You have /5 articles left.
Sign up for a free account or log in.
Faculty at the University of Delaware can personalize study materials based on their own course content using generative AI technology developed by the institution.
AJ_Watt/E+/Getty Images
Many students turn to online resources to help them prepare for exams, but the University of Delaware is making it easier for learners to utilize relevant digital tools to supplement course lectures.
Starting this fall, UD is piloting an initiative that will transform recorded lectures into study guides, flash cards and practice quizzes. The work, developed in-house at the university, leads with ethical principles and prioritizes faculty content ownership to protect all participants, as well.
The background: The University of Delaware has used its own software to record professor lectures for over a decade, says Jevonia Harris, educational software engineer and leader of Academic Technology Systems (ATS) at the university. Fifteen years ago, faculty members were slow to adopt the tech, but it’s pretty popular now.
In 2022, when ChatGPT launched, Harris was considering ways that students and faculty members have used those lecture recordings previously for studying and learning, and how generative artificial intelligence could improve those processes.
Some professors have taught multiple sections of the same course for years, often every semester, providing a wealth of repetitive data, “which is great for AI,” Harris explains.
Harris hypothesized that she could use recorded lectures to train AI and transform lectures into study materials and outlines.
How it works: The artificial intelligence builds a knowledge graph, connecting information with different relationships. Staff feed the AI any and all transcripts from a single course, which could be as many as a few hundred, and then ask the tech to identify topics and subtopics.
Once their lectures are processed, faculty members review the exported information for accuracy. Often, they’ll flag information as not as relevant as indicated by AI, outdated or not taught in the class anymore, Harris says.
After identifying the top-line information, the data is then stored in the LMS as is or modified into learning tools such as flash cards or practice quizzes. From there, students can access the information on demand to supplement the recorded lecture for each module.
UD uses Amazon Web Services Bedrock to encrypt data and ensure faculty privacy.
The development team includes two software engineers, some instructional designers, a user-interface developer and a Ph.D. student who used to work as a software developer. Having instructional designers has guided some of the digital tool processes and improved pedagogy, Harris says.
Ethics in focus: Before Harris and her team ever got to working with AI and professors’ content, they outlined ethical use and the principles that would guide their work.
“Technology always has the very shiny benefits, but also, what are the potential drawbacks?” Harris explains.
ATS consulted with campus librarians, a group of instructional designers and philosophy professors to glean their insights and consider all perspectives.
“This is before we’ve written any code,” Harris says. “We’ve made sure to be pretty transparent about, ‘This is what we want to do; if there’s any things that we’re not thinking of, please let us know.’”
UD was also uniquely positioned for this work because of the culture of trust ATS has established and the relationships staff have built with faculty. The process is entirely opt-in, meaning UD does not scrub transcripts without faculty knowledge and approval and does not work with their content otherwise.
What’s next: One faculty member is piloting this project in two courses, Psychology 100 and 300. The introductory course has basic flash cards for rote memorization, and the advanced course has more complex questions, requiring critical thinking and problem-solving.
A future application for this work may be faculty review of course content.
“If you have been teaching for a long time, maybe some of your information is older; you may not want that to be something that’s more prevalent in these learning objectives,” Harris says. The knowledge graph provides a time-stamped database of all discussion points, helping professors identify where their material could stand to be updated or reimagined.
Students and faculty in the educational technology program are also creating research studies to understand the impact and outcomes of the pilot.
“We’re inviting collaboration—we’re very open about this—and if we can become a standard of intentional and ethical way of using AI, then I would consider this a success,” Harris says.
How is your university using generative AI for teaching and learning? Tell us here.
