Blog

Conversational Learning with AI

By Michael Faber // December 27, 2024

I’m going to out myself a little bit here...

 

I like to talk to ChatGPT in the car.

I mean, usually I like listening to music, but sometimes, I like talking to a robot, I guess.

 

Hear me out.

Every day, I drive down the beautiful Highway 751 to get to campus. This picturesque winding road is a fantastic opportunity to let the mind wander. Maybe I’m thinking about what I am going to be working on that day. Maybe I’m thinking about the latest tech device that we should play with. Maybe I’m wondering what Sauron’s backstory is and how Middle Earth came to be overrun by darkness. I can meander through the mind’s winding road but my own knowledge can only take me so far. If I really want to know what the best glue is for re-attaching a car door gasket that has come loose, I probably need to look that up later (for the sake of myself and everyone else on the actual road).

But lately I’ve found that adding an AI (ChatGPT's Advanced Voice Mode) to the passenger seat turns that wandering into something better - a process of learning and creative discovery.  While it’s not perfect, and we do talk on top of each other from time to time, it's very close to natural back-and-forth conversation. However, instead of just me and my own thoughts, I have access to the entire pre-trained knowledge of the world’s smartest AI with me. So, as I’m driving through the morning fog, I can have it tell me that, no, superglue is not a good choice to fix that gasket, you’re going to want to get the 3M stuff that is made for that task. Or that Sauron was once a powerful Maia spirit named Mairon, and served the godlike smith Aulë before being seduced by the dark lord Morgoth and turning to evil. 

But most usefully, I find it to be a learning and brainstorming tool.

The idea of using dialog as a pedagogical approach is hardly new. Plato and Socrates were demonstrating the power of conversation as a way to stimulate critical thinking, challenge assumptions, and co-construct knowledge over 2000 years ago. And modern researchers like Robin Alexander are even now showing how this paradigm can be effective in education and learning.

So what happens when we replace the Socratic dialog with that of an AI, who has access to the entire wealth of human knowledge. The AI won’t have the life experiences or desire for truth that Socrates has, or the innate empathy or teaching experience that a great professor has, but interesting things can start to happen. I ask better questions. I think more deeply about the responses. I am forced to articulate my thoughts and ideas better, so that I can get better responses. Often, I am dissatisfied with the answers, and am forced to provide deeper constraints or wrinkles to the questions. Or the commentary provided sparks my curiosity to explore a lateral idea, even if the AI’s ideas are common or uninspiring. The goal here is not that an AI can give me everything, it’s that we can work together in a process of discovery and learning. I follow threads, provide signposts, ask for examples and pastiche, and through this improvisational dialog, arrive at interesting or thought-provoking conclusions.  I also get better at doing this over time, which produces faster or more aligned results.

Researchers like Robin Alexander call the human version of this kind of learning dialogic teaching—a process where ideas are built collaboratively, tested, and refined. It’s not about memorizing facts but thinking through ideas in real time, through conversation. What’s surprising is that this same process seems to work even when one partner in the dialogue is, well, a machine. The knowledge that the AI is providing is good, but perhaps more importantly is the creativity and insight that I unlock in the process.

Ethan Mollick calls this “Co-Intelligence,” where human and AI are greater than the sum of their parts. The AI brings it’s strengths, like repetition, knowledge, synthesis, and re-contextualization, while I bring creativity, curiosity, constraints, and most importantly, qualitative judgement of the AI. And when it is all working together, this process feels surprisingly… human.

While this works great with ChatGPT "out of the box", I also built a custom GPT with a system prompt that is designed to be a stimulating dialogic instructor. You can try it out here: https://chatgpt.com/g/g-676ef109b78c81918d46c0df3ee279ed-dialogic-instructor

There’s lots of ink and discussion these days about how AI will disrupt education, change the university, and maybe even change humanity, and those conversations are valid, especially in light of the hyperbolic pace of AI progression. But at least for the moment, AI is in a pretty great sweet spot, where conversation and co-intelligence can unlock a new, flexible form of learning. Whether it’s solving a specific problem, satisfying a curiosity, or asking Sauron what he would make with a Raspberry Pi (mini nazgul drones?), these conversations are more than just fun - they are transformative.

So next time you're on a drive, strike up a conversation with ChatGPT about... anything... and see where the road takes you.