Motorways with no speed limits and car-free town centres

A manifesto for technology in schools

Germany has motorways with no speed limits. It also has medieval town centres where cars are banned.

So is Germany a pro-car or an anti-car society? Do people love cars or hate them? Is Germany’s attitude to cars libertarian or authoritarian?

Obviously, the context matters. In some contexts, cars are the solution to a problem. In other cases, cars are the problem.

In my 2020 book Teachers vs Tech, I made a similar argument about education technology. I thought that ed tech was indispensable in some places, but a terrible idea in others.

Since then, we’ve had an explosion of interest in how AI can help students learn. You have boosters arguing that kids can learn anything they need to on screen, but you also have sceptics saying that AI taints everything it touches. I disagree with both these positions.

In this post, I will update my 2020 argument for why we need both more and less technology, and also put forward three approaches that embody my car-free town centre / limit-free motorway concept.¹

The limit-free motorway: why we should persist with ed tech

For me, there is one fundamental reason why we should persist with ed tech: scale.

Imagine the best possible human-led classroom lesson, the one where every student is engaged and enthused and learning at every point. Whatever your vision looks like, one thing is true: it will be incredibly hard to scale up. There are massive variations in quality between classrooms. Even within schools, there are variations in quality between classrooms, to the extent that OECD data shows that variations within schools are greater than variations between them. This suggests that more learning can happen in the best classroom of a struggling school than in the weakest classroom of a high-performing one.

On its own, money does not solve this problem. Expensive private schools have not solved the variation problem. If you reduce class sizes you often end up exacerbating the problem, because you then have to recruit and train thousands more teachers who are unlikely to immediately (or, perhaps, ever) be as good as the existing teachers in the system.

It is this problem which is, for me, the biggest and most legitimate justification for education technology. Technology can be a great equaliser. In its ability to deliver consistent quality at scale, it is perhaps the most democratic force in existence. The printing press made books available for the masses. Recorded music allows everyone to listen to world-class performances. The internet makes information freely accessible. These technologies have already helped democratise access to education, both inside and outside of traditional educational institutions.

But given that the variation problem remains, these technologies have not done enough. For many, it can seem as though the next step is a fully personalised screen-based education that cuts out the human teacher. Instead of learning in a traditional classroom with human teachers and paper-based materials, perhaps in the future students will learn using a combination of personalised devices, AI chatbots, eye-tracking technology and adaptive algorithms.

This is where I depart from the pro-technology line. I think we have to find ways of using technology to get scale, but for the majority of school education I don’t think that will involve kids sitting at screens.

The car-free town centre: learning is social & embodied

Right now, education still does mostly take place in human-scale, human-led physical environments. No-one has yet found a way of successfully educating students in large numbers with a human-light and tech-heavy alternative that involves students learning from screens most of the time.

We also recently had the natural experiment of Covid, the type of external shock which often does accelerate systemic technological change. And in many areas, that is exactly what happened: Covid massively accelerated remote working, telemedicine, online shopping and cashless payments. Six years on, these trends are all permanently higher than they were before Covid. But in education, this did not happen. Yes, there were plenty of attempts to learn online and plenty of interesting innovations. But most parents were desperate to get kids back in school and most of the data (ours included) showed that kids did not learn much at home. So for all of the pro-tech arguments about scale, what Covid showed is that even the inconsistent in-person school is still better than learning remotely on screen.

Why is this? In 2020, I argued that there were subtle ways in which learning was social and physically embodied, and that these mechanisms were particularly important for younger students. Post-Covid, I feel even more confident that this is the case.

What do I mean by “learning is social” and “learning is embodied”?

For very young children (ie toddlers and infants), there is a lot of evidence that learning relies on eye contact, shared attention, imitation, emotional connection, and interactive back-and-forth conversation. Researchers have been trying for decades to get young children to develop first or second language skills by video, and they have found there is a “video deficit” which makes it very hard. Children pay more attention to words spoken by a physical human being in their presence than to the same human saying the same words on a screen. They copy human adults. They care about what their teachers and their peers think is important, and are more likely to pay attention to that. Even when they are watching a video or listening to a recording on their own, often their motivation or interest derives from knowing that other humans are interested in it too.

These mechanisms don’t disappear for older students. “Social contingency” is the term used in the research literature to describe interactions that are dynamic, where your response or behaviour changes the other person’s. Research on instructional design suggests that this kind of responsive back-and-forth interaction remains important for learning well beyond early childhood. One of the reasons direct instruction programmes are so successful is that they are highly socially contingent, with multiple opportunities for the student response to change the teacher’s behaviour. The best screen-based programmes attempt to recreate this dynamism too, but doing so reliably and at scale is difficult. Intelligent tutoring systems tend to use banks of pre-written statements and questions, which aren’t as dynamic as humans. Even the tiny delays on video calls can subtly disrupt turn-taking.

In terms of learning being physically embodied, I’ve frequently summarised the research showing the difference between learning on screen and learning on paper. We skim and scan more when we read on screen, we think differently when we take notes on laptops, and there are powerful “mode effects” that lead to students doing worse when they take tests on-screen than on paper.

