Personalizing Learning in Large Classes through Just-in-Time-Teaching
| Course Subject: | Engineering: Mechanics of Materials |
| Student Level: | Sophomore |
| Number of Students: | 100 total (two sections of 50 students) |
| Developed by: | Marguerite Matherne, Assistant Teaching Professor, Mechanical and Industrial Engineering, College of Engineering |
What Instructor Did
The instructor applied Just-in-Time-Teaching (JiTT), with generative AI as an assistant, to quickly personalize students’ learning experience in large classes. A few hours before each class, the instructor analyzed student-submitted pre-work. Based on aspects of the pre-work that they found to be especially confusing and/or interesting, the instructor adjusted the lesson plan.
Purpose
The goal of JiTT is to help students prepare for class while also creating a time-sensitive generative feedback loop between the instructor and the students about their understanding. The method helps the instructor determine where students are in their learning process in relation to the attainment of learning goals for the week and be immediately responsive. JiTT is an established teaching method, but it is extremely difficult to implement effectively in large classes because of the large volume of data the students create on a regular basis. The instructor’s use of generative AI as an assistant made it feasible for her to personalize the learning experience effectively.
Assessment
The instructor finds this method to be effective based on observations that students are asking more and better questions in class, taking the discussion to a deeper level. Student surveys and the examination of their work indicate that JiTT has increased the perceived value of course readings and has improved the attainment of specific learning goals. [see Marguerite’s essay linked under Related Materials].
The pre-class “warm-up” work is graded for completion. Students are not graded down for errors, but the warm-ups are factored into the homework calculation that is 10% of the final grade.
Faculty Reflections
I really wanted to use the JiTT method, but there was no way I could implement it each week at scale in a large class. Everyone was talking about AI, and so I decided to try it to see if it could help.
I am now able to ask students to respond in writing to open-ended “muddiest point” questions, which would otherwise generate far too much text to analyze each week. I can meet my students where they are, closing the loop on student understanding.
I would advise instructors who implement this strategy to engage in trial and error, don’t give up on it, and don’t be afraid of AI. It’s not coming for your job because you have expertise that is important, such as critical thinking. I determined that generative AI could not accurately identify patterns of error in student work, but it could provide a nice thematic summary of what they perceived to be confusing and what they found most interesting. It’s a great analysis tool, but it’s not really thinking – definitely not thinking creatively – it’s just good at analyzing words.
Step-by-Step Instructor Directions for Generating the Case Studies
| Step 1 | Develop warm-up exercises based on the readings and materials for the week. Assign these to students in Canvas. Set the exercises to become available to students at least 4 days before class and to be due 2 hours before class begins. They are graded for completion, not accuracy, and contribute along with weekly problem sets to a 10% homework grade. |
| Step 2 | Download student warm-up work 2 hours prior to class (deadline for student completion). Provide the following prompt to ChatGPT: “I asked students what they found most confusing or interesting about an assigned reading. Their responses are below. Summarize them according to what was interesting and what was confusing.” Paste the text from student responses, then run the query. Note: The first few weeks I analyzed the responses manually and compared them with the ChatGPT output. The results were very similar, and so from then on I only used the ChatGPT analysis. |
| Step 3 | Use ChatGPT output to inform lesson revisions. Begin the session by sharing what you learned from the warm-up responses so that students know their input is valued and informing the lessons. |
Related Materials
- Effectiveness-of-Just-In-Time-Teaching (Marguerite’s write-up of a study on JiTT that she conducted during her time as a 2023-24 CATLR Teaching and Learning Scholar)
Download PDF version of this assignment
Return to AI Gallery Home