Let’s Talk about Artificial Intelligence
At the Davis Institute for Artificial Intelligence, conversations about the consequential technology abound
As artificial intelligence powers more and more of our daily lives, it’s fitting that conversations about the technology are on the rise. At Colby, they’ve ballooned since a year ago.
“The vibe on campus about AI feels very different this year,” said Michael Donihue, interim director of Colby’s Davis Institute for Artificial Intelligence and the Herbert E. Wadsworth 1892 Professor of Economics. “Before, students were often afraid to mention AI, and faculty largely kept their heads down. Now, everybody’s talking about it across campus.”
There to guide the momentum is the Davis Institute for Artificial Intelligence, or Davis AI, a national leader facilitating conversations about AI with faculty and students. Davis AI is in its fourth year promoting interdisciplinary, human-centered approaches to AI in the classroom and faculty scholarship. It remains the only facility of its kind at a liberal arts college.
Committed to deeply embedding its work in the academic program, Davis AI is expanding its resources and programs to promote hands-on learning, dialogue, and interdisciplinary collaboration. New this year is Mule Chat, a tutoring program designed to lower the anxiety level of those curious about or fearful of AI. It’s a safe space to experiment, ask questions, and discuss AI usage policies.
Trained student tutors enrolled in the new Artificial Intelligence Lab course meet with users for one-on-one sessions to introduce a custom-built platform, teach prompt engineering (designing inputs or questions to produce optimal outputs), and have meaningful conversations about appropriate uses for AI, said Donihue.
Tutor Maddie Puzon ’26 understands people’s hesitancy around AI. “It’s this whole new thing that will really change how people do work,” said the junior from Glen Ellyn, Ill., double majoring in philosophy and computer science with a concentration in AI. “I think Colby is doing a good job of turning it into an asset. Instead of it being something they have to damage control, saying students are not allowed to use AI, it’s saying, let’s teach people how to use it responsibly.”
Equity in access
Mule Chat is designed to level the playing field and provide equal access to advanced large language models such as Open AI’s ChatGPT 4, Google’s Gemini, Anthropic’s Claude, and Meta’s LLaMA. These AI tools are forms of generative artificial intelligence that generate new material such as images, text, or music based on patterns learned from large datasets.
The Mule Chat platform allows users to compare these models side by side, which was alluring to Professor of Psychology Chris Soto. A leading personality psychologist, Soto researches the Big Five personality traits and has used AI to help him fine-tune his personality tests for various populations or contexts.
Even with his experience, Soto signed up for a Mule Chat session and was paired with Puzon, who gave him a tour of the platform and allowed him to explore generative AI models new to him. Soto feels it’s important for the College to prepare for how AI will change student and faculty work. It’s essential, he said, to provide a baseline set of AI tools to become familiar with them, and from an equity perspective.
“I think Davis AI and Mule Chat specifically are doing a good job of testing things out and putting guardrails in place so that at some point in the next couple of years it will make sense to open it up and make sure that everybody has access to some of the baseline tools.”
Equity is at the heart of the Mule Chat program. Although free versions of generative AI models exist, they have severe limitations. Some students can afford paid versions, others cannot. Faculty may not have funds or want to spend them for advanced versions if they’re unsure it’s the right fit. With Mule Chat, Davis AI allows users to explore advanced versions without paying anything upfront.
“Generative AI can be a tool or partner in the learning process, but not a replacement.”
Michael Donihue, Interim Director of the Davis Institute for Artificial Intelligence
AI in the classroom
Donihue reports that the number of professors who prohibit the use of AI in the classroom is diminishing. Still, many need assistance figuring out policies. Mule Chat can help educate faculty on how generative AI works so they can better set guidelines for how it can or can’t be used in their courses.
Linshan Jiang, a visiting assistant professor of history, is a case in point. This semester, she has prohibited students from using AI in her courses. She believes that students aren’t always aware that AI-generated ideas are often vague or overly general.
In her Mule Chat session, Jiang found tutor Elsa Grant ’25, a psychology major, helpful and professional.
“Since she is both a student and a tutor, she can see the benefits and drawbacks of using AI for learning from both perspectives of professors and students,” said Jiang. “Her idea that AI cannot think for us, but can help us think is really helpful for me to sustain the idea of critical thinking in an AI-filled environment.”
In the spring, Jiang will take a different approach.
“Next semester I will definitely engage with AI. I think we can’t escape it. That’s my attitude now. Davis AI suggested I discuss with students how we should engage with AI instead of me setting up the rules for them. We should talk about it together.”
Across campus, Soto is experimenting with a use-and-disclose policy in his 200-level psychology course this semester. While students can’t ask ChatGPT to write a paper for them, they’re free to experiment with AI tools. If students use them to help generate ideas or provide feedback, Soto requires them to submit their chat history as a supplemental document. Using AI won’t affect anyone’s grade—Soto is simply gathering information.
“My hope is that by having this kind of transparent, non-punitive policy we’ll get a clear sense of how many students are using it and how they are using it.”
AI a partner, not a replacement
As a Colby student, Puzon has loved “exploring the space of what AI can do,” including research with computer science professors. Initially, much of what she learned about AI was largely on her own, which compelled her to become a Mule Chat tutor.
“I wish I had someone to help me figure out how to use AI models effectively. I know as much as a junior undergrad student does, but I felt like I know enough to guide people in the right direction if they’re looking for help.”
Student interest like Puzon’s is a large part of what’s driving the push toward embracing AI. For faculty, this student-driven charge differs from change that generally evolves within the liberal arts curriculum or specific disciplines.
To gain traction with faculty hesitant about the impacts of AI on their teaching, Donihue emphasizes that the usefulness of generative AI depends on how you ask the question and then validate and contextualize the results. Teaching students to ask good questions is what we do at Colby, he said, and critical thinking skills are inherent in the learning goals for all of us.
“Generative AI can be a tool or partner in the learning process, but not a replacement,” said Donihue, drawing on decades in the classroom and his current post in Davis AI. “We’ll all be challenged to think about the ways we’ve taught in the past, but this isn’t new for us. It’s what we do well as Colby faculty.”
