How one college is leveraging AI for educators and students

The explosion of artificial intelligence tools like chatbots has rocked the education world in the last couple years. It’s spurred efforts to prohibit, detect or otherwise build guardrails around these powerful new tools.

Some educators, though are embracing them, and Colby College is doing it on an institutional level. Four years ago, before most of the public had ever heard about large language models, this private liberal arts college in Maine established a cross-disciplinary institute for AI to help educators and students integrate the technology into their curricula in an ethical way.

We had the college president on back then to discuss, and today we wanted to check back in. This time with Michael Donihue, interim director of the Davis Institute for AI at Colby College. The following is an edited transcript of his conversation with Marketplace’s Meghan McCarty Carino:

Michael Donihue: As a liberal arts institution, these cross-disciplinary conversations are a natural part of what we do, and AI just provided an opportunity to think more deeply about conversations around ethics and moral uses of this particular technology. I think what made it different, though, for us, was that instead of being driven by the discipline, these are being driven by our students. They’re all using AI, and so figuring out how to align that with what we’re trying to do in the classroom, I think, presented the biggest challenge for faculty here.

Meghan McCarty Carino: Right, there’s been a lot of angst, I think, in the context of higher education about students using these. How has Colby [College] turned things on its head?

Donihue: So we’ve done that in a couple different ways. These generative AI applications, they only work if you provide a really good prompt, and there’s a whole field now of prompt engineering. But teaching students to ask good questions is something we’ve done, and we do it really well, and then the results you get from a generative AI application requires you to deconstruct the response and put it together in ways that are understandable and might facilitate learning. And we call that critical thinking. And again, that’s part and parcel of what we’ve done well at Colby for a long, long time.

McCarty Carino: What are some of the ways that the [Davis Institute for AI] is helping Colby to integrate AI education into the broader experience?

Donihue: You know, one of the most effective things we’ve done this year has been to create a portal. So Colby College’s mascot is called the Mule, and we created something called Mule Chat, where under a paradigm of equity of access we’ve enabled faculty and students and staff to come to the Institute and experience the most advanced versions of several different generative AI models in the context of what they’re doing. For students, it might be a research project or a class assignment, and for faculty, it’s sort of trying to align their learning goals with this tool, and also think about perhaps other ways of crafting assignments that might take advantage of the potential of generative AI. It’s possible to get free versions of these things, but it’s not possible for everyone to have access to the most advanced ones, and you get very different results for both teaching and scholarship, and it’s proved to be remarkably effective.

McCarty Carino: What kind of feedback have you gotten from students and instructors?

Donihue: It’s interesting. We did a survey this year of students, and we had about 659 responses, about 30% of the campus, and it’s broadly represented by major and class year, and turned out over 90% had used generative AI, with more than half of those students identifying themselves as being average or above average in their ability. And the primary uses have been for editing written work and solving math problems. But 92% of the students reported that their professors have provided at least some guidance on the ethical uses of AI in academics, and 80% indicated that generative AI has had either a positive or a neutral effect on their academic performance. There’s some real skepticism, though, about what it’s doing to some of their social lives. And 20% said outright, it’s inappropriate to use generative AI for academic assignments. So they’re having some healthy conversations as well.

McCarty Carino: And have you been contacted by other schools interested in kind of replicating Colby’s model?

Donihue: We have. We’ve been contacted by a lot. We just established an academic partnership with the Max Planck School in Germany, and they’re interested in some help in developing their AI strategy. We had a mini symposium, if you will, with a collection of our four peer institutions, and had a gathering of directors of centers for teaching and learning, writing departments, academic technology support services to come to Colby and see what we’re doing at [the] Davis [Institute for] AI, and then to share sort of strategies and ways we might cooperate going forward in developing AI applications on our campus.

McCarty Carino: Are there any pain points that you see in kind of integrating this technology into the broader education, particularly when it comes to just the reliability of the tools as they are?

Donihue: Yeah, I think the pain points primarily are when faculty don’t talk to their students about AI and appropriate uses. We don’t have one institution-wide policy on the use of AI. What we’ve opted instead to do is to allow departments and programs and individual faculty to develop their own AI policies. We provided a guiding principles document that helped them do this. So we have lots of conversations here at Davis AI with individual faculty about how to incorporate AI into their courses. But we found that if you have an initial conversation with your students and engage them in this notion of developing appropriate uses, the anxiety disappears almost immediately, and you move quickly to conversations around, well, here’s how I would use it to achieve this learning goal, or here’s why I’m asking you not to use it for this piece. And in my own teaching, in my own classes, we’ve had the same conversations, and you can focus on things like communication skills and trying to think for yourselves.

McCarty Carino: Where do you see these efforts to increase AI literacy and embedded in your curriculum going in the coming years as this technology continues to evolve?

Donihue: Yeah, well, we have this Mule Chat portal, and what we’re working on now is an offspring we’re calling Mule Bot, which is a trained assistant to Mule Chat, in which you can take Colby data, like the course catalog, and train it up so we can ask very specific questions about what’s going on at Colby. And then, we’re moving from that to more what are called on-premises applications. These can include things like Mule Bot, but might be things where you don’t want images or your research or your teaching materials uploaded to one of the commercially available vendors. And so we’re establishing some of these. I had one of my students in my lab here demonstrate small you could create a useful on-prem large language model on an existing laptop. And so that’s pretty exciting. The other area that looks to be disruptive probably are the notion of AI assistance for course design, and we’re going to pilot some studies here at Colby offering faculty small stipends to experiment with these in their classrooms and then report back. And it’s that peer-to-peer sharing of experiences that presents the most fruitful opportunities for engagement.

Stories You Might Like

Should every B.A. include some AI?

The ABCs of AI, algorithms and machine learning (re-air)

The ABCs of AI, algorithms and machine learning

Parents, educators are unaware how their students use generative AI, report finds

What happened when an entire class of college students had ChatGPT write their essays

States and schools are learning how to manage AI in education