The Conversation We're Not Having About AI in Higher Education

I've been teaching long enough to recognize when something is genuinely transformative rather than just the latest shiny object in edtech. When I first encountered generative AI in my classroom through student submissions, my reaction was somewhere between fascination and mild panic.

But here's what's been nagging at me: we're so caught up in debates about academic integrity that we're missing something fundamental about how learning actually happens. As someone who's championed dialogic approaches to teaching for years, I keep wondering whether AI might actually open up new spaces for rich, generative dialogue or whether it's going to shut those spaces down entirely.

What Gets Lost in Translation

Dialogic teaching isn't just about discussion. It's about creating conditions where thinking happens through conversation, where ideas are tested, reshaped, and built upon collectively. It's messy. It requires students to articulate half-formed thoughts, to stumble, to hear themselves think out loud.

The best moments in my seminars happen when a student says, "Wait, I'm not sure this makes sense, but..." and then the whole room leans in. That uncertainty, that willingness to think publicly, is where deep learning lives.

So when AI entered our classrooms, my first question wasn't really about plagiarism. It was: what happens to this dialogic space?

The Unexpected Bits

I've been watching closely, and it's more complicated than I expected. A recent study from the University of Liverpool looked at architecture students using AI image generators. Students who crafted more conceptually rich prompts produced more compelling visual outputs. Nothing particularly surprising there.

But what caught my attention was this: students reported high levels of creative support, particularly in exploration. They weren't passively consuming AI outputs. They were iterating through language to refine their design intentions. It looked suspiciously like dialogue.

This challenged my initial assumption that AI would short-circuit learning. Instead, students were learning a new literacy that required them to articulate their thinking with both precision and creativity.

Where I'm Worried

The Liverpool study found that collaboration and immersion scores were notably lower than other dimensions. That matters. Learning isn't just about producing good outputs. It's about the struggle, the shared frustration of working through difficult ideas with other people.

I'm also worried about student voice. In dialogic learning, we talk about students developing their unique perspective. But when you're constantly negotiating with an AI trained on millions of texts, whose voice is actually emerging? There's a real risk that students start sounding like slightly customized versions of a language model.

And then there's intellectual dependency. If students outsource the hard cognitive work of drafting to AI, are they actually developing as thinkers? Writing isn't just about producing text. It's about wrestling with ideas until they make sense.

What I'm Trying Instead

I've started experimenting. Instead of banning AI, I ask students to use it as a thinking partner, then bring that conversation into our classroom dialogue. "Show me what the AI suggested, then tell me where it's wrong or incomplete."

The results have been surprising. Students who normally stay quiet seem more confident when they've rehearsed ideas with AI first. Others use AI to generate provocative counterarguments that push discussion in unexpected directions. One student told me it was like "having a really patient but occasionally clueless study partner who helps you figure out what you actually think."

That last bit is crucial: what you actually think. The AI doesn't replace thinking. It becomes a tool for externalizing and examining it.

Putting Theory into Practice: The LEAP Project

These aren't just theoretical concerns for me. I'm currently leading LEAP (Learning Enhancement through Adaptive Pathways), a two-year Knowledge Transfer Partnership funded by Innovate UK. We're developing an AI platform specifically designed to support teachers and enhance student learning through dialogue-driven thinking.

The goal isn't to replace human teaching but to amplify educator expertise. LEAP uses adaptive pathways to create spaces where AI acts as a scaffold for deeper thinking, not a shortcut around it. It's about designing AI tools that preserve and enhance the dialogic spaces I've been talking about, rather than collapsing them.

If you're interested in how we're approaching this challenge practically, you can learn more about the LEAP project here.

What Needs to Change

Most universities have responded to AI with policies about when it can be used, focusing on detection and prevention. That's treating AI as a problem to manage rather than a shift that needs thoughtful integration.

We need to redesign assessments around process, not just product. If students can document their dialogue with AI, showing how they tested ideas and rejected suggestions, that becomes evidence of learning.

We should teach prompt engineering as a rhetorical practice. How do you articulate a question well? How do you push back against an answer that seems plausible but feels wrong? These are exactly the skills we've always wanted students to develop.

And critically, we need to preserve spaces for human-only dialogue. Not because AI is bad, but because certain kinds of learning, particularly those involving emotion and ethics, require human presence and vulnerability.

What Keeps Me Up

Can students develop genuine intellectual autonomy if they're always in dialogue with a tool that seems to know everything? What happens to productive confusion if AI is always there to resolve ambiguity?

And perhaps most importantly: are we asking students to engage in dialogue with AI, or are we asking AI to simulate the dialogue that should be happening between humans?

No Easy Answers

Those architecture students had 45 minutes to design a residential space using only text prompts. The researchers found that richer, more varied language in prompts correlated with better outcomes. But the real insight was about what happened between intention and output, in that iterative dance of articulation and revision.

That's where learning happens. That's what we need to protect as we navigate this shift.

I'm not arguing for or against AI in education. That debate feels beside the point. AI is here. The question is whether we can shape that integration in ways that deepen rather than diminish learning.

My hope is that we can. But it requires us to be thoughtful, experimental, and willing to admit when we're unsure. It requires treating AI not as an answer but as a new participant in an ongoing conversation about what education is for.

What's your experience been? I'm genuinely curious how other educators are navigating this.