At a recent lunch gathering hosted by HAIL (the Hub for AI and Data Science Leadership) and the Mascaro Center for Sustainable Innovation, this summer’s MCSI research cohort sat down with faculty advisors and industry-connected guests to share project updates and dig into one of the most pressing questions in higher education today: What does it mean to learn in the age of AI?
The cohort's work reflects the full breadth of sustainability research at Pitt. Nathan Smith (advised by Dr. John Brigham, Computational Diagnostics & Inverse Mechanics Lab) is modeling how wind, light, and temperature interact with building facades. Sharvi Arora (advised by Dr. Setareh Kheiri) is developing a privacy-preserving, machine learning-based Federated Learning model to detect worker fatigue in industrial settings. Andrew Chen (advised by Dr. Justin Kitzes and Dr. Amin Rahimian) and Chloe Herron (advised by Dr. Justin Kitzes) are both working in bioacoustics, using computational tools to study ruffed grouse behavior and individual identification. Jiakai Dai (advised by Dr. Natasa Miskov-Zivanov) is simulating layered materials for solar panel optimization. Tengyue Zhang (advised by Dr. Paul Leu, Industrial Engineering) is developing antireflection glass for bifacial solar panels. And William Peralta (advised by Dr. Amin Rahimian) is examining how tools like ChatGPT are reshaping — and potentially eroding — the peer-driven knowledge communities of platforms like Stack Overflow. Each project brought its own set of challenges to the table, and together they set the stage for a conversation that was anything but abstract.
Hot Takes on AI and Education
Our most recent lunch turned into a lively structured discussion, with participants placing sticky notes along an agree-to-disagree spectrum in response to a series of statements about AI in education. We used color-coding (green for non-students, blue for students) to see whether perspectives diverged by role.
Statement 1: “If AI can answer an exam question correctly, there’s no reason a student needs to be able to answer it themselves.”
Nearly everyone landed somewhere between “disagree” and “strongly disagree,” but the reasoning varied in interesting ways. One faculty advisor framed it as a design problem: “The exam should be designed to test students’ proficiency with the material or their ability to think.” William Peralta, who is studying AI’s impact on online knowledge communities under Dr. Amin Rahimian, pushed back on the premise entirely: “It’s irrelevant if AI can do it. That’s not the point of the test. That’s like saying, if my professor can do it, why do I need to?” Sharvi Arora, researching worker fatigue detection with Dr. Setareh Kheiri, offered a more measured take: a student doesn’t need to answer with complete precision, “but they should ideally be able to reason through the topic itself.”
Statement 2: “Using AI to write code during a research project is no different from using a calculator to do arithmetic. It’s just a tool.”
This one generated the most debate. Dr. Ernie Benchek, part-time teaching faculty in Strategic Management, Entrepreneurship, and Ethics, challenged the analogy directly: “A calculator was built specifically and coded specifically to give an output. Generative AI is pulling from all kinds of things.” His point wasn’t that AI is the wrong tool — but that treating it with the same confidence as a calculator overlooks two important realities: it isn’t the right tool for every task, and its outputs always need to be verified.
“I’ve been using Claude to learn a research code for optimization. It definitely still hallucinates, and that’s the scariest part. If I didn’t know the field well, there’s no way I would catch it.” — Dr. John Brigham, Computational Diagnostics & Inverse Mechanics (CDIM) Lab
The group converged on a useful distinction: the function you’re using AI for matters just as much as the output it produces. Using it as a brainstorming partner or a starting point is fundamentally different from trusting it as a final authority.
Statement 3: “A student who uses AI to understand a concept has learned just as much as one who struggled through it on their own.”
The word “depends” appeared more than once on the sticky notes — and for good reason. The group agreed that cognitive engagement, not the tool itself, is the key variable. How much intellectual effort you bring to the interaction matters far more than whether AI was involved at all.
“AI is really good at gathering information from the public domain and summarizing it in an understandable way. Asking AI to explain a math concept can be very helpful — but it still depends on how you use it.” — Chloe Herron, bioacoustics researcher, advised by Dr. Justin Kitzes
Statement 4: “In five years, educators who refuse to integrate AI into teaching will be doing their students a disservice.”
Benchek didn’t think the field has five years: “I think people need to get on it now.” He described a conversation with the CEO of a Fortune 200 company who asked point-blank whether students were being trained to use AI and noted that some companies are already tracking employee AI usage against specific time expectations.
Dr. Michael Blackhurst, Teaching Assistant Professor and Environmental Studies Program Coordinator, raised a fairness concern that resonated with the room: “There are students who say they don’t want to use it but then use it anyway. It creates unfair classroom dynamics. Without some kind of institutional foundation, it’s really hard to achieve fairness.”
Dr. Brigham offered a historically grounded counterpoint, drawing on the arc of CAD software in civil engineering: “Twenty years ago, every curriculum had to include AutoCAD. Now those courses are gone because it’s no longer seen as a university’s job to teach a tool people can figure out on their own. I could see a very similar pathway here.”
“It’s too much of a norm in industry to not know how to use it. If it’s not introduced at the college or high school level, students are going to become too reliant on it on their own. Just knowing the basic skills of how to prompt and debug anything AI-related is important.” — Sharvi Arora, worker fatigue researcher, advised by Dr. Setareh Kheiri
Keep the Conversation Going
We want to hear what you think. Engage with us on LinkedIn to share your perspective, and stay tuned for future lunches — next up: industry–university collaborations and AI regulation.
Learn more about the MCSI Summer Research Training Program →