Educator urges integrating AI into assignments to teach students, not ban AI use in grading

AdamLGRing used an X thread on May 23 to argue that educators should stop treating generative AI as something to ban outright and instead design assignments around its use. His proposal centered on making students show their work with AI: what they asked, what the system returned, what they kept or rejected, and how they evaluated the result. The post described that approach as a way to teach students to use AI skillfully rather than hide it. The thread was still live on X on May 24. ### What, exactly, did he tell teachers to grade? AdamLGRing said assignments should require students to document prompts, reasoning steps and reflections on AI-generated output, according to the May 23 thread referenced in the source briefing. The point of assessment, in that framing, is not whether a student touched an AI tool, but whether the student can explain how the tool was used and judge the quality of what it produced. (teachingcommons.stanford.edu) That approach matches guidance published by university teaching centers that have been revising assessment design around generative AI. Stanford’s Teaching Commons says instructors can integrate AI into assignments and activities used to assess student learning, while Berkeley’s Center for Teaching and Learning says rubrics can reward visible reasoning and reflection as part of a more transparent assessment process. ### How is that different from a simple “AI allowed” policy? Stanford’s Teaching Commons says assignment design should spell out how AI may be used and what evidence of learning students must still provide. Harvard’s Bok Center similarly says instructors should focus on whether work remains reliable evidence of student learning, rather than trying only to police AI use. (teachingcommons.stanford.edu) AdamLGRing’s thread, as described in the briefing, pushed that logic into grading practice. Instead of a blanket permission or blanket ban, the model he outlined would make process evidence part of the assignment itself: prompt logs, notes on revisions, and written reflection on where AI helped or failed. ### Are schools already moving toward that kind of assignment design? Harvard Magazine reported that faculty members across Harvard were shifting from debating whether to allow AI to creating clearer rules for using it effectively in coursework. (teachingcommons.stanford.edu) The article said instructors were redesigning assignments as generative AI became more common in academic work. The University of Virginia, Indiana University and the University of Illinois Chicago have also published guidance on integrating AI into assignments rather than relying only on prohibition. Those materials describe concrete options such as AI-assisted drafting, critique of AI responses, process documentation and revised rubrics that distinguish between the final product and the student’s decision-making. (harvardmagazine.com) ### Why does prompt documentation keep coming up? Berkeley’s teaching guidance says making reasoning and reflection visible helps instructors assess the learning process as well as the final product. Liverpool’s prompt-template guide for teaching and learning similarly frames prompt design as something that can be customized and taught, not just used privately. (teaching.virginia.edu) That makes prompt records useful for a teacher in two ways. First, they show whether a student can direct a tool toward a specific task. Second, they create a basis for asking why a student accepted one output, rejected another, or revised the prompt after seeing weak results. ### Where can readers see the next step in this discussion? The May 23 AdamLGRing thread on X included examples and links to classroom prompt templates, according to the source briefing. (teaching.berkeley.edu) Stanford, Berkeley and Harvard also maintain public teaching pages with assignment-design examples and AI policy guidance for instructors revising courses now. (teachingcommons.stanford.edu)