Note here that the anti-tech aspect of my argument is different from a lot of other anti-tech arguments. For example, I hear a lot of tech sceptics say that learning on screen may well be efficient, but efficiency isn’t everything. That is not my argument! I care a lot about efficiency! I think efficient learning is exceptionally important and too often it gets downgraded and neglected. When I hear someone say “efficiency isn’t everything”, I get nervous because I think that is too often used to justify some quite vague and nebulous classroom activities that add very little social or academic value.

My argument is that screens are currently not an efficient way of learning for under-11s, and that they may never be, in the same way that cars are not an efficient way of getting around a medieval town centre. I think the most efficient way of educating children under 11 involves whole-class explicit instruction, paper-based materials and teacher read-alouds.

Isn’t this all quite waffly?

How important are these effects? Are they really crucial, or are they just nice to have? I am sure there were travel agents in the 1990s who said “no one will want to book a holiday on a screen! They will want to come into the social and embodied space of the warm and inviting travel agent on the high street and book it with a real human being instead!” Now maybe there was some truth to that. Probably a lot of customers would have even said something like that if you’d asked them. But ultimately, whilst they might have valued the high street travel agent, they didn’t value it that much. They preferred getting 50 quid off their next package holiday more than they preferred booking it in a shop with a human agent.

Am I at risk of making the same mistake? I am talking up the social and the embodied aspects of learning, and there is a lot of evidence that they are real. But technology has benefits that humans can’t provide, and it is always possible that it is worth losing the social and embodied in return for the benefits of scale and consistency.

Let’s think this through using two different scenarios.

Let’s imagine you are aged 25 and you need to take a course on a specific type of database in order to get a promotion at work. Option A is to do the course in person. The next course starts in six months, costs £5000 and involves a 2-hour round commute. Option B is to do the course online. The next course starts when you want it to, and costs £500. Given this choice, I would pick Option B – the online screen-based no-human option.

Here is another scenario. You have a child aged 6 who is learning to read. Option A is to attend the median primary school in England which will teach them to read in a class of about 25 using one of the standard phonics programmes in England – paper-based, led by a human teacher who has been trained in the use of that particular programme. Option B is to attend a school where the student learns to read by sitting at a screen with headphones on following an app. A human teacher is around to supervise and motivate, but they do not know anything about what the student is learning. Given this choice, I would pick Option A – the in-person paper-based no-tech option.

I think the major factor here is the age of the student. The physical and social aspects of learning matter a lot more for younger students than for adults. With younger students, it is not worth trading off the physical and social. As students get older, the trade-offs start to make more sense.

You also have the extra wrinkle which is that LLMs have made cheating far easier for everyone, making a lot of unsupervised screen-based tasks untenable even for older students. There is a lot of interest at the moment in “online proctoring” where you use screen monitoring software and cameras to prevent a student from cheating. However, in many contexts I think we might conclude that human-supervised paper-based tasks provide an easier and more effective solution.

Ultimately, I am making an empirical argument. If it turns out I am wrong, then, like the travel agents of the 1990s, we will soon see it show up in practice. Schools will start closing as parents choose to educate their children at home on screens, or schools will restructure themselves around individual computer pods.

There are a lot of experiments like this happening at the moment. Most famously, Alpha School in the US have gained a lot of attention over the last couple of months for their particular approach to technology and education. I haven’t visited an Alpha School, although I have used one of the apps they’ve developed – Math Academy, which I thought was very good. I remain sceptical that students can learn all the academic content they need from apps like this, however.

So what do I suggest instead?

Here are three specific approaches that I think embody my “limit-free motorway and car-free town centre” concept.

1. Paper-based AI assessment. Our approach at No More Marking embodies these principles. We use AI-enhanced comparative judgement to assess student writing, but our assessments are paper-based, so students never need to see a screen. Handwriting recognition technology has improved dramatically in the past couple of years, which is a great example of how new technological advances can improve the efficiency of older media.

2. Professional development at scale. If it’s easier for adults to learn on-screen than students, maybe the solution is to create high-quality online professional development resources for teachers. Over the last decade or so, England has had a lot of success scaling up quality instruction of phonics. There are several government-approved programmes that consist of detailed student resources and teacher training that explains how the resources are used. I think we could copy this approach for other concepts and subjects, but use technology to make the training easier for all teachers to access.

3. LLM tutors. LLMs are making screen-based instruction harder because, as noted above, they make cheating easier. However, I think LLM tutors do offer some promise because they are more responsive than traditional learning apps. In older year groups where a little bit of screen time is more acceptable, I think there is a role for them. The best model I have come across is Google and Eedi’s collaboration. A human teacher teaches a maths concept to a class, and then each student uses an LLM tutor for individual practice of the concept.

The basic challenge here is common to a lot of other fields: how do you maintain the human essence of a particular service but also scale it up so that it is available for millions of people? This is not easy, and it requires careful thought and design. But it is also unavoidable. Children can only learn with substantial human input, but there are eight million children in England alone, and we have to find ways of providing high-quality instruction to all of them.

We need to be more sceptical about screens in the classroom, at the same time as being more ambitious about the power of technology to raise standards.

1. I am not pretending to be an expert on German car policies, and given that, this metaphor could backfire. Perhaps there are huge controversies about these laws, and perhaps even now the Bundestag are debating proposals to change them. However, I think the central point stands, regardless of what happens in Germany. In my own context of London, it is possible to be in favour of Covent Garden being pedestrianised, and also in favour of more rapid construction of the Lower Thames Crossing.