Grounding in course materials
Faculty-provided resources establish a course-specific basis for generated responses.
HUMAN-CENTERED AI · EDUCATIONAL TECHNOLOGY
Meet NORMAN.
Teaching assistance with traceable sources.
A retrieval-augmented generation teaching assistant designed around faculty-provided course materials, transparency, and responsible access.
THE CHALLENGE
Useful answers need an understandable foundation.
AI teaching assistants raise questions about hallucinations, source transparency, and the handling of educational information. NORMAN addresses these challenges through a human-centered system design that grounds assistance in faculty-provided course materials.
I co-authored this work with M. Donovan, S. Shadangi, D. Ganskop, and C. Beaton. The contribution connects retrieval-augmented generation with the practical requirements of teaching and learning.
THE SYSTEM
Course-grounded assistance.
Retrieval-augmented generation, or RAG, supplies relevant material to an AI model as context for generating a response. In NORMAN, faculty-provided course materials form that knowledge base, and traceable source attribution helps make the basis of a response visible.
01
Faculty materials
Course-specific knowledge
02
Retrieval + generation
Relevant context for assistance
03
Traceable response
Answers with source attribution
Conceptual overview of the approach.
DESIGN PRIORITIES
Transparency, security,
and usability.
Traceable source attribution
Visible sources help users understand the information behind an answer.
Role-based access
Access controls are part of the system’s approach to responsible use in an educational setting.
On-premise AI infrastructure
Institutionally situated infrastructure supports the project’s security and data-handling goals.
CONTRIBUTION
Responsible AI as a design practice.
NORMAN brings course grounding, source visibility, access control, and infrastructure choices into the design of an educational assistant. Its goals include reducing hallucinations and improving transparency, security, and usability.
The project complements my research on institutional readiness by considering how responsible AI principles appear in a concrete system. The focus here is on the published system and its design rationale.
RELATED PUBLICATION
NORMAN: A Retrieval Artificial Intelligence Augmented Teaching Assistant for Higher Education
Donovan, M., Shadangi, S., Ganskop, D., Beaton, C. & Mahajan, R.
EDULEARN 2026 Proceedings · Conference paper