GenAI Teaching and Learning Fellowship Project Update

GenAI Fellowship Project Updates

Project 1: Scalable Conversational Language Platform

Developing customizable AI-powered language practices for students

This project is exploring an instructor-facing platform that would make it easier to create course-specific language using chatbots. Using reusable templates and course materials, instructors can customize bots to reflect their teaching approaches, including their preferred tone and style of feedback.

Progress to date

The team has selected and tested an initial approach to chatbot development and has been working with participating instructors to build agents for their courses. Setup guides for prompts and voice components have been created, research ethics clearance has been received, and student recruitment has begun in participating French, Spanish, and German courses.

Learning to date

Developing effective GenAI-supported language activities requires considerable testing and instructor involvement. Early experimentation has shown that exercises can easily be too simple or too complex, making it important to refine prompts and learning activities to support an appropriate progression of difficulty. The team has also experienced firsthand how quickly changes to GenAI platforms can require adjustments to project design and implementation.

What's next

The team will continue recruiting students, implement the chatbots in participating courses, and gather student feedback. Evaluation will examine usability, interaction quality, linguistic accuracy, and alignment with course learning goals.

Project 2: GenAI Grading for Large Survey Courses: Enhancing Efficiency and Fairness

Exploring AI-supported feedback in large first-year courses

Project leads are investigating how a custom Microsoft 365 Copilot agent can support grading and feedback in a large first-year course. The agent generates suggested grades and feedback, while teaching assistants retain responsibility for assigning all official course grades. AI-generated feedback is supplemental and does not affect students' grades.

Progress to date

The project has received research ethics clearance, finalized recruitment and consent materials, and completed classroom visits introducing the study. Study information and a consent quiz are now available to students through LEARN. The team has also developed a working version of the agent and explored options for hosting it.

Learning to date

Initial testing suggests that the structure of an assignment can significantly influence the quality and accuracy of AI-generated feedback. Assignments that ask students to provide structured responses may be easier for the agent to assess consistently than more open-ended submissions. The team is also considering how technical infrastructure and hosting decisions may affect the feasibility of scaling this approach.

What's next

The team will identify consenting participants, prepare eligible assignment submissions, and provide AI-generated feedback alongside regular course feedback. Student perspectives will be gathered through a post-course survey and optional focus group, examining the usefulness, fairness, trustworthiness, clarity, and acceptability of AI-supported feedback.

Project 3: TALens: A Customized Teaching Assistant Training AI Tool

Helping instructors develop course-specific TA training resources

This project is developing TALens, a GenAI tool designed to help instructors create customized training resources for their teaching assistants (TAs). Drawing on existing course materials, including course outlines, assessments, policies, and training resources, the tool will ask instructors targeted questions and help generate resources such as training documents, meeting materials, checklists, and reference guides.

Progress to date

The team has completed qualitative coding of six instructor interviews and identified key TA training needs. These findings informed an initial set of questions addressing TA responsibilities, communication, accommodations, grading, course policies, and course knowledge. An early version of the agent has also been developed and is undergoing testing.

Learning to date

Instructor approaches to TA training vary considerably. While some provide comprehensive training, others take a more hands-off approach. Interviews also highlighted the considerable time involved in preparing training materials. Some instructors expressed interest in a tool that could improve efficiency while maintaining the quality of TA preparation.

What's next

The team will refine the prototype's questions and outputs, work with the development team to improve its content and prompting approach, and test how reliably course materials can be uploaded and incorporated. An initial version of the tool is expected to be shared for feedback in late November or December.

Project 4: Improving Student Communication Skills Through More Frequent and Formative Feedback

Developing context-aware AI feedback for student presentations

This project is developing an AI-supported feedback tool that gives engineering students more opportunities to practise and improve their oral communication before final presentations. The tool is being designed to review formative presentation recordings and provide feedback on clarity, concision, completeness, and the use of concrete evidence. Instructors will be able to provide course and assignment context so that feedback reflects specific learning outcomes, assessment criteria, and audiences.

Progress to date

The project has received research ethics approval, and the team has begun interviewing students and instructors to better understand user needs. A focused literature review on oral communication, formative feedback, and human-AI interaction has also informed the design. Development is underway in Azure, alongside work to establish appropriate technical, privacy, and security requirements.

Learning to date

Early proof-of-concept work suggests that presentation transcripts and slides can provide sufficient context for GenAI to generate formative feedback relatively quickly. However, effective feedback involves more than simply generating comments: it must respond to students' learning needs and encourage reflection and revision. The team has also identified privacy and responsible student-data handling as essential considerations from the outset.

What's next

The team will continue consulting instructors, complete its targeted literature review, and develop and test the initial Azure-based tool. An initial classroom pilot is targeted for January 2027, with findings informing further development, evaluation, and dissemination through April.

Project 5: LabReady: Scalable GenAI-Powered Readiness Platform for Experiential Learning

Helping students arrive better prepared for labs and hands-on learning

This project team is developing LabReady, a GenAI-supported platform designed to help students prepare for laboratory and other experiential learning activities. By identifying gaps in student understanding before a session, the tool aims to reduce time spent reviewing basic concepts and procedures, allowing more time for troubleshooting, discussion, and hands-on learning.

Progress to date

The team has developed a working prototype that transforms course materials and instructor knowledge into short, written-response pre-lab questions, model answers, and assessment rubrics. The tool can also grade student responses, draft individualized feedback, and generate class-wide readiness reports highlighting common misconceptions.

A new web interface guides instructors through three stages: preparation, quiz creation, and feedback. The platform integrates with LEARN, allowing quizzes to be uploaded and student feedback and scores to be returned directly to the course gradebook. Student identities are replaced with anonymous keys before information is sent to the AI model, and instructors can review and edit all prompts used by the tool.

In October, the team conducted its first optional, ungraded preparatory quiz in PHYS 121L and has been working through the research ethics process.

Learning to date

The team has been exploring how standalone web technologies, including HTML, JavaScript, and CSS, can be used to create interactive, customizable educational tools. Their experience has highlighted the value of human expertise and iterative refinement in turning AI-generated starting points into functional, user-friendly applications.

The team has also been examining research on AI-supported assessment. Existing studies suggest that large language models can assess short written responses with consistency comparable to a second human grader when supported by well-designed rubrics. However, results can vary, and models may be vulnerable to misleading instructions embedded in student submissions. LabReady addresses these risks by treating student responses strictly as content to be assessed and flagging unclear or potentially suspicious answers for instructor review.

An additional opportunity for the project is its focus on diagnostic assessment before experiential learning sessions, an area that has received less attention in existing research on AI-supported grading.

What's next

The team will complete the ethics process, expand testing in PHYS 121L, and pilot full lab sessions in additional courses. Further development will focus on refining prompts and templates and establishing ways to use instructor feedback and student results to improve the tool over time.

The pilot will examine student readiness, the diagnostic value of pre-lab quizzes, the effect on instructor and teaching assistant time during lab sessions, and the efficiency and scalability of the GenAI-supported workflow.