School Project
Redesigning UM-GPT by Students for Students
Deliverable
For SI307, Introduction to Design, we were tasked with designing a landing page and interface for the University of Michigan’s LLM UM-GPT. The individual stages were broken up into bi-weekly assignments, starting from ideation and sketches and ending with hi-fidelity wireframes and a presentation.
Old Design
Thinking
As a user who preferred using a leading LLM service such as ChatGPT over UM-GPT, my ideation process started out with the following questions:
What did UM-GPT lack compared to ChatGPT?
- Personalized agents
- Organized project folders
- Chat history search functionality
What strengths does UM-GPT have over ChatGPT?
- Privacy - chats are not logged or used to train the model
- Tailored for students - context and access to University materials is integrated
- Free for students, staff, and faculty
What are three practical strengths of generative AI?
- Automating and reducing the amount of time spent on menial tasks
- Idea generation or problem solving
- Personalization
These questions guided me throughout my design process, determining which feature ideas to eliminate and how to implement the ones I kept.
Outside of these questions, I based certain UI decisions on interaction logic. Since the input field is the center of any LLM or chatbot, I emphasized the central call-to-action, reduced friction where possible, and clarified how new users would navigate between starting chats, choosing models, and accessing University resources.
