Cornell instructors are developing new ways to teach with artificial intelligence, from a robot planned for a largely device-free class to historical chatbots that students can question and challenge.
This year, 33 instructors are taking part in the AI Experimenters Faculty Fellowship, a Cornell AI Initiative program for trying new classroom tools and rethinking teaching methods, according to the university’s announcement.
Steven Jackson, vice provost for academic innovation and leader of the initiative’s education work, said faculty are learning from experiments and sharing results while recognizing both AI’s benefits and harms.
Robot lessons without screens
Scott Seidenberger, a postdoctoral associate in human resource studies in Cornell’s ILR School, is building a robot and custom chatbot for his AI and HR course next semester.
The agent, named Hopper after computing pioneer Grace Hopper, is being built from scratch so students can examine its logic and help update and improve it. Seidenberger wants future human resources professionals to participate in decisions about AI systems that increasingly handle sensitive tasks and information.
Hopper can answer basic administrative questions about the class, though Seidenberger said it is far less capable than a person. Giving the system a robot body lets students explore how a physical presence might change people’s relationships with AI.
His lectures and exams will otherwise be device-free. Students will put away laptops, phones and tablets during lectures, using their two 75-minute class sessions each week over 15 weeks to focus on learning together.
Checking answers in biomedical engineering
Suélia Fleury Rosa, an associate teaching professor in Cornell’s Duffield College of Engineering, is designing assignments that start without AI before asking students to use it and check its answers.
Her Digital Control Strategies for Biomedical Applications students are preparing to work on devices such as surgical robots and prosthetics, which must process information from the body accurately and quickly. The exercises are intended to help them recognize answers that are incomplete, misleading or technically wrong.
“A response that sounds plausible is not enough,” Rosa said. “Students need to demonstrate that it is mathematically, computationally and physiologically reasonable.”
Rosa, an AI Experimenters fellow, said preparing classes takes far more time than when she began teaching 30 years ago because she tests prompts and anticipates problems students may encounter.
Debating with historical figures
Pamela Karimi, an associate professor and associate dean in the College of Architecture, Art and Planning, developed historical chatbots for Architectural History I, a first-year survey spanning antiquity to the Enlightenment.
Students study John Evelyn’s 1661 work “Fumifugium,” which proposed changes to clear London’s smoky air long before the Industrial Revolution. With help from Andre Hafner, the college’s director of IT Solutions, Karimi created AI versions of King Charles II and Parliament, to whom Evelyn addressed his proposals.
Students took Evelyn’s role and debated green spaces and regulation of smoke-producing trades such as lime burning. His proposals were not implemented at the time; the exercise simulated discussions that might have occurred.
Karimi kept the tool focused on trusted primary and secondary sources so students could question its historically informed responses. Even errors can give students a reason to examine the simulation’s accuracy, she said.
She also sees potential for AI versions of figures such as Leonardo da Vinci and Filippo Brunelleschi to offer another critical voice in architectural reviews, rather than dictate whether student projects are good or bad.
Writing, bias and student judgment
Kate Navickas began teaching an AI and Writing First-Year Writing Seminar in fall 2024. An associate teaching professor in the John S. Knight Institute for Writing in the Disciplines and director of Cornell Writing Centers, she welcomes students with different views about AI.
Students set writing goals, meet one-on-one with Navickas, study how AI tools work and the biases built into them, and experiment with the tools while considering their uses and limits. Her goal is to help students gain confidence and make their own decisions about using generative AI in writing.
Katie Lu, a computer science and mathematics student who took the course in fall 2025, said one classroom experiment showed how similar AI-generated writing could become: Students used the same prompt, and nine of 10 responses began with “both.”
Lu said the exercise made her question whether using AI for creative work would leave room for her own creativity.
Supporting faculty experiments
Cornell’s Center for Teaching Innovation organizes the fellowship and offers consultations, training, webinars and department workshops. More than 1,000 instructors, about half the faculty, used its experts and programs last year, many for help with AI.
Executive Director Robert Vanderlan said some instructors have returned to assigning take-home essays after initially worrying about AI-generated work. They pair the essays with brief in-person conferences, where questions from a professor or teaching assistant can reveal whether students understand what they submitted.
Lu said reminders about the value of working through difficult material help her resist taking shortcuts and focus on learning.
FingerLakes1.com is the region’s leading all-digital news publication. The company was founded in 1998 and has been keeping residents informed for more than two decades. Have a lead? Send it to news@fingerlakes1.com.
