MIT recently released the final report from its Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training. If you work in education and are thinking seriously about artificial intelligence, I highly recommend reading it.
The report is packed with ideas, recommendations, questions, and provocations about what AI means for teaching and learning. It is the kind of report I expect to keep revisiting as I think through some of the larger questions it raises.
That may be what I like most about it.
MIT’s report focuses on artificial intelligence, yet much of what it raises reaches into the core of education itself.
What are we trying to teach students? How do people actually learn? What kinds of intellectual struggle are important to that process? How do we know when learning has occurred? What should the relationship between students and faculty look like? Which parts of education become even more important as technology becomes increasingly capable?
Several ideas from the report have particularly stuck with me.
1. AI can expose weaknesses in education that were already there.
When an AI system can successfully complete an assignment, it creates a reason to look closely at what that assignment was intended to accomplish.
That does not make every AI-completable assignment a bad assignment. It does mean we have to become much more explicit about what students are supposed to learn from doing the work.
2. Some struggle is necessary.
One of the ideas I keep returning to is the importance of productive intellectual struggle.
Learning is not always supposed to be efficient. Wrestling with a problem, getting something wrong, trying again, and figuring out why an approach failed can be central to the learning process.
AI can remove friction. Sometimes that is enormously valuable. Educators still need to identify which forms of friction get in the way of learning and which ones contribute to it.
3. “Augmentation, not automation” is a useful way to think about AI.
This may be one of the simplest and most useful principles in the report.
A productive question is what AI is enabling students to do.
Is it helping students think more deeply, explore more possibilities, receive better feedback, or attempt work that previously would have been beyond their reach? Is it reducing the amount of thinking students have to do for themselves?
Those distinctions matter when we decide where AI belongs in a learning experience.
4. AI should allow us to raise the ceiling.
The report offers examples of students using AI to undertake more ambitious work than they could previously accomplish within a semester.
I love that idea.
The promise of AI in education includes the possibility that students can tackle harder problems, create more ambitious projects, test more possibilities, and spend more time on higher-level decisions.
That gives us an opportunity to reconsider the level of work we ask students to do.
5. We need to rethink assessment.
AI makes it increasingly difficult to assume that a finished product tells us what a student knows or can do.
That creates obvious challenges, and it also gives us a reason to reconsider how we assess learning. Oral assessments, demonstrations, portfolios, authentic projects, iterative work, reflection, and other approaches may become increasingly important.
The central question becomes much more focused on what a student can demonstrate they know and can do.
6. Human interaction may become more important as AI becomes better.
This was another part of the report that stayed with me.
AI can provide explanations, feedback, tutoring, brainstorming, and assistance at virtually any hour. Those capabilities are extraordinary.
Learning is also deeply social.
Students learn through conversations with faculty, arguments with classmates, study groups, mentoring, collaboration, office hours, and simply spending time around other people who are wrestling with the same ideas.
As AI becomes more embedded in education, we need to pay close attention to the role of human connection in the learning experience.
7. There probably should not be one universal rule for AI.
Different disciplines teach different things. Different assignments serve different purposes. Students at different stages of their education need different kinds of support.
Simple institution-wide declarations that AI is either “allowed” or “prohibited” become increasingly difficult to apply meaningfully across all of those contexts.
The learning objective offers a better place to begin.
What are we trying to help students learn? How does AI affect that goal? Does it support the learning, interfere with it, or change what we should be asking students to do?
8. AI may force us to become more intentional about education.
This may ultimately be my biggest takeaway.
For years, higher education has relied on familiar structures: lectures, papers, problem sets, exams, grades, credit hours, and semesters. AI is putting new pressure on some of those structures.
That pressure can lead us back to a fundamental question:
What is education for?
I started reading MIT’s report because I am deeply interested in AI and education. I came away with a long list of questions I expect to keep thinking about.
What should students know? What should they be able to do? What experiences help them grow? What deserves to remain difficult? What should technology make easier? What happens between human beings in the learning process that we need to protect and cultivate?
These are enduring questions in education. Generative AI has given them new urgency and made the consequences of our answers much more immediate.
The MIT report offers a thoughtful framework for wrestling with them. That is one reason I expect to return to it.
If you are thinking about AI, teaching, learning, or the future of higher education, put this one on your reading list.
The distinction between productive struggle and unnecessary friction really resonates.
AI is also forcing higher education to confront a different kind of friction. Society is simultaneously questioning the core value proposition of a college education while demanding something increasingly transactional: immediately relevant career readiness and skill development for a rapidly changing economy.
The challenge is that the speed of change outside higher education is very different from the speed of responsible change within it. Developing sound educational content takes expertise, testing, assessment and discipline. Then come the regulatory and oversight systems designed to ensure quality... processes that can mean months or years before something reaches students.
Society doesn't see those slow and deliberate gears turning. But people absolutely see and feel the consequences of being left behind in a changing career and economic landscape.
That may be one of the more difficult questions AI is forcing higher education to confront: How do we preserve the rigor and safeguards that give education value while becoming responsive enough to deliver that value when students actually need it?