Earlier this year, two lecturers at Vilnius University in Lithuania deployed an unusual teaching assistant: an AI chatbot version of themselves.
Instructors Paul Jurcis and Goda Strykaite-Latusinskaya created an AI chatbot trained exclusively on academic publications, PowerPoint slides, and other materials they've developed over the years, calling the two the “AI Knowledge Twins,” one named Paul AI and the other Goda AI.
They told students that if they had any questions during class or on their homework, they should ask the bot first before asking a human instructor. Rather than discouraging students from asking questions, the idea was to encourage them to try out their chatbot double.
“We brought them on as assistants — research assistants who help people use our knowledge in new and unique ways,” Jurcys says.
Artificial intelligence experts have been experimenting with the idea of creating chatbots that can fulfill this support role in the classroom for years, and the rise of ChatGPT and other generative AI tools has sparked a new wave of experimenting with robot TAs.
“From a faculty perspective, this is very appealing, especially for faculty who are overwhelmed with classes and need teaching assistants, so they can focus on research instead of teaching,” says Mark Watkins, a lecturer in writing and rhetoric at the University of Mississippi and director of the university's AI Summer Institute for Writing Teachers.
But just because Watkins thinks some faculty might like it doesn't mean he thinks it's a good idea.
“That's exactly why it's so dangerous, because it essentially puts the human relationships that we're trying to build with students and between teachers and students into the hands of an algorithm,” he said.
On this week's EdSurge podcast, we hear from the professors about how the experiment worked and how it changed — and sometimes disrupted — classroom discussions. Maria Ignacia, a student in the class, also shares her thoughts on what it was like to have a chatbot TA.
And we hear Jurcys ask the chatbot a question, and the bot acknowledges that he phrases it a little differently than he does.
You can listen to this episode on Spotify, Apple Podcasts, or in the player on this page.
