On Turning the Tables and Disconnecting the Cables

“Señor, señor,
Let's overturn these tables
Disconnect these cables
This place don't make sense to me no more
Can you tell me what we're waiting for, señor?” Bob Dylan, Señor

This post discusses an interesting pedagogical strategy involving AI in literature classrooms centered around Brad's suggestion (see below) about using AI for real-time discussions of literature. Brad raised a powerful idea for AI use in response to my Pygmalion post. It would be useful to all of us to get perspectives from other readers about this idea.

One thing I really like about it is how students have to understand something about the assigned reading before coming to class to participate in this ‘teacher on the hot seat’ event—or do some motivated cold reading of a text in class.

We know from long-standing research that students at all levels sometimes have trouble getting motivated to read their homework assignments. This strategy creates a built-in incentive for students to complete their reading assignments because they need to understand the material to participate meaningfully in the AI-assisted discussion which turns the tables on the teacher.

Another is the contribution it could make to students regarding the importance of human expertise to bring out the value of AI. A paradox sits in the middle of the AI in learning problem. As a learning tool AI works best when used by subject matter experts. Yet by definition learners in classrooms are there because they are not yet experts. How do they come to engage this paradox for themselves?

There's also an interesting social dynamic at play. The AI potentially acts as a confidence buffer, allowing students to challenge their teacher's interpretations through the intermediary of the AI rather than directly, which might feel more comfortable for some students at least at first. It could motivate the desire to actually become an expert!

This strategy could impacts student learning about this expert paradox and give them insight into general LLM limitations. They get what will later become rare opportunities to work with the bot in the immediate presence of a subject matter expert.

Little user content expertise means user hyper vigilance in uncertainty. It’s hard to commit limited cognitive bandwidth to fact-checking without prior expertise in the topic. Strong expertise means 20/20 semantic vision enabling grounded critical processes. Users must be able to assess their level of expertise with each bot interchange and adjust their truth-o-meter accordingly.

In Brad’s scenario students borrow expertise from their teacher to witness this association unfolding in real time. It’s almost an artificial extension of a teacher think aloud.

Have you seen this approach or any resembling it implemented in practice? I think all of us would be interested in hearing about any examples of how students and teachers have navigated this three-way dialogue.

Brad’s comment on yesterday’s Pygmalion post:

“Do you envision students using AI in real time conversations to challenge and even counterargue teachers' ideas? I am thinking about literature classes. That dynamic could be quite lively, with everyone involved having to qualify ideas verbally.”

Thanks, Brad!