An MIT committee is urging the institute to overhaul how it teaches and evaluates students, arguing that the rise of AI demands a shift away from traditional tests and toward hands-on projects, oral exams, and in-person collaboration. The report from MIT's Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training, released several weeks ago, outlines specific changes to keep assessment meaningful as AI tools become more capable and widely used.
The committee warned that AI has introduced risks to the core student experience. It pointed to greater isolation, the erosion of social contact between students and instructors, and a weakening of student mastery and confidence. The report also said AI is making it harder to gauge what students actually know. Every subject at MIT, the committee concluded, needs to be re-examined to account for how students learn and demonstrate knowledge in an environment where AI is readily available.
Moving beyond exams and problem sets
The report pushes instructors toward assessment methods that are harder for AI to substitute. Oral exams, semester portfolios, and out-of-class assignments paired with in-class discussions all received specific mention. These approaches would make teaching assistants and class time more central to how student progress gets measured. The shift moves evaluation away from take-home work that an AI can complete and toward direct, observable demonstration of understanding.
Grading itself needs a rethink, the committee said. It called on MIT to explore alternative grading systems that place more weight on projects and oral presentations. Traditional test-based evaluation, the report noted, is more open to AI assistance in ways that undermine what the grade actually reflects.
Hands-on work and research take center stage
Experiential and project-based learning should play a larger role across the curriculum, the report recommended. To make that happen, MIT needs to support instructors in building the skills to facilitate hands-on learning effectively. The committee also urged the institute to expand out-of-class research opportunities and career experiences that emphasize mentorship, collaboration, and learning by doing.
Every subject, regardless of size, should include a regular in-person social component. The report said instructors must structure these interactions intentionally to hit specific learning goals. The point is not just to get students in the same room but to design exchanges that deepen understanding in ways that solitary, screen-based work cannot replicate.
Clear rules for AI use
The committee wants MIT to develop a clear, consistent set of AI use guidelines. Departments would choose from a menu of options and adapt them as needed, but the report stressed that a coordinated approach matters. Students moving through a major should encounter guidelines and rationales that are largely consistent from course to course, rather than a patchwork of individual instructor policies.
For educators navigating similar questions, resources on AI for Teachers offer practical strategies for integrating AI into teaching while preserving academic integrity. Broader discussions on AI for Education examine how institutions are rethinking curriculum and assessment in response to these tools.
Why this matters for educators
MIT's recommendations signal a direction that other institutions are likely to watch closely. The core tension is familiar to any teacher: if AI can produce a passing essay or solve a problem set, what does the grade actually measure? The report's answer is to lean into methods that require students to think and articulate their reasoning in real time, in front of people. For educators, that means building more discussion, presentation, and project milestones into course design, and pushing departments to agree on AI policies rather than leaving every instructor to figure it out alone.
Your membership also unlocks:
