnership is not the only way Columbia is reportedly using tech to screen or shape students’ convictions. The university is also reportedly using Schoolhouse Dialogues, a tool offered by Sal Khan of Khan Academy’s nonprofit, to pair high school students with opposite viewpoints on controversial topics, then rank each other’s “civility” — and Columbia could use that feedback in its admissions decisions.
To Howley, who has also taught a course on AI labor and knowledge work, there’s an influx of groupthink about “the magical promise of AI” that university leadership isn’t immune to.
“Some of the people at the very top, who … do not do the work of knowledge, creation, or education, have convinced themselves that this kind of software is a magic bullet, and we are in a political moment where everyone is looking for magic bullets,” he said. “It all just couldn’t be more disconnected from what I think of as — sorry to sound hyperbolic — the sacred charge of a university.”
Correction, September 5th: This story previously stated that Cullen is an an assistant professor at CMU; he left the university in June. His affiliation has been updated.
