After the Policy

Abstract

If you teach at a college or university in 2026, your institution has probably given you a policy about AI. What it probably has not given you is help: time, training that fits your discipline, working examples, or a colleague who has figured any of this out. You received governance; you did not receive capability. Drawing on thirty-six published studies, this paper argues that the assessment crisis dramatized by episodes like the spring 2026 Brown University take-home exam is a problem of measurement, not morality — surveillance, detection software, and bans demonstrably fail — and that the productive response is design, grounded in a working knowledge of what these tools actually do well and badly.

Written for faculty rather than administrators, the paper offers a practical map: a set of hands-on experiments for calibrating the tool in your own discipline, a menu of documented assessment redesigns that make delegation unprofitable without burning down your course, the abandoned half of the conversation — what AI can do for your teaching and your own scholarship — and a specific, evidence-backed list of what you and your colleagues are entitled to ask of your institution. It closes with the conviction that teaching is relational work, and that the same technology can serve or diminish it depending entirely on who holds it.

This paper is written for faculty and is distributed freely. Share it with colleagues in unmodified form.