Guardrails in the Classroom: Possible responses by Higher Education to the current moment

Abstract

This closing discussion seeks to draw out the larger questions about higher education's role in this current moment of AI technology debates, based on the takeaways and ideas for next steps by presenters at this summit. This year's summit submissions centered on practical, classroom-level questions such as: how do we structure assignments so AI supports rather than replaces thinking, how do we teach students to catch AI's confident errors, and how do we build institutional policy on articulated values rather than reactive rules? The presentations on these questions arrive at a moment of a much larger and more unsettled conversation. This year has seen prominent AI researchers — including some from the companies building these systems — publicly disagree about whether increasingly autonomous AI agents pose risks at the scale of catastrophic societal disruption. We cannot resolve that debate here, and none of this summit's sessions were designed to, but this is the backdrop against which conversations about AI in knowledge creation now takes place. This closing discussion asks about the role of higher education when the tools adopted are described, even by some building them, as not fully understood and possibly dangerous — and needs to be taught about AI now in higher education. Presenters: Dialogue Facilitator: Yvonne Houy Panelists: Nakisa Ghanbarian, Diana Moss, Elisabeth Linn, Elena Azadbakht, Sushma Alphonsa, Brock Casselman, Zachary Porter, Alethea Inns, Jean Hyo Jin Jeon, Sierra Adare-Tasiwoopa api, Michael Chamberlain, Jesse Fitts, Delight Ejiaka Objectives: Synthesize cross-session themes into actionable guidance for participants' institutions Examine how classroom practice connects to unresolved, higher-stakes AI safety questions Surface areas of consensus and disagreement among panelists regarding higher education's role Activities: Facilitated dialogue among thirteen summit contributors, drawing on their session findings, followed by open discussion with attendees. Expected Outcomes: Participants will leave with a clearer framework for weighing pedagogical decisions against broader uncertainties in AI development, along with concrete next steps for values-based policy work at their own institutions.