Does education need Artificial Intelligence?

Carousel Learning
5 min readAug 9, 2022
Photo by Alex Knight

“Artificial intelligence,” it was recently put to me, “will obviously never replace teachers. The decisions they make around sequencing ideas, communicating concepts, responding to students and building relationships cannot be replicated by a machine.”

Sounds sensible. Short of tech fantasists who have watched too much sci-fi, I don’t think anyone really believes there is a plausible scenario whereby teachers are replaced by artificial intelligence (AI) in the foreseeable future.

The person I was talking to did not stop there though:

“Where I do think AI can be useful is in the diagnosing and tracking of student knowledge as it builds over time. The AI figures out what the student knows and what they don’t know, and ensures that students are re-exposed to things they don’t know at appropriate times.”

This is a better point, and it is one worth discussing, especially with Prime Ministerial candidates waxing lyrical about AI’s implementation in education. The case that was put to me sounds sensible, and it doesn’t just use AI for the hell of it — to sound sexy and progressive but to fill no real need. In this case, we have passed what I call the “Perry Test,” a test which essentially asks:

Does your use of educational AI solve a real problem in the learning process that cannot be solved more easily in other ways?

In this case, the “problem” is keeping track of things students do and don’t know. In GCSE Chemistry, for example, there are upwards of 700 knowledge items a student needs to know and be able to do in order to get a good grade. Each of those individual items needs to be connected to various others and can be in varying states of decay — one student might have clean forgotten the names for the different subatomic particles, whereas another might have remembered the names but can’t remember which ones are in the nucleus of an atom and which ones orbit the atom. I think it’s obvious that keeping track of all this for a class of 30 students would be impossible for most teachers.

AI could certainly help with this. First, the system could repeat particular items depending on how well students know them. If a student gets question X wrong, then the AI re-exposes them to X after 3 days, then 7, then 21 and so on. Then, the AI looks at all the other students who got X wrong, and works out which other questions they got wrong. The AI then looks at question Y and calculates the probability that a student would know Y if they didn’t know X. If the probability is low, the AI only asks them about Y once X is secured. The AI then looks at the optimal intervals between repeat episodes, i.e. if I want students to answer Y, how many times do they need to have answered X, and what should the gaps between those answers be?

Whether this really classifies as “AI” could keep us busy for quite some time, but it doesn’t really matter: the point is this is a complex activity that a computer can do that a human teacher couldn’t reasonably do. As such, we could easily argue in favour of education needing Artificial Intelligence.

The single most important factor is still the teacher who sets it, despite the quality of the homework platform.

There is, however, cause to be wary. I’ve spent a long time thinking hard about homework and how students work when we aren’t physically in front of them, and the sophistication of the student-facing platform is only one factor in determining the quality of students’ learning. For sure, if you have two identical students doing homework, and one is using a good platform and the other is using a bad platform, the former will learn more. But the platform isn’t the only factor, and it isn’t even the most important factor. Despite the fact that the work being done is homework, the single most important factor is still the teacher who sets it, despite the quality of the homework platform.

Consider two identical students, following an identical curriculum and using the same online learning platform for their homework. The only difference is their teachers: student A has a good teacher, and student B does not. Student B’s teacher doesn’t frequently check that the homework is done. When they do, they don’t always notice that some students have done it to a low standard. They rarely give consequences to students who haven’t done the homework. Most importantly, perhaps, they do not integrate the homework with the class work: they do not use it as a springboard for in-class work and do not mention it in class. They do not talk about the things that have been learnt as a result of the homework. They do not verbally note that students who are doing well in class are doing well in part as a result of their homework.

Student A’s teacher is different. They set the work regularly, always check it for completion and quality, and hold students accountable. They use what they have seen in the homework as a springboard for classwork, and deliver whole class feedback. They use phrases like “that’s a brilliant answer, and it tells me that you have been doing your homework properly.”

In my experience, the difference between these two teachers accounts for an extremely large difference in the outcomes between A and B, and perhaps a greater difference than is made by having AI in your homework platform. If a student doesn’t do the homework or does it to a poor standard or doesn’t value it, then the AI will count for little.

This is why we’ve developed Carousel to be a tool that integrates classwork with homework. Through setting Whiteboard Quizzes, and using the Mark, Analysis and Whole Class Feedback functions teachers show students the value of the homework, integrate it with classwork and hold them to account (see more on making homework effective here). It’s also why we have an entire section of our website called “Teaching and Learning”: because we think that Carousel and the teacher have to work in tandem if students are to progress in their learning. It’s why we’ve produced a guide for our users called Retrieving Better which aims not to go through the technical details of how to use this or that function, but how to make it work for the students and their learning. It’s why our stated aim is to help you teach smarter and your students learn better: we want to work with teachers to secure better learning for their students.

So does education need Artificial Intelligence? Maybe. But even if we had one, without a good teacher to direct and use it, nothing would be achieved.






Carousel Learning

Carousel is a retrieval practice and online quizzing tool that helps students to embed knowledge in their long-term memory.