HelpMe AI is an instructor-guided chatbot that pairs Retrieval-Augmented Generation with a Human-in-the-Loop validation framework. In MET, we are studying how it can support critical AI literacy in graduate education.
About the research →
Students are already using AI tools, often invisibly. This research asks how graduate educators can make that invisible literacy development pedagogically intentional.
Most commercial AI tools operate as black boxes: students send a prompt, receive a response, and have no visibility into what the system was trained on or how its answers are shaped. HelpMe AI works differently. It is designed around human–AI co-agency: instructors validate AI responses, review student–AI interaction logs, and iteratively guide how the system adapts to their course — making the human-in-the-loop visible as a pedagogical practice rather than hidden as backend quality control. Each course instance is anchored in its Canvas materials and an instructor-curated library of course-specific texts, semantically searchable through retrieval-augmented generation.
Instructors may also offer students examples of ways to use HelpMe and critically engage with its outputs as part of course activities — modelling the kind of reflective, questioning stance that AI literacy requires. Studying these structured engagements helps us understand how students take up the system: what they ask, what they trust, what they push back on, and what they ignore. At the end of the term, a short survey invites students to reflect on their experience.
Pedagogy & approach →
Three commitments shape how we use HelpMe:
- Transparency — instructor validation is visible to students.
- Co-agency — students engage as co-investigators.
- Inquiry — the tool itself is an object of study.
Participating instructors embed HelpMe into a course activity rather than leaving it as a passive sidebar tool.
Tutorials & guides →
FAQ for students →
Do I have to use HelpMe?
No — it’s voluntary and has no impact on your grade.
How is it different from tools like ChatGPT?
HelpMe is grounded in course materials, and instructors can validate or refine its responses. A few things set it apart from commercial tools:
- Course-grounded answers. HelpMe draws on your Canvas course materials and an instructor-curated library of course-specific texts. When you ask a question, it retrieves relevant passages from those sources and uses them to shape its response — a method called retrieval-augmented generation (RAG). Answers are anchored in what your instructor has chosen, not the open internet.
- Instructor oversight. Instructors can review student–AI interaction logs, validate responses, and refine how the system answers future questions. The “human-in-the-loop” is a visible part of the system, not a hidden backend process.
- Runs on UBC infrastructure. HelpMe uses open-weight language models — including Qwen, DeepSeek, Gemma, Llama, Phi, and GPT-OSS — hosted locally on a UBCO GPU server through Ollama. Your questions and conversations are processed within UBC’s environment rather than sent to an external commercial provider.
- Not a commercial product. HelpMe is a UBC-developed research and teaching tool, not a service sold by a private company. Your interactions are not used to train commercial models, generate advertising data, or feed any third-party system.
- Open source. The HelpMe codebase is openly available, meaning its design, logic, and data handling can be inspected, studied, and improved by the academic community.
What happens to my interactions?
Logged for research and course improvement, handled per UBC ethics protocols. Full details in the consent form.
Can I use HelpMe alongside other AI tools?
Yes — comparing tools is part of the pedagogy.
Team →
Rachel Horst — MET pilot lead · rachel.horst@ubc.ca
Ramon Lawrence — HelpMe principal investigator, UBC Okanagan
Graduate Research Assistant: Sooyoung Yang