I came across MIT's report on AI in Education last week and consumed it fully and rapidly. I think it was excellent and fulfilled something I, and likely many educators, are longing for.

The report was developed by a committee written to inform the university's administration about generative AI in education and what to possibly do about it. The committee was primarily made up of professors, but also included a recent PhD graduate, a graduate student, undergraduate students, a librarian, and a teaching support staff director. There were no administrative staff in the committee. The professors are the front line educators and it is important that the chancellor tasked them to provide information via a bottom up approach.

There are several things that stand out to me in the report. Firstly, the take is very thorough in coverage of the pertinent topics and it is very thoughtful with a central focus on learning. They have concrete actionable recommendations that the administration can choose to enact. [1]

They do not hold back. They clearly state that generative AI is having significant negative effects to MIT's educational approach and student learning. This is in spite of it being a transformative technology that unlocks new possibilities. They highlight that the technology is growing faster than many other preceding technologies, giving educators and universities much less time to adapt. MIT has a longstanding "learning by doing" educational ethos that is detrimentally affected by generative AI. This is because AI can, in some cases, wholly relieve the doing, thus negating the learning.

They discuss how learning objectives need to be reassessed and that objectives may need to adapt. But they also recognize that not all learning objectives necessarily should change. Traditional assessment of student work is incredibly challenging when we no longer know if students have even written anything they submit. The role of teacher-to-student live contact and new forms of assessment are necessary to observe and extract the actual learning. To do this, the university needs to provide more teaching resources (instructors, space, funds, communities of practice, technical support, etc.) for in-person work and evaluation.

Their recommendations further include:

• Leverage project-based learning and social learning activities.
• Don't deprive students of learning experiences via internships and research by replacing them with AI substitutes.
• Reconsider if grades are even needed.
• Make AI use policies but explain why the policy is as it is.
• Be wary of AI detectors.
• Instructors should also have to disclose AI use to students to ensure the teacher-student social contract is not violated.
• Teach responsible AI use.
• Have AI leads, committees, and implementation teams and provide AI training to instructors.
• Ensure equitable access to technology for students and ensure their privacy.
• Monitor environmental costs and financial costs.

The only comment on theses and reports is that AI should be disclosed. I do not think this is sufficient. We already see that a large percentage of students using AI to write for them and it isn't clear to me that honest and full disclosure will occur or that the disclosuers will be useful in knowing what a student did and didn't write. There are other weird legal issues. For example, does a student own the copyright to a thesis that is partially generated by AI? There are lawsuits and law making around the world that range from output being public domain to the training data copyrights propagating to the output.

Generative AI is too easy and convenient for students under graduation and grade pressure to use. There is unfortunately a significant relaxation of ethics that seem to be a function of this convenience. It is important that students continue learning the fundamentals and that we can assess their learning. Instructors need tangible resources more than ever, most importantly additional instructor labor to engage with students in new live assessment approaches. Yet here we only see these resources decreasing and class sizes getting larger.

I'm curious what such a report generated by TU Delft instructors would look like. We may not find out, as it seems our policy may soon arrive directly from the administrative policy-makers with little voice from the front line.

[1]	Hackernews and Reddit had many commenters stating the report was fluff, which is just absurd.