By Elizabeth McAlpin, Director of Educational Technology Research, NYU IT
At a recent SoTL TeachTalks session, I was struck by a recurring theme that sat quietly beneath the iterative design, technical work, research frameworks, and data visualizations: Students are not waiting for the perfect learning technologies to arrive. Instead, they’re designing them.
Two NYU student teams, working in very different disciplines, shared projects they developed to address challenges in higher education: how to build confident presenters, and how to provide meaningful feedback when time is limited. Their tools, CaPS and GradeGPT (renamed OSCE-AI), frame AI not as a disruptive force to be managed, but as a set of co-design materials for shaping better learning experiences.
Centering Practice, Not Performance
The Communication and Presentation Skill Advancement (CaPS) team, consisting of six students and two supporting faculty from the Steinhardt School of Culture, Education, and Human Development, and one content expert faculty from Stern School of Business, began with a deceptively simple observation: We often assume students “just have” communication skills, when in reality these skills require intentional practice, structured feedback, and supportive environments to develop them. The team responded by creating a program that breaks presentation skills into manageable subskills, pairing expert guidance via video and repeated, low-stakes practice.
What makes the work compelling is not the novelty of the activities, but the way the design makes practice feel possible. When the team introduced the AI CaPS Companion, students described it as supportive, personalized, even confidence-building, with one calling it “A real companion—you feel continual guidance.”
Feedback wasn’t simply corrective; it felt engaging, empathetic, and generative. And importantly, students felt less alone in the work. As one user explained, “If [activities are] purely self-directed and online, it’s a lot more isolating . . . with this I feel motivated to keep up with the work.”
The team’s research highlights something many instructors know intuitively but struggle to operationalize: learning to communicate requires psychological safety, low-stakes practice, and feedback that creates momentum rather than anxiety.
The fact that these insights emerged from NYU students as designers—not external vendors—feels meaningful. Their work illuminates not just what students need, but how they want to feel while learning.
The AI CaPS program has scalability to other departments and schools (e.g., professional studies, business, law) where students need to build their communication and presentation skills not only for their course requirements but for employment and the workplace. If you would like a demonstration or more information contact student team members Eden Lee, Yinuo Ma, Xixi Tan, Chole Huang, Ziyun Chen, or Zulsyika Nurfaiz, or faculty team member Maaike Bouwmeester.
Reimagining Feedback as a Scalable Resource
Where the CaPS team asked, “How do we help students practice practical skills for work?”, the GradeGPT team asked, “How do we help students improve and get timely feedback on OSCE exams when faculty time is finite?”
In the context of medical training, students spend hours writing OSCE notes but often receive limited narrative feedback, making it difficult to understand strengths, gaps, or next steps.
Three students from Grossman School of Medicine responded to this need with GradeGPT (renamed OSCE-AI), an AI-driven feedback tool that evaluates OSCE notes using rubrics and generates meaningful narrative comments tailored to the student’s work.
Early research suggests that GradeGPT grades similarly to human evaluators and can differentiate high- and low-performing notes. During the TeachTalk, one team member reported that “GradeGPT and human graders are in agreement on high-scoring versus low-scoring notes.” Just as importantly, the tool gives students something they rarely receive at scale: readable, varied narrative feedback.
There is something both pragmatic and optimistic in this work. Rather than accept feedback scarcity as inevitable, these students asked: What if reflection and improvement didn’t depend on time and labor being infinitely available?
The GradeGPT (OSCE-AI) program also has scalability to other departments and schools (e.g., Dental, Nursing, and Social Work) where students also take OSCE exams and could benefit from immediate, actionable, and meaningful feedback. If you would like a demonstration or more information contact student designers Caroline Magro, Olivia Schaffer, or Andy Qiao, or So-Young Oh, Director, Division of Digital Learning, NYU Grossman.
. . . Innovation in education doesn’t need to come from corporate labs or top-down mandates. Sometimes it emerges from students asking, “What do we wish existed, and what would it take to build it?”
The Pedagogy Behind the Innovation
What stands out in both efforts is their grounding in learning research and validated with research to demonstrate learning impact.
The CaPS team drew on self-determination theory, deliberate practice, and social learning theory, translating them into design features that foster competence, autonomy, and belonging.
The GradeGPT team emphasized iterative design, stakeholder feedback, and dissemination, not only to build a usable tool, but to consider adoption, sustainability, and impact.
It’s easy to view student-created technologies as quick hacks or clever prototypes. These projects resist that framing. They are thoughtful, theory-informed, research-based attempts to make education more humane, scalable, and equitable.
What I Keep Thinking About
I left the session reflecting on two quiet truths:
- Students have an inside view of the pain points we often overlook or don’t have the bandwidth to address.
They understand not just where systems break, but how those breaks feel. - When students design solutions, they tend to optimize for agency, not efficiency.
Their tools help learners practice, reflect, try again, without fear of failure or lack of oversight.
The AI conversation in higher ed often centers on risk, disruption, and policy. These projects offer a different narrative: AI as a medium for student-driven problem solving and human-centered design.
A Future That Feels More Collaborative Than Competitive
I am reminded that innovation in education doesn’t need to come from corporate labs or top-down mandates. Sometimes it emerges from students asking, “What do we wish existed, and what would it take to build it?”
CaPS and GradeGPT (OSCE-AI) don’t try to replace people. They try to help students learn in ways that feel more supported, empowered, and connected to their aspirations.
If anything, these tools model what students seem to be asking of us: less emphasis on binary scores; more emphasis on guided practice, meaningful feedback, and sustained growth.
Perhaps that is the real innovation, not the AI itself, but the shift toward learning environments where students can experiment, design, and lead.
Lastly, the talk reminded me why I love NYU so much — the students who have grit, perseverance, and intelligence, and the faculty who guide them along the way.
Hear more from the student design teams at Grossman and Steinhardt by watching their recent TeachTalk, Students as Educational Innovators: User-Designed Tools for Learning.
For a deep dive on OSCE-AI (formerly GradeGPT), check out the team’s widely circulated article: Magro C, Attal K, Li V, Schaffer O, Qiao A. GradeGPT—Generative AI for grading post-OSCE notes. Med Educ. 2025. https://doi.org/10.1111/medu.70044