Teaching Dialogic Intelligence with AI

Recent studies split almost 50-50. Some warn that chatbots dull critical faculties through easy copy-paste answers. Others show the opposite: used dialogically, AI nudges students into deeper reflection. The difference is not the tool but the frame. This post outlines a two-loop educational strategy for teaching dialogic intelligence—the art of thinking with others (human or machine).

Step 1: Reframing AI Use

In the Thinking Together approach, we begin by helping students notice how the way they talk affects group thinking. This might involve watching short clips of classroom dialogue and asking: What’s going on? What helps learning? What gets in the way?

This reveals the hidden ground rules shaping their talk—some dominate, some avoid speaking, some try to win rather than explore. We work together to co-construct better ground rules: listen with respect, ask questions, give reasons, explore alternatives, share knowledge.

We can apply the same approach to the use of AI. Instead of just learning how to prompt better, we can begin by asking: How are we using AI? What helps learning—and what gets in the way?

Less helpful uses:

• Copy-pasting answers without thinking

• Using AI to confirm existing beliefs

• Treating it like Google

• Accepting the first answer uncritically

More helpful uses:

• Asking follow-up questions to check and probe

• Comparing multiple perspectives

• Exploring what others have said

• Asking AI to challenge assumptions

This isn’t just about getting better practical tips—it’s also an opportunity for reframing. Asking these questions students are given the opportunity to rethink what they imagine that their education is for.

Step 2: Dialogue with Cultural Voices

Across cultures, mastery often begins by entering into dialogue with an ancestral guide — the spirit who “knows” how to carve a canoe, weave a basket, or navigate the stars. Generative AI lets modern learners do something similar. With the right training data, an AI can personify any discipline—quantum physics, medieval history, Python coding, even rose cultivation—becoming a responsive partner rather than a static textbook. It not only shares the wisdom accumulated so far; it plugs students into the live, ever-evolving conversation of each field. Using AI in this way, learning becomes participatory. Students engage directly with traditions, asking questions, offering ideas, and positioning themselves within a living conversation.

My previous post, argued that we should reorient education towards the goal of teaching dialogic intelligence. This sounds a bit like teaching thinking. A common criticism of the aspiration to teach thinking is that thinking requires knowledge—you can’t think well about maths or history unless you know the material. That’s very true. But shifting from a monologic to a dialogic frame helps us see that thinking and knowing are not separate things but two fruits of dialogue. ‘Knowledge’ consists of answers to the questions we ask and ‘thinking’ is how we ask those questions and how we listen to the answers.

Bakhtin pointed out that dialogue always involves a third voice—a "super-addressee"—a witness who hears and understands everything, even when your immediate partner doesn’t. This generation of a witness position is a necessary function of dialogue. We often sense this when we speak: we anticipate being understood (or misunderstood) by an imagined listener.

Bakhtin also argued that all language is "heteroglossal": full of inherited meanings from traditions, genres, and disciplines. We’re never just speaking as ourselves—we speak through cultural voices.

In some conversations, the superaddressee can, as Bakhtin suggested, be a kind of God-like presence—a higher authority capable of hearing both sides of a case with perfect justice. But in disciplinary contexts, such as a classroom discussion in mathematics, the superaddressee takes the form of the voice of the discipline itself. When students make claims, they are accountable not just to their peers or teacher, but to the long-term cultural dialogue of mathematics and its evolving standards of what counts as good reasoning.

Step 3: Dialogue with the Infinite Other

But learning to think through dialogue with defined cultural voices isn't enough. Real thinking requires knowing how to ask questions and listen to answers even when there is no tradition to call upon. Real world challenges like responding to global warming or maybe just creating something constructive for young people to do in your neighbourhood demand transdisciplinary thinking, thinking outside of any context where confidence and creativity are needed to shape the future.

Bakhtin’s idea of the super-addressee is not really a specific voice but in context it will always take on form. The image of science for example that we find ourselves in dialogue with in a lab might for some be a bit masculine, emotionally cold and linked to the idea that there are ‘laws of nature’. But engaging in dialogue with this image generates a further superaddressee type personified voice. Our dialogue with that new voice will generate still another superaddressee and so on in an infinite regress. Real thinking, thinking that questions all assumptions and listens deeply, is not a dialogue with this or that cultural voice but a dialogue with the unknown horizon – and when we are building things this becomes a dialogue with the unknown future. Although it might seem crazy to ask questions of the unknown horizon and expect it to answer back, experience shows that it often does.

AI cannot personify the Infinite Other. But it can gesture toward it—prompting learners to challenge assumptions, step outside their frames, and embrace uncertainty. Used this way, AI helps shift students beyond defending positions or conforming to consensus. It helps them hold the tension between viewpoints—where new insights can emerge.

Step 4: Becoming Double, Becoming Dialogue

Becoming dialogically intelligent involves a shift in identity. At first, students hear only their own voice, not the dialogue’s response. But learning to engage in dialogue means becoming double – you speak but as you speak you listen to yourself speak and in that listening you incarnate the voice of the whole dialogue that you are participating in.

Ultimately, dialogic intelligence doesn’t reside in a person, a tradition, or even some image one might have of the universe. It is the arc of tension between an inside perspective and the outside horizon. Sometimes it seems like you hear your voice asking questions and finding answering insights: sometimes it might be the other way around with the universe seeming to ask you questions to which you must somehow respond.

No Final Words

Bakhtin reminds us: there are no final words—the dialogue always continues. This post offers a few modest and provisional suggestions for teaching dialogic intelligence with AI through a two-loop model:

• First loop: Teach students to think by opening, widening and deepening, dialogic space—AI can open a space by prompting, can widen by sharing multiple voices and perspectives on the topic and deepen by exposing and challenging the assumptions that are being made behind every question.

• Second loop: Induct students into larger cultural dialogues. AI can authentically personify traditions, inviting learners in and feeding them with what they need to know to be able to actively participate.

Skeptics insist AI can’t be a true dialogue partner because it lacks empathy. Yet its very other-ness lets it serve as education’s outside voice—embodying everything ever said in a field and inviting students into that living conversation. From there it can prod them to leap beyond inherited ideas and co-create tomorrow’s knowledge. That is the promise of double-dialogic education.