Scaling AI-Aware Teaching from Course to Campus (or, What I Thought About That MIT Report)

What does it take to help the entire faculty of a college or university learn and adopt new teaching practices?

This is a question I’ve had to grapple with twice in my career. The first time was during the COVID-19 pandemic in 2020. I was serving as the director of a center for teaching and learning at a research university whose faculty who needed to learn how to teach online. My team and I worked with hundreds of those faculty to learn new teaching skills and design effective online and hybrid courses in advance of a very unusual fall semester.

The second time came after the launch of ChatGPT in 2022. It was immediately clear, first to writing instructors then to all faculty, that generative AI posed significant challenges to traditional forms of assessment and evaluation. As AI technologies have advanced since then, many of us in higher ed have realized that responding to AI meaningfully involves much more than updating assessments–it requires reevaluating and sometimes redesigning all aspects of teaching.

That’s one of the central findings of a new “call to action” report from the Massachusetts Institute of Technology on navigating the “challenges and opportunities” posed by generative AI to the teaching mission of MIT. The ad hoc committee behind the report writes, “Most classes taught at MIT will require careful review and, in many cases, substantial adaptation to survive an encounter with AI without serious disruption.” The report argues that learning goals, teaching methods, and assessment strategies are all on the table for change.

Having worked with hundreds of faculty at the University of Virginia and elsewhere over the last four years, helping them adapt to those AI challenges and opportunities, I was glad to see the sense of scale and urgency in the MIT report. The report isn’t advocating that all faculty adopt AI technologies into their teaching, but it is arguing that all faculty should adapt to AI, finding the right set of AI-aware teaching practices for their unique contexts. I very much agree.

The MIT report seems to reflect well the local context: the nature of MIT students, commonly used assessments at MIT, the opportunities for undergraduate research at MIT. I’m sure this makes the report more practical for its intended audience, but it means the report is somewhat less generalizable to other institutional contexts. The recommendations about enhancing residential education, for instance, won’t resonate at places where online education is mission-central.

What kind of organizational development does it take to help the entire faculty of a university change their teaching practices? The MIT report outlines big plans: creating “AI Lead” positions for every department, launching an “AI Pilot Fund” to support course redesign, staffing an “AI Implementation Team” to provide needed expertise in AI and in pedagogy, and more. These are great ideas, and we are doing versions of them at the University of Virginia. Other institutions have similarly launched a variety of AI-focused programs for faculty. But how do you go from a small number of faculty-facing programs to institution-wide teaching change?

There are a few principles that I’ve found to be true in my decades of faculty development work, and these principles that can inform the choices made by academic leaders at all kinds of higher ed institutions when the goal is adapting teaching at scale.

If they will come, we will build it.

In the 1989 movie Field of Dreams, Kevin Costner’s character builds a baseball diamond in his cornfield. In a climatic scene, James Earl Jones’ character delivers a stirring speech encouraging the Costner’s farmer to persist in his seemingly irrational dream. “Build it… and people will come.” That advice might work great in a magical realism sports film, but I have found it isn’t the best advice when it comes to faculty development. “If they will come, we will build it” works better.

It’s true that if you offer workshops and lunch-and-learn seminars and collections of AI-aware assignments, faculty will take advantage and use what they learn to rethink and revise their teaching. There are 40 faculty and staff in the learning community I’m running this fall, and we’re going to learn a lot through our discussions. However, there seems to be a ceiling for participation in come-one-come-all events for faculty, even for those dealing with a challenge as universal as adapting instruction to AI. 

To reach past that ceiling, connect local academic leaders (department chairs, directors of undergraduate studies, assistant academic deans, and so on) who can gather faculty together with the people at your campus who can provide the professional development that those faculty need. When the chair of a department rings up the teaching center director and asks, “Can you lead a discussion about AI-aware assignment design with our faculty?”, the resulting conversation can help all the faculty in that department make more informed, intentional choices in their teaching, not just the ones who would opt into a campus-wide teaching workshop. And the conversation can continue long after the meeting itself, as faculty develop and use a shared language around their AI-aware teaching practices. 

