Can artificial intelligence genuinely play a role in moderating some of society’s most divisive debates—those surrounding topics such as abortion, systemic racism, immigration policy, or the decades‑long conflict between Israel and Palestine? Columbia University appears convinced that such a possibility is worth exploring. According to reporting by *The Verge,* the university has very recently initiated experimental trials with a program called **Sway**, a still‑in‑development, AI‑powered debate platform. Designed by two researchers affiliated with Carnegie Mellon University, Sway brings together individuals holding opposing viewpoints and facilitates one‑on‑one discussions that focus on empathy, argument clarity, and civil engagement. The platform’s stated mission, as described on its own website, is not to direct discussions toward a predetermined conclusion but rather to cultivate more constructive, less adversarial dialogue between interlocutors. Nicholas DiBella, a postdoctoral scholar at Carnegie Mellon and one of Sway’s primary architects, disclosed that roughly 3,000 students spanning more than 30 separate universities and colleges have already engaged with the tool—and Columbia may soon be the next major institution to join.
This development arrives amid an extraordinarily turbulent period for Columbia University, which during the past two years has faced heightened scrutiny and ongoing clashes not only among its student body but also between its administration and the federal government. Situated at the heart of numerous controversies, the school has endured episodes ranging from expulsions of pro‑Palestinian demonstrators to a series of police interventions, alongside direct demands from federal authorities concerning university policy. Against this backdrop of escalating hostilities, Columbia’s Teachers College has chosen to pilot Sway in the hope of weaving it into both its curriculum on conflict resolution and its broader campus programs designed to foster bridge‑building. DiBella revealed that beyond Teachers College, other groups within Columbia have also expressed interest in rolling out Sway for regular use by the fall semester of 2026. Simon Cullen, an assistant professor at Carnegie Mellon who partnered in the development of Sway, confirmed to *The Verge* that ongoing conversations are underway with Columbia University Life about potential integrations.
At its core, Sway positions an “AI Guide” into each dialogue session. This guide intervenes when necessary by posing probing questions aimed at sharpening student reasoning and, importantly, rephrasing remarks that the system interprets as disrespectful, thereby attempting to neutralize inflammatory language before it derails the conversation. An illustration provided in the platform’s introductory materials shows a test debate around U.S. foreign policy: whether the United States ought to prioritize the rights of Palestinians by ending weapons transfers to Israel. Columbia University, when invited to comment publicly about the partnership, opted not to issue a statement by the time of publication, even after being given additional time to respond.
Columbia’s interest in this initiative corresponds with sweeping institutional reforms linked to a substantial $200 million settlement with the Trump administration. That agreement, framed as a measure to confront antisemitism on campus, restored the university’s eligibility for as much as $1.3 billion in federal funding but also imposed onerous conditions. These included unprecedented data‑reporting requirements, strict enforcement against disruptive demonstrations, and intensive oversight of international student populations. Additionally, the administration pledged to collaborate with external organizations dedicated to fostering what it term constructive dialogue. It is in this context of federally mandated reform that Columbia’s exploration of the Sway platform appears to reside.
Yet not everyone within the university community is optimistic about such strategies. One Columbia source, who spoke to *The Verge* anonymously out of fear of professional retaliation, argued that these initiatives reflect a recurring institutional pattern. According to this individual, administrative attempts at conflict resolution often dispense with essential historical and political contexts, reducing deeply rooted issues to a set of sanitized “difficult conversations.” They contended that while the academic environment at Columbia excels in rigorously analyzing the nuanced political and historical dynamics underlying divisive questions, the administration too often reframes these challenges as logistical or psychological problems to be managed, rather than serious intellectual and moral confrontations to be reckoned with.
Parallel efforts already in place further reveal this tendency. For instance, Columbia’s Student Leadership Engagement Initiative (SLEI), which convened more than 70 students over a series of meetings last academic year, was promoted as a vehicle for dialogue between students and senior administrators. Participants, hand‑selected by deans and provided with generous financial compensation, were ostensibly tasked with examining divergent perspectives. However, some critics—including the same anonymous source—described the program as less about genuine scholarly exchange and more about trustees employing financial incentives to suppress dissent and quell unrest. Joseph Howley, an associate professor in Columbia’s Classics department, reinforced this skepticism, asserting that university leadership increasingly borrows frameworks from corporate crisis management, policing, and law enforcement. As a result, dissent and protest are treated not as vital democratic acts but as crises to be neutralized.
Further complicating the picture are questions about Sway’s funding and affiliations. Research associated with the project has drawn partial support from the U.S. intelligence community, which annually finances unclassified postdoctoral research in areas deemed potentially relevant to national interest. Additionally, the venture has received contributions from foundations such as the Arthur Vining Davis Foundations, the Snyder Foundation, the Omidyar Network, and Carnegie Mellon itself. DiBella clarified that while anonymized data collected through Sway are shared publicly and may be reviewed by intelligence entities, individual transcripts or identifiable responses remain confidential. Instructors, however, are given results of a brief five‑question quiz that measures student comprehension following a debate, though the transcripts of the conversation remain inaccessible.
Sway’s trial runs have raised profound philosophical concerns. In early empirical tests, students confronted claims surrounding the 2020 U.S. presidential election and whether it had been “stolen.” Critics noted that such experiments highlight a significant dilemma: is it constructive—or even ethical—for discourse platforms to project balance in cases where one position demonstrably rests on falsehoods? And if moderation between factually unequal viewpoints is presented as a virtue, who determines the parameters of fairness? Such questions carry particular urgency when applied to politically charged or historically entrenched issues like Israel‑Palestine.
Still, Sway’s developers point to measurable impacts. The post‑discussion surveys report that many students, sometimes nearly half, acknowledge altering at least one of their prior beliefs after an exchange. Though DiBella emphasizes that changing opinions alone does not constitute success—since participants might adopt misinformed stances—the overarching goal is to reduce hostility and increase openness to engagement. In practice, the system appears to cultivate humility, with students emerging from discussions less entrenched in their own convictions and more receptive to opposing arguments. It is precisely this shift in cognitive flexibility that inspired the program’s name: to “sway” minds so they become more adaptable, though not necessarily uniform.
In addition to Sway, Columbia is reportedly evaluating other technology‑centric methods for shaping campus dialogue. One such tool, Schoolhouse Dialogues, a program developed under the umbrella of Khan Academy by its founder Sal Khan, pairs high school students with opposing viewpoints and has them assess each other’s civility. Columbia allegedly considers incorporating the results of such exercises into its admissions process, potentially tying perceived civility to applicants’ candidacies.
For faculty members like Howley, such trends point to a worrisome embrace of an almost enchanted belief in artificial intelligence as a “magic bullet” solution to deep‑seated political, social, and cultural rifts. He argues that certain university leaders—those far removed from the everyday work of teaching, research, and knowledge production—risk fetishizing technology as a cure‑all. In his words, this reliance on AI represents a fundamental divergence from what he terms the “sacred charge” of a university: namely, the serious and sometimes uncomfortable pursuit of truth through debate, scholarship, and critical thought. In the current political climate, he suggests, the temptation to chase rapid technological fixes may overwhelm the slower but indispensable processes of genuine academic inquiry and human dialogue.
Sourse: https://www.theverge.com/ai-artificial-intelligence/770510/columbia-university-sway-ai-to-cool-off-student-tensions-israel-palestine-protests
