My mum’s favourite TV show when she was a child was The Flowerpot Men.
As she tells it, she would sit in front of the TV watching the test card, waiting eagerly for the programme to start.
She would sit entranced for 15 minutes, and then it would end, the test card would come back on, and that was it for Bill & Ben until the next day.
That was the media environment of the early 1960s.
A few years ago I suggested to my mum that she might like to try out Duolingo. She has always wanted to learn another language and never got very far. I quite like Duolingo and thought she might too.1
After downloading it and trying it out she rang me up. “This app is AMAZING! I would have LOVED this as a child! But what I don’t understand is - it’s so good and so much fun - why aren’t all the kids on it all the time? Why aren’t they all fluent in ten languages?”
Because the kids today have more options than watching a ropey black-and-white puppet show.
The attention arms race
Let’s imagine a thought experiment where we can take Duolingo back in time to the 1960s - but no other modern technology or entertainment. Just Duolingo. I think then that probably every kid in 1960s Britain would have completed ten languages by the time they were 12. It would have been more entertaining, more addictive and more exciting than any kids’ media product of the 1960s.
But of course, that is not how it works. Modern technology is not restricted to Duolingo, or to learning apps. It is used to create a whole range of entertainment products.
Most of these entertainment products are trying to do just one thing: hold the user’s attention.
Teachers and educational apps are also trying to hold their students’ attention. But they are trying to do another thing too: they want their students to learn.
This basic dynamic is why, in a straight fight with an entertainment product, teachers and education apps will always lose out. The entertainers are optimising for one parameter, and the educators are optimising for two.
However much fun you make learning, someone else will use the same techniques minus the constraint of learning. You are in an arms race where you have one arm tied behind your back. Your rival can use all the techniques you can, plus several more that you can’t.
Is learning a constraint? Am I being unduly negative here? You will hear people argue that with a good teacher, learning is intrinsically fun. And my response to that is you know what is actually, literally intrinsically fun? Fun!
There are serious academics who are proposing that one of the reasons people are having less sex is because the online world is so compelling and addictive.
If online content is so addictive that it can lead to adults having less sex, it’s probably fair to speculate that it will lead to kids studying less algebra.
The maths of app design
We can think about this mathematically, using exactly the same methods that the app designers themselves will use. An entertainment creator will review millions of data points showing the exact moment in a video that people switch off, and the exact moments that people rewind and watch again and again. They can then design future content to have less of the former and more of the latter.
Online learning creators will review similar data on attention, but they will also review learning data: how many students get this question right, how many get that question right, how long it takes for a memory to fade, etc.
The online entertainment creator therefore has a simpler task: more of whatever sustained attention, less of whatever did not.
The online learning creator can also see that engagement spiked at point x and dipped at point y. But what if learning dipped at x and spiked at y? Or what if learning spiked at point z? What if you realise that students need 10 repetitions of content in order to truly understand it, but that engagement drops precipitously after 7?
You will have to optimise for the two parameters of learning and engagement, and decide (for example) that 90% engagement and 70% understanding is better than 30% engagement and 100% understanding.
Your engagement rates will inevitably be lower than those of an app that is purely optimising for engagement, because optimising across two parameters is always going to result in some compromises. Suppose you ask one company to build a car that is really really fast, and another company to build a car that is really really fast and also capable of carrying 8 passengers. Who is going to build the faster car?
Even worse than that, we are not living in a static world where the trade-offs you make remain constant. We are in a dynamic world where the rivals for student attention are constantly and ruthlessly optimising on just one of your parameters in ways that will progressively distort your entire equation. Keeping your engagement levels up is going to require more and more trade-offs, more and more compromises. Suppose by some miracle you do create a learning app that is more engaging than a pure entertainment app. The techniques you have used will be instantly copied by your rivals and if there is any way at all that stripping out the learning content will make them more engaging, they will find that way. They can copy all of your techniques, but you can’t copy all of theirs.2
Slow and steady does not always win the race
If lots of viewers turn off at a certain point in a MrBeast video, he can ditch that content in future videos. If lots of viewers turn off at a certain point in a maths video, you can’t just ditch that content. You can obviously try to rework it and improve it and make it more engaging, but you are far more constrained than MrBeast. You can’t replace your explanation of simultaneous equations with a video of a homeless man being given £10,000 without seriously compromising learning.
