The challenge for EdTech

Can you take the teacher out of the geography classroom?

Every EdTech pitch makes more or less the same promise. Infinite patience. Perfect personalisation. Available at 11pm the night before a mock exam, when no teacher in their right mind would be. It’s a seductive proposition, and not an unreasonable one. So let’s take it seriously rather than dismiss it: what, exactly, would we be removing if we removed the teacher?

That’s a question worth being honest about from the start. The answer isn’t “everything.” A lot of what happens in a geography classroom could, in principle, be done by a well-designed digital tool. But some of it can’t, and the difference between the two turns out to be a useful diagnostic for what good geography teaching actually consists of.

The Future 1 trap

Most digital geography tools, however slick the interface, end up doing the same thing: delivering content. Capitals, case studies, key terms, a quiz at the end to check it’s gone in. Michael Young and Johan Muller’s curriculum futures model gives this a name: Future 1, where knowledge is treated as fixed and uncontested, and the pupil’s job is to receive and reproduce it. Some apps reach for Future 2, with engaging activities and project-based exploration draped over a geography theme but without much disciplinary substance underneath. Almost none get anywhere near Future 3, where disciplinary knowledge and the pedagogy that makes it usable are designed together.

This isn’t really an argument against digital tools. It’s an observation about which version of the job is easiest to build, and therefore which version gets built. A simple test for any geography app or worksheet, digital or not: does it reward getting through the content, or does it reward the quality of a learner’s explanation? Most things, if we’re honest, reward the former. That’s not a teacher problem. But there are several things good teachers do that genuinely resist this kind of automation, and they’re worth looking at in turn.

1. Reading and steering dialogue in the moment

Neil Mercer’s distinction between disputational, cumulative and exploratory talk is well known to most geography teachers, even if not always by name. Disputational talk is assertion without resolution. Cumulative talk is uncritical agreement, pleasant but uncritical. Exploratory talk is where the learning actually happens: ideas offered up for challenge, reasoning made visible, disagreement treated as useful rather than something to smooth over.

Getting a class into exploratory talk isn’t a script you can run. It’s a judgement made in real time: when to let a half-right answer sit unresolved a moment longer because the discomfort is doing work, when to step in because the group has drifted into cumulative agreement and stopped actually thinking, when a counter-perspective needs introducing because nobody in the room is going to offer one themselves. Rupert Wegerif calls this holding a “dialogic space,” an environment where competing ideas stay in tension rather than getting resolved too quickly. That’s a live read of a specific room of specific pupils on a specific day. An app can prompt a learner to respond, or simulate a counter-argument. What it struggles to do is notice that this class, today, has settled into comfortable agreement and needs unsettling.

2. Withdrawing scaffolding as a judgement, not a schedule

Scaffolding is supposed to be temporary. Sentence stems, graphic organisers, guiding questions: useful while a pupil is building capability, actively unhelpful once they don’t need them any more. The expertise reversal effect is the evidence behind this, showing that the same scaffold that helps a novice can measurably hinder someone who’s moved past needing it.

The trouble is that “moved past needing it” isn’t a fixed point you can schedule in advance. A good teacher is constantly, almost unconsciously, deciding whether to hand a pupil the sentence stem or hold it back and see what they do without it, based on what they’ve just watched that pupil do, not on a generic model of where a typical Year 9 should be in week six of a unit. A digital tool has to build its scaffolding progression in advance, against a model of the learner. A teacher is reacting to the learner actually in front of them. Those are different problems, and the second one is much harder to specify in software.

3. Surfacing and restructuring misconceptions

This is maybe the clearest example. Pupils arrive in a geography classroom carrying scale errors, single-cause explanations, and media-formed stereotypes about places. Delivering correct content does not reliably fix this, since a pupil can reproduce the correct terminology while still holding the underlying misconception underneath it, untouched.

What actually shifts it is dialogue that makes the faulty reasoning visible and then applies the right pressure to it: does this explanation still work at a different scale? What makes you think that? Can anyone challenge that idea? Knowing which question to ask requires having heard the specific wrong idea a specific pupil just expressed, and recognising what’s actually wrong with it: not “this answer is incorrect” but “this is a scale error” or “this is treating a global pattern as if it has one local cause.” That’s diagnosis in real time, not a presentation of correct information. An app can present the correct version. It’s much harder for it to catch and name the precise shape of the misunderstanding it’s competing with.

What I’m leaving out

Three examples make for a tidy structure, but a reader who’s spent time in a classroom will reasonably object that this only scratches the surface. There’s the pastoral judgement of knowing why a pupil has gone quiet today and isn’t usually quiet. There’s the curriculum-level decision-making behind why this unit comes before that one, built from years of watching what pupils arrive able to do. There’s behaviour management woven invisibly into every one of the three examples above, since none of that dialogue happens at all in a room that hasn’t been made safe enough for pupils to risk being wrong out loud. I’ve picked three things that connect cleanly to the cognitive science and dialogic teaching literature, because they’re the ones easiest to argue precisely. That’s a deliberate narrowing, not a claim that they’re the whole job.

The honest counter-case

None of this means digital tools are useless, and it would be a cheap and slightly dishonest conclusion to land there. Cognitive science gives us tools that teachers are bad at delivering reliably even when they know the theory perfectly well: spaced retrieval practice across a term, interleaving topics rather than blocking them, giving every pupil in a class of thirty an actual individual turn at retrieval rather than the same five hands going up. A teacher’s attention is one resource, split thirty ways, for one hour, once a week. A well-built app doesn’t have that constraint. There’s a real and useful job to be done here. It’s just not the same job as the three above.

So what should an app actually be trying to do?

Alaric Maude’s typology of powerful geographic knowledge gives a useful way to sort this out. He describes five types of “power” that geographic knowledge can give a learner: new ways of seeing the world, the ability to explain natural and social processes, knowledge of distant places and processes, power over how knowledge itself is produced and contested, and the capacity to join debates of real significance. The first three are well within reach of a thoughtfully designed app: that’s largely a content and explanation problem, and content and explanation are things software is good at. The last two, power over your own knowledge and the capacity to argue your corner in a genuine debate, are exactly the capabilities built from the three things above: live dialogue, responsive scaffolding, real-time diagnosis of faulty reasoning. They’re also, not coincidentally, the capabilities AI is most likely to bypass entirely if app designers aren’t deliberate about building for them.

So the more useful question isn’t “can an app replace a teacher.” It’s “which of these five types of power is this app actually trying to build,” and an honest answer about why the last two are so much harder to reach than the first three.

Geography is a good test case for all this precisely because it’s a discipline built on contestation and uncertainty rather than settled fact: climate change, migration, development, who gets to decide what counts as a resource. If a digital tool can’t hold dialogic space, read live reasoning, and respond to the specific misconception in front of it for a subject like this, that tells us something about the limits of the current approach more broadly, not just about geography apps.

I don’t think the answer is to give up on the attempt to support geography learning with digital tools. But it does mean being honest about which parts of teaching are a content problem, eminently solvable in software, and which parts are a judgement problem, made fresh in the room, every single lesson.

---

References

Kalyuga, S., Ayres, P., Chandler, P. & Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23-31.

Maude, A. (2016). What might powerful geographical knowledge look like? Geography, 101(2), 70-76.

Mercer, N. & Littleton, K. (2007). Dialogue and the Development of Children’s Thinking. Routledge.

Wegerif, R. (2013). Dialogic: Education for the Internet Age. Routledge.

Young, M. & Muller, J. (2010). Three educational scenarios for the future. European Journal of Education, 45(1), 11-27.