GenAI Lecture Summary Evaluation

In this assignment, students used a generative AI tool before each guest lecture to learn key concepts, predict the lecture topics, or generate study materials based on the speaker's biography and session description. After the lecture, they reflected on how accurate and helpful the AI's response was.

Authors: Isabelle Athena Qian, Undergraduate Student & Chrystal Chern, Program Strategist, Research Mentorship and Student Impact, Berkeley Discovery, and Instructor for Brilliance of Berkeley 
(Supervisor: Oliver O'Reilly)

Course Number & Title: LS110: Brilliance of Berkeley

This work is licensed under CC BY-NC-SA 4.0

Details
Assignment Title
 
GenAI Lecture Summary Evaluation
Delivery FormatIn-person
Learning Objectives
  • Practice crafting effective and context-rich prompts for generative AI models
  • Evaluate the accuracy and limitations of AI-generated predictions and summaries
  • Increase active engagement and preparedness for lectures
Brief Summary of AssignmentBefore each guest lecture, students select a GenAI model (e.g., Gemini, ChatGPT, Claude) and prompt it with the professor’s biography and lecture session abstract, asking it to (1) walk them through concepts they should know beforehand and (2) predict the top five topics the lecture will cover. Students record their model, prompt, and the AI’s output. After attending the lecture, students write a reflection critiquing the AI’s response. Across the semester, the format varied: pure text, an AI-generated infographic, and a co-created lecture notes outline that students annotated live during class. The exercise replaced a traditional lecture-reflection assignment and was designed to sharpen students’ judgment about when AI helps versus hinders learning.
Innovative Teaching Reflection:The assignment gives students structured, low-stakes practice using GenAI as a study companion rather than a shortcut, while explicitly training them to identify AI’s limitations. Group meetings also enable students to engage in conversation and learn from others’ AI usage. The nature of the assignment in this class is unique, as it brings together students from majors and year groups across Berkeley, and the GenAI prompts have been applied to lecture topics across a wide range of academic disciplines. The result has shown that students’ attitudes towards the quality of GenAI output for different lecture topics have been fairly uniform.
Assignment Length/Time EstimateThis is a weekly assignment, completed once per class session across the 14-week course (implemented as a pilot in Weeks 5, 7, 9, and 11). Each cycle takes roughly 20–30 minutes outside class (pre-lecture AI prompting) plus a short post-lecture reflection (10–15 minutes), for about 30–45 minutes of student time per lecture; no additional in-class time is required beyond the lecture itself, except in the notes-outline variant, where students actively annotate during the live session. Students are also asked to meet virtually three times throughout the semester for instructors to collect more detailed feedback on the student learning experience.
Assignment Details