How will AI transform the education experience?

How will AI transform the education experience?

Reported by Patrick McAlary, Policy Research Coordinator, CSaP

Debate about the role of AI in school is polarised. Many parents think that schools should be an and ‘AI free’ zone while others view AI as a catalyst for changing an education system that has not moved very far from when they themselves were schooled. Professor Anna Vignoles (Director, Leverhulme Trust) chaired a discussion with Dr Kevin Martin and Dr Imogen Casebourne from the Digital Education Futures Initiative (DEFI), University of Cambridge to unpack just how AI could transform the educational experience.

To listen to a recording of the panel session see below:

Issues facing education in an AI world

Professor Vignoles highlighted some of the core issues surrounding the potential role of AI in education noting that it brings up uncomfortable questions of access and resource­. It is clear that many students are using this technology, but those with socio-economic capital are disproportionately benefitting and there is a risk that those with less resource will be left behind in the AI-ification of education. Moreover, when people talk about ‘AI use’, they rarely add the caveat that this looks very different depending on the age and stage of the particular student and there is a whole set of questions relating to an emerging generational gap where younger ‘digital natives’ are more confident in using these technologies than their teachers who require upskilling and retraining. However, Professor Vignoles also issued a call for realism about AI technologies and their potential impacts. Many schools lack basic digital infrastructure and skills and while some stakeholders see great things on the horizon for AI in education, it is important to be realistic about what can be achieved in the current climate.

The rise of the Large Language Models (LLMs)

As Dr Casebourne pointed out—AI is not new. Until recently AI in education was ‘good old-fashioned AI’, these are rule-based systems that are relatively explainable. The focus has shifted to generative AI and Large Language Models which rather than being designed and trained by experts are trained on huge data sets. This raises a range of questions such as where does the data come from and who gives permission for its use—legal cases based on the use of material without permission are still ongoing. Wherever these end up, generative AI has been embedded into a range of mainstream platforms and business processes including those operating in the education realm. Google has been testing how Google LM can act as a tutor in education rather than simply spitting out answers while Merlyn Mind has released an LLM specifically trained for education on a curated data set (rather than simply piggybacking on another LLM and attempting to apply guardrails to limit certain kinds of content). These are just some specific developments that have taken place over the last 18 months and the fast pace of developments raises a host of questions: what will we be teaching students about AI and what/how will teachers learn about AI?

Can AI be more than just an efficiency tool?

AI is clearly having a moment—it is quite literally everywhere—but Dr Martin urged participants to ask themselves whether the avalanche of new AI tools are actually improving teaching, learning, and education in general: in his view, the answer is largely no.

The focus of these tools is overwhelmingly on efficiency—that is capturing time back for teachers—and this should not be ignored, but Dr Martin argued that there is a missed opportunity on what these tools could do for education. The Digital Education Futures Initiative are interested in the potential to create more dialogue-based approaches to education that will promote collaboration and conversation-based learning. New tools include the Google Illuminate project that allows the user to engage in a dialogue with an academic journal paper or book in a conversational style and these kinds of advances are beginning to leverage the real opportunities presented by AI. AI is not going to ‘replace’ teachers and the preferred future is one where AI enhances what the teacher already does and helps to facilitate better dialogue opportunities for students. Dr Martin and Dr Casebourne outlined various scenarios where emerging technology could spark learning including for teachers that need more support coming together and using an AI-based dialogic system to incite learning from one another. A persistent issue is that AI-system architects struggle to ‘tell’ LLMs what good pedagogy actually is. There is more to this than the data inputted into the system. As one panelist noted, there is a genuine risk of “garbage in, garbage out”. However, in a setting where teachers have experience to draw from, they may be able to engage with guided questions through an AI-type system to help them better articulate their experiences to one another and create opportunities to develop as a teacher.

As Professor Vignoles summaried:

"what you are describing is actually using the technology to develop relationships and communications as opposed to facts involving things which, coming into this, a lot of people would think would be the advantage of AI."

An opportunity for greater educational equity?

As a self-proclaimed ‘luddite’, Dr Martin explained that his interest in these technologies is in the opportunities that they present for creating greater educational equity. This belief is informed by his own work in East Africa where he explained how technology had created greater educational equity and access in some of the most under resourced and difficult to reach environments. Pointing to the deployment of AI tutors across school districts in Houston, Dr Martin noted that such technologies could help ‘level up’ students coming from lower resource backgrounds in countries the likes of UK and US.

AI and assessment

One aspect that was raised in the session was the threat that AI poses to standard assessment models and academic integrity. Dr Casebourne noted that there are potential routes forward in this space beyond a simple return to ‘old school’ exam-style testing. She pointed to work being done on adaptive digital assessments that responds with questions to the level that a given student operates at. On the plagiarism issue, one participant noted that an uncomfortable discussion around AI and plagiarism is necessary, but it is also important to keep in mind that plagiarism has always been an issue citing personal experience with individuals whose job it is to write essays for Oxbridge students. While Chat GPT and similar systems have arguably made plagiarism more accessible, panelists noted that in some cases plagiarism via AI was more detectable than ‘old fashioned’ plagiarism (such as essay mills).

The impression given by the panelists is that the future of education will look the same but different. The application of AI tools to education carries risks and uncertainties, but this moment also presents a real opportunity to reorientate education in a way that could support both students and teachers and help address educational disadvantage.

Image by Su Ford, CSaP Centre Coordinator