Will AI revolutionise education for the better?

Much of the public conversation about artificial intelligence in education has focused on the perils of plagiarism — students using AI tools, such as ChatGPT, to write convincing essays that they pass off as their own, and academic institutions, in turn, deploying AI tools to try to catch these cheaters. 

But, as new technologies including AI break into the mainstream, they also promise to revolutionise learning for the better. 

According to Hadi Partovi, chief executive of education non-profit Code.org, AI “is going to be one of the largest education platforms outside of school.”

Already, there are AI-powered platforms and tools emerging that can support learners — particularly those who are struggling, or from low-income backgrounds — and potentially improve equity across the board.

For example, there are already AI-powered ‘tutors’, that can offer a personalised experience to the tutee, adapting to their needs in real time. One app — Khanmigo made by the Khan Academy, a US-based non-profit online educational organisation — dubs itself “your always available study-buddy”, promising to challenge learners “to think critically and solve problems without giving you direct answers”. The app is free for teachers, and $44 a year for users and their families.

“Until now, expert, personalised tutoring has been cost-prohibitive for all but the most elite learners and institutions,” points out Joseph South, chief innovation officer at the International Society for Technology in Education, an education non-profit. “Generative AI has the potential to scale this experience for online learners at an affordable price.” 

One 2024 study, led by the University of Oxford, found that school students in Ghana who used an AI-powered maths tutor accessible via WhatsApp for half an hour, twice a week, over eight months had “substantially higher” maths growth scores than peers in a control group.

Emily DeJeu, assistant teaching professor of business management communication at Carnegie Mellon University notes that there is further research to show that low performers “get a significant boost” when using the guidance powered by AI. “It fills in their gaps, it makes them more productive,” she says.

The technology can also be used in novel ways for simulating working environments or training. “Transferring knowledge from the classroom to the workplace has always suffered from the artificial nature of a classroom setting, which lacks the complexities of real-world application, says South. “No more.” He adds that AI-generated simulations mean that students can enter workplaces “having pressure-tested their performance in simulated environments”.

As well as helping students directly, AI can be wielded by educators, themselves. It has the ability to help generate personalised learning plans, and prepare individualised assessments.

“AI transforms the educator’s role, saving significant time on lesson planning and assessment, thus allowing more classroom interaction and individualised support,” says Christophe Mallet, chief executive of Bodyswaps, a virtual reality soft skills training provider.

For example, AI can swiftly generate problem sets on a chosen topic. “It can also triage student work to identify individual errors,” adds South, “as well as patterns of performance that need to be corrected.” 

In business schools, it can generate innovative and original cases studies, too, says DeJeu.

All of this has the potential to save staff significant amounts of time. A report by educational publisher Twinkl found that adopting AI could prevent $77bn of unpaid teacher overtime across the US. That costing was based on McKinsey data suggesting teachers could regain as many as 13 hours a week by embracing AI tools.

New, up-and-coming tools include the lesson planner MagicSchool, which says it has 2mn educators signed up, and Pressto, which will generate writing prompts for teachers based on the subject topics or level they want. Merlyn Mind is another an AI-powered assistant that teachers can use in the classroom. 

However, there are still potential problems to address. AI is still unreliable and prone to hallucinations — presenting fiction as fact — which could mean students might ‘learn’ incorrect information. Similarly, AI can develop biases based on the data used to train it, and potentially pass them on to learners.

In addition, experts warn that any educational institution needs to have restrictions on what student data and personal information can be fed into AI systems, to mitigate privacy risks. 

But, to some, the biggest risk is to the learning process. DeJeu says that, as students become increasingly used to relying on technology from a young age, AI tools could mean they are “shortcutting” learning. 

“Is that good for them long term?” she asks. “Some of that is still an open question. The challenge is constantly narrating to students ways to use these tools for their net benefit, professionally, and tips to avoid using it in ways that shortcut their growth.”

DeJeu believes teachers will need to be “sufficiently tool agnostic” in introducing their students to AI, as well as in their own usage of technology. She advises them to seek out “timeless” use cases for AI because the tools and capabilities are evolving so rapidly. 

Mallet notes the risk of “blindly replacing human educators by AI for learning on the basis of cost-efficiency”. He also warns that a “rapidly widening AI skills gap among students” could emerge, as some schools and institutions are banning AI altogether over concerns about cheating, while others are embracing it. 

To ensure the correct guardrails are in place to protect students, Partovi of Code.org suggests “we need government policies — national and regional — to provide guidance to schools.” He has helped to create a consortium called TeachAI to this end. 

Nevertheless, he adds: “The biggest risk of all with AI is doing nothing. The world has changed and is changing. We should change how we teach and what we teach to embrace AI.”