UConn Leverages AI to Personalize Learning for Student Success
UConn is building an AI-driven learning ecosystem that supports every student, starting with neurodiverse learners and expanding across the university. The work is led by Professor Arash Zaghi (civil and environmental engineering) with a five-year NSF RED grant. Dean JC Zhao serves as the overall principal investigator to scale the approach across the College of Engineering and beyond.
The team's early research focused on students with autism, ADHD, and dyslexia in engineering courses. They found AI can contextualize content, deliver timely support, and create virtual coaching that improves retention and performance. The same tools now help a broader set of learners with different strengths and needs.
From one-size-fits-all to neuroadaptive learning
"Our research examined the challenges of trying to impose a one-size-fits-all model in teaching, and the capacity of AI to customize course material for diverse learners," says Zaghi. "We soon realized that AI could benefit all students, regardless of their learning difficulties, and we saw the value of trying to shift people's thinking from seeing neurodiversity as only a disruptive medical condition to viewing it as an untapped asset."
Fabiana Cardetti (mathematics), Sarira Motaref (professor-in-residence and associate director of innovation and student success), and Connie Syharat (research assistant and Ph.D. candidate in engineering education) round out the core team. Motaref's focus is course redesign for inclusion and neurodiversity.
Ethical, purposeful use of AI in the classroom
Motaref is clear about the instructional approach: "By integrating AI into coursework, we help students use it purposefully and ethically: as a personalized learning aid, a bridge between class content and their disciplinary interests, and a tool whose outputs must be verified through professional judgment. Students also learn where human expertise is essential and why copying AI output blindly undermines learning and jeopardizes careers."
For institutions and employers, this approach develops practical AI literacy and reduces misuse. It also offers a model that can be scaled across departments without losing academic rigor.
Math as a proving ground
Cardetti is leading AI integration in math courses. "This work is important for UConn students because it helps them overcome real barriers that research has shown they face," she says. "By building on that research, we can focus on easing challenges in their learning experiences, so they can make timely progress and succeed in reaching their educational goals."
She adds that the project is structured as a feedback loop: evidence informs practice, and practice informs research. The goal is simple-figure out what actually helps students learn and keep improving it.
AI4ALL: a campus onramp for first-year success
As an outgrowth of the research, UConn launched a new course called AI4ALL. Nearly 500 students enrolled in Fall 2025, with plans to offer it to all incoming freshmen by 2028.
The course helps with day-to-day hurdles: planning a schedule, staying on top of assignments, finding coaching and mentoring, integrating with classmates, communicating across cultures, and getting interactive support for anxiety and mental health concerns. It's a practical layer of support students actually use.
Stronger support, better engagement
"By providing students with more meaningful support and contextualized learning experiences, we hope to capitalize on students' assets to enhance their engagement and build key engineering skills during their time at UConn," says Syharat. "We are also building on our prior work in which we found that shifting our mindset about neurodiversity allows us to create more inclusive learning environments for a wide range of learners."
The team is now exploring how AI can create experiences that reflect individual strengths and challenges for all students, not just a subset.
Staying agile as AI changes fast
Zaghi points out how quickly tools shift: "When we started this project, AI was evolving so quickly that we had to scrap the first models we had created; I can't even imagine what things will look like in five years, by the time we finish this grant," he says. "Our goal is to remain agile and update our work as we go, continuously focused on mental health, general wellbeing, and improved ways of learning."
Dean Zhao emphasizes the outcome educators care about most: personalized learning with AI can help students reach their full potential, especially those who learn differently.
How educators can apply these ideas now
- Start with a focused pilot in a high-enrollment course. Measure engagement, persistence, and outcomes for different learner profiles.
- Publish an AI use policy for students: allowed use cases, required verification, and examples of misuse. Reinforce it in assignments.
- Build AI "explain and verify" steps into coursework. Require students to document prompts, outputs, and what they accepted, edited, or rejected-and why.
- Adopt an assets-based approach to neurodiversity. Pair AI supports with UDL principles to increase access without lowering standards.
- Create lightweight virtual coaching: study planning, time management nudges, and "ask for help" pathways that route to humans when needed.
- Run continuous improvement cycles. Collect data, share findings with faculty, iterate each term.
Further learning
- Explore additional AI training options by role: AI courses by job.
The bottom line: UConn's approach treats AI as a practical aid for inclusion, persistence, and performance. Start small, keep ethics front and center, and improve the system with every cohort.
Your membership also unlocks:
