AI in education: what those buzzwords mean

You’ll be hearing a great deal about artificial intelligence (AI) and education in 2025.

The UK government unveiled its “AI opportunities action plan” in mid-January. As part of the plan it has awarded funding of £1 million (about US$1.2 million) to 16 educational technology companies to “build teacher AI tools for feedback and marking, driving high and rising education standards”. Schools in some US states are testing AI tools in their classrooms. A Moroccan university has become the first in Africa to introduce an AI-powered learning system across the institution.

And the theme for this year’s United Nations International Day of Education, observed annually on 24 January, is “AI and education: Preserving human agency in a world of automation”.

But what does AI mean in this context? It’s often used as a catch-all term in education, frequently mixed with digital skills, online learning platforms, software development, or even basic digital automation.

This mischaracterisation can warp perceptions and obscure the true potential and meaning of AI-driven technologies. These technologies were developed by scientists and experts in the field, and brought to scale through big tech companies. For many people, the term AI reminds them of systems like OpenAI’s ChatGPT, which is capable of writing essays or answering complex queries. However, AI’s capabilities extend far beyond these applications – and each has unique implications for education.

I am an expert in AI, machine learning, infodemiology – where I study large amounts of information using AI to combat misinformation – knowledge mapping (discovering and visualising the contents of different areas of knowledge), and Human Language Technology (building) models that use AI to advance human language, such as live translation tools. I do all of this as the head of the Knowledge Mapping Lab, a research group within the Faculty of Economics and Management Sciences, and co-director of the Interdisciplinary Centre for Digital Futures at the University of the Free State.

In this article I explain the technologies and science behind the buzzwords to shed light on what terms like machine learning and deep learning mean in education, how such technologies can be – or are already being – used in education, and their benefits and pitfalls.

Machine learning: personalisation in action

Machine learning is a subset of AI involving algorithms that learn from data to make predictions or decisions. In education, this can be used to adapt content to individual learners – what’s known as adaptive learning platforms. These can, for example, assess students’ strengths and weaknesses, tailoring lessons to their pace and style.

Imagine a mathematics app that asks questions based on the curriculum, then uses a learner’s answers to identify where they struggle and adjusts its curriculum to focus on foundational skills before advancing. Although the science is still being explored, that level of personalisation could improve educational outcomes.

Deep learning: assessment and accessibility

Deep learning is a branch of machine learning. It mimics the human brain through neural networks, enabling more complex tasks such as image and speech recognition. In education, this technology has opened new avenues for assessment and accessibility.

When it comes to assessment, AI-driven tools can assist in marking, analyse handwritten assignments, evaluate speech patterns in language learning, or translate content into multiple languages in real time. Such technologies can both help teachers to lessen their administrative loads and contribute to the learning journey.

Then there’s inclusivity. Speech-to-text and text-to-speech applications allow students with disabilities to engage with material in ways that were previously impossible.

Natural language processing: beyond ChatGPT

Natural language processing is a branch of AI that allows computers to aid in the understanding, interpretation and generation of human language. ChatGPT is the most familiar example but it is just one of many such applications.

The field’s potential for education is huge.

Natural language processing can be used to:

analyse student writing for sentiment and style to provide real time feedback into the thinking, tone and quality of writing. This extends beyond syntax and semantics
identify plagiarism
provide pre-class feedback to learners, which will deepen classroom discussions
summarise papers
translate complex texts into more digestible formats.

Reinforcement learning: simulating and gamifying education

In reinforcement learning, computer systems learn through trial and error.

This is particularly promising in gamified educational environments. These are platforms where the principles of gamification and education are applied in a virtual world that students “play” through. They learn through playing. Over time, the system learns how to adapt itself to make the content more challenging based on what the student has already learned.

Challenges

Of course, these technologies aren’t without their flaws and ethical issues. They raise questions about equity, for instance: what happens when students without access to such tools fall further behind? How can algorithms be prevented from reinforcing biases already present in educational data? In the earlier mathematical example this might not be as much of an issue – but imagine the unintended consequences of reinforcing bias in subjects like history.

Accuracy and fairness are key concerns, too. A poorly designed model could misinterpret accents or dialects, disadvantaging specific groups of learners.

An over-reliance on such tools could also lead to an erosion of critical thinking skills among both students and educators. How do we strike the right balance between assistance and autonomy?

And, from an ethical point of view, what if AI is allowed to track and adapt to a student’s emotional state? How do we ensure that the data collected in such systems is used responsibly and securely?

Experimentation

AI’s potential needs to be explored through experimentation. But this works best if managed under controlled environments. One way to do this is through regulatory AI “sandboxes” – spaces in which educators and designers can experiment with new tools and explore applications.

This approach has been used at the University of the Free State since 2023. As part of the Interdisciplinary Centre for Digital Futures, the sandboxes serve as open educational resources, offering videos, guides and tools to help educators and institutional leaders understand and responsibly implement AI technologies. The resource is open to both students and educators at the university, but our primary focus is on improving educators’ skills.

AI in education is here to stay. If its components are properly understood, and its implementation is driven by good research and experimentation, it has the potential to augment learning while education remains human-centred, inclusive and empowering.