Machine learning is helping children to learn – and teachers to teach

A few years ago, I became an uncle for the first time and I’m honestly amazed at how fast my nephew seems to be growing; to coin a phrase “they grow up so fast”. Speaking as a Millennial, technology has certainly influenced and had a part to play in my development and growth. For my nephew, and the Generation Alpha’s, it’s hard to truly comprehend how great the sphere of influence will be given how fast technology has progressed. The use of technology in education is no longer just the use of a Casio calculator, it’s the use of machine learning.

Post-lockdown, Zoom assemblies may be fading from memory, but technology is playing a bigger role than ever in helping teachers educate their students. Artificial intelligence (AI), in particular, provides essential classroom support in the post-pandemic world of growing class sizes and diverse educational needs.

Forget about the apocalyptic AI of sci-fi. In real life, AI has grown to become a powerful and helpful part of our everyday experience. It's been years since we started using spam filters and autocorrect functions powered by machine learning (ML), a type of AI that constantly learns in response to data and human input, such as when you correct its occasional mistakes.

This smart interactive process has proved valuable in the classroom, and of course outside the classroom during the Covid lockdown. Developers in the growing area of EdTech (education technology) have harnessed ML and used it to help children learn in clever, engaging ways.

How ML helps children to learn

Smart online tools such as Fingerspelling.xyz, which uses ML to teach sign language, came into their own during lockdown, when children were deprived of the usual interactive feedback of the classroom. Language app Duolingo and writing guide Grammarly are other well-known educational apps powered by ML.

We're now seeing more and more educational tools designed to teach children about machine learning itself, and to introduce them to AI concepts such as image recognition and augmented reality (AR). After all, tomorrow's adults will live in a world where AI and ML power everything from cars to healthcare decisions, so it's important that young people both understand how this technology works and how to get the best out of it.

Google's Teachable Machine is another, wonderfully accessible tool that lets children create their own machine learning models, then train them to complete tasks like pattern recognition, without the need to write or learn any code at all. For children, their parents and the teachers who do want to learn about coding, there are plenty of tools to help them on their programming journey.

It's also worth knowing that schools and other educational organisations can create their own AI and ML tools with no programming expertise needed, thanks to platforms like Amazon Personalize, part of Amazon Web Services (AWS). Personalize allows educational providers to design and develop smart ML tools on a pay-as-you-go basis, without having to worry about running their own IT systems.

These are, however, many others which are free to use or try. They include, Machine Learning for Kids, which is already used by thousands of schools in the UK and beyond. The platform guides children through the process of designing and building ML games and other projects with it. Completely free.

Dalton Learning Lab is another new platform that teaches coding in the widely-used ML language Scratch, while Kano Computing invites youngsters to create their own versions of computers, and then complete tasks by programming their computer. Finally, Wonderville Viper explains machine learning in a fun, easy-to-understand way, and then challenges young users to program a robot to explore a virtual planet. For the next generation, it's all a lot more fun than learning times tables.

How ML supports enhanced teaching

Machine learning has been a huge help for teachers as well as for pupils. Software such as Gradescope, which marks homework and exams with the help of students, and Otter, which automatically transcribes audio recordings, take care of time-consuming tasks, and enables teachers to focus on more qualitative teaching work.

AI or ML will never replace human teachers, says distinguished Freeform Dynamics analyst Tony Lock, who worked as a volunteer classroom assistant early in his career. But he thinks ML observation tools in children's computers can help teachers monitor and nurture ever-growing classes of children with diverse needs.

"You could potentially use cameras to detect behaviour changes," says Lock. "They can help spot who needs more help, or who's bored, and who needs more work because they're ahead of the game. ML can then offer suggestions based on the content, and on strategies other teachers have found helpful in similar scenarios."

It's the equivalent of having 30 pairs of well-informed eyes looking at the class, says Lock. "By offering personalised recommendations, it's got the potential to make teaching more individualistic. That's particularly useful in the UK, where teacher-pupil ratio numbers are so high now."

The machine learning element of this AI software would ensure that the recommendations continue to improve as time goes on, because it adapts to their outcomes. Like a good student, AI takes feedback on board. For example, if it offers the pupil a certain reading recommendation that she doesn't get on with, then that feedback will help inform and tailor its future suggestions.

The role of data

For any ML system to deliver accurate, useful predictions, it needs to draw upon large, unbiased sets of data about children's behaviour and achievement, as well as teachers' own suggestions. Data collection has to be done at scale to help avoid biased recommendations that get entrenched over time, but it needs to be done carefully and sensitively, and all of which needs to be stored safely in a high performance data centre.

"Gathering data will need to involve parental permission," says Lock. "PhD students, teaching assistants or postgrads can probably help here. They could collect data on pupils' reactions in class, for example when they're using computer systems, and then work on anonymising and analysing those data sets to feed into the ML. It's clearly something that needs funding."

With the right governance, and large enough data sets of sufficient size to avoid algorithmic bias, ML could be an enormous help for teachers and pupils in future – and by drawing upon data from other teachers, ML effectively turns teaching into a remote collaborative process. Further, by taking care of much of the monitoring and repetitive work of teaching, the technology also frees them to focus on the rich, interactive strategies that make every teacher unique.