Whatever your position in your organization, look for the people who can convene faculty and for those who can support those faculty–and connect them.

Motivation = Value x Expectation – Cost

That’s an “equation” we frequently use in our course design institutes at UVA as a way to talk about the levers we can pull to increase student motivation: when students value the work they’re doing in a course and they have confidence that they can do that work and the cost involved (often cost in time) is manageable, then those students are more likely to do the hard work of learning. Although this equation (based on expectancy-value theory) is targeting student motivation, it also offers a useful framework for faculty motivation.

If you’re planning initiatives to help faculty respond to the challenges and opportunities posed by AI, consider what motivates your faculty. Do they see the need for AI-aware teaching? As with the need to teach online in 2020, the need to respond to AI is very clear for many faculty members, so the “value” part of the equation is often a given. But if it’s not, how can you make the case to faculty? Faculty can be persuaded by student stories and voices, by decisions made by their faculty peers, and sometimes by the changing needs of the organizations that employ the students we graduate. 

What expectation of success do faculty have when redesigning their instructional practices and assessment methods to account for AI? Faculty have deep domain expertise, which is crucial to making effective teaching decisions, but responding to AI also requires pedagogical expertise and at least some level of expertise in AI technologies. How can you provide those forms of expertise to your instructors? The Norton Guide to AI-Aware Teaching, which I’ve co-authored with Annette Vee and Marc Watkins, is one source of expertise you can share with faculty. Consider also the people at your campus who can lend this expertise. Faculty might not have the time to keep pace with the latest AI models or pull together examples of AI-aware assignments or research best practices in oral assessment, but you likely have teaching center staff, academic technologists, librarians, or instructional designers at your institution who do. Build their presence into the programs you offer faculty.

The third piece of the equation is cost, and, as with students, that often means cost in time for faculty. Time is maybe the most precious resource faculty have, and they are naturally selective about how they spend it. How can you create the time that instructors need to do this work? The best solutions take significant resources (course release, student assistants, smaller classes). Pursue those if you can, but if not, look for smaller efficiencies. If faculty are going to spend an hour in a workshop, be sure to provide the curated ideas and resources they need to move quickly into action. If a small group of instructors are going to revamp a multi-section intro course, then build AI-aware teaching practices into that revision. Use the department meetings faculty were already planning to attend to discuss important questions about new learning goals in light of AI.

To be clear, the self-determination theory equation isn’t magical. It does, however, provide a way to think pragmatically about the resources faculty need to practice AI-aware teaching at scale.

We have a coordination problem.

Since the launch of ChatGPT, most of the faculty development work I’ve seen, including my own, has focused on helping individual instructors make informed, intentional choices about how to respond to AI in their individual courses. That’s valuable, important, and challenging work, but our students don’t take individual courses in isolation. They take multiple courses concurrently and over time. And, as the MIT report notes about the AI course policies that students encounter, “the rules and the way they’re conveyed vary widely from instructor to instructor.” I hear this again and again in student surveys and focus group. This point about conflicting course AI policies sounds quite typical for colleges and universities in 2026.

Which is to say, we have a coordination problem, so I’ll say plainly what the MIT report only hints at: Departments and schools need to identify shared values that will guide their responses to AI, then develop more coherent program-level AI policies, and, critically, program-level learning outcomes that reflect how AI is changing disciplines and professions. I’ve seen very few examples of program-level policies (although the University of Chicago seems to be on this) and even fewer examples of updated, AI-aware curricula, but that is work that needs to be done. And it will be, as more faculty find their feet in this AI world and more academic leaders see the need for meso-level responses. What are best practices for this kind of work? That’s to be determined, but I’m excited to figure that out.

Leave a Reply

Your email address will not be published. Required fields are marked *