I mention MrBeast deliberately because there was a recent Guardian Long Read outlining some of his methods. He has turned attention-grabbing into an art and a science. To watch one of his videos is the equivalent of watching a cheetah chase its prey. The cheetah has been optimised for speed through millennia of evolutionary pressures. MrBeast’s YouTube videos have been optimised to grab your attention through millions of data points. Expecting learning to compete is like expecting a Galapagos tortoise to evolve to outrun a cheetah. Even if it could do so, it wouldn’t be a tortoise any more by the time it had finished.
I think the basic dynamic I have described above has always been theoretically true. However, it was practically less relevant before social media provided content creators with such rich feedback data, and before children carried online content portals with them 24/7. So 20 years ago, if someone said “if kids are bored it’s the teacher’s fault” I would still have disagreed with them, but I would have conceded that an above-average teacher or above-average learning website could produce high levels of student engagement and learning.
We are beyond that point now. My theoretical point is now also practically true. Learning is the slow, ponderous and beautiful Galapagos tortoise, and online content is the invasive predator which will inevitably drive it to extinction.
I was surprised to read the education writer Andrew Rotherham complain about school phone bans recently on the grounds that they are an admission of defeat. He thinks that educators should focus more on making their lessons exciting, and less on banning phones. In one sense he is right – banning phones IS an admission of defeat. But he is profoundly wrong to assume that there is a way teachers can win this fight by just “being more exciting”. The reason schools are banning phones is that they are realising this is a fight you can never win, and the most sensible thing to do in a fight you can never win is to change the terms of engagement. Change the battlefield to one that gives you a better chance of winning.
The Galapagos Islands of the mind
Go back to my thought experiment of “Duolingo in the 1960s.”
You could artificially create such an environment by banning all non-educational uses of smartphones and computer games. I have heard a rumour that this is what happens with TikTok in China - videos for kids are restricted to ones about maths and piano lessons. I have no idea if this is true or not, but is essentially a similar tactic to the one used to save the Galapagos tortoise. You don’t change the Galapagos tortoise, you change the ecosystem. You don’t change the human, you change the environment.
I am not advocating for this kind of action to be taken by governments in democracies. But I am advocating for schools and individuals to make decisions that help create such environments. And I think in the next few years, whatever governments do, we are going to see more and more schools and individuals voluntarily opt in to a “Galapagos Islands” of the mind.
I realise a lot of people don’t like Duolingo. I think it’s pretty good and does a decent job of embedding basic principles from learning science. I am not claiming it is the best language learning app, but it is probably the best known. I would tend to agree with these thoughtful criticisms from Doug Lemov.
An analogy is Rangers FC’s signing policies. Up until the late 1980s, Rangers had a self-imposed ban on signing Catholic footballers. Celtic had no equivalent ban on signing Protestants. The legendary Celtic manager Jock Stein said that if he had a choice between Catholic and Protestant players of equivalent abilities, he would pick the Protestant, because then Rangers couldn’t have the Catholic. Essentially, if your opponents can copy you, but you can’t copy them, then they have an in-built structural advantage.
Great piece. A bit tangential, but on the subject of making learning fun—I saw Hamilton on stage last year. On the way out, I overheard a particularly grumpy audience member chastising secondary school history teachers for not making their lessons as fun and engaging as the hip hop musical. I didn’t have the heart to point out that if teachers could write a global smash hit stage show, they probably wouldn’t be stuck marking essays—and I don’t think anyone could pass a test on the American Revolution based on watching the musical.
Very thought provoking as always. But I think the analogy is a bit different. Classrooms absolutely can be like Duolingo. The engagement in Duolingo is not that it is entertaining (I am one of the daily users who hates it). The engagement is that you are constantly successful because of the way information is chunked and practice is organised and sequenced. Lessons can absolutely be organised this way.
Likewise the incentives of points gained, league tables and slightly annoying celebrations of streaks within lessons and across days and weeks creates proof of success and progress, which motivates despite the lack of 'fun'.
I'll go a step further - students would learn languages in school much better if 50% of lesson time was devoted to Duolingo. 25 hours took me to grade 5 on a reading paper, which is the equivalent of January in a year 7 curriculum.