Personalised learning – there’s more to it than I thought
In April I attended the Durham Blackboard User Group conference which followed the theme of personalised learning. As someone more focussed on the technical side of things, the theme was a bit of a challenge for me, but it did get me thinking.
What is personalised learning?
Before the Durham event, my concept of personalised learning was simply providing a framework for students who are perhaps struggling with their studies to improve their learning. Or being able to choose optional courses as part of their programme of study.
Both Microsoft’s Copilot and University of Edinburgh’s ELM platform have similar definitions of personalised Learning to each other:
Microsoft Copilot
Personalised learning is an educational approach that tailors instruction, content, pace, and learning environments to meet the unique needs, preferences, and interests of each learner. Instead of a one-size-fits-all model, personalised learning recognizes that students learn in different ways and at different speeds.
University of Edinburgh’s ELM
Personalised learning refers to educational strategies and practices that tailor learning experiences to individual learners’ needs, capabilities, and preferences. The aim is to enhance learning by making education more relevant, engaging, and effective for each student. Personalised learning takes into account students’ previous knowledge, their learning speeds, preferred learning styles, and even their personal interests and future career goals.
While I wasn’t necessarily wrong with my definition, there is clearly a lot more to it than I realised.
Individual Learning Plans
Perhaps the closest example to what I initially thought of personalised learning is Individual Learning Plans. At the highest level, this would be allowing a student to pick their own optional courses away from the core required for their programme. For example, when I was a student, I chose Exploring the Cosmos and Earth Science as optional courses. Very different from my Computer Science programme and dare I say it, a little more interesting!
Alternatively, an instructor can curate a curriculum for individual students based on various factors including personal interests. For example, a physics student may have a particular interest in nuclear physics. They instructor could mould some of the course content specifically around that for the student by giving them a project around that topic.
Depending on just how personalised the curriculum for each student will determine how much additional work will be required for the instructor. But is there a way personalised learning could be delivered efficiently at scale? Artificial intelligence for example?
Using Artificial Intelligence Models
The growth of artificial intelligence has exploded in the last couple of years and doesn’t show any signs of slowing down any time soon. So how can AI be deployed to drive personalised learning?
There are many natural language models available such as GPT developed by OpenAI, Microsoft’s Copilot developed by OpenAI & Microsoft and Google’s own Gemini created by Google Deepmind. These can be deployed in different ways including building custom models based on them such as Edinburgh’s ELM which is built from GPT-4.
These models allow Socratic questions to be used and a “conversation” to happen. Using the open ended nature of the these models, you can explore topics in ways that feel natural to you. The path taken will be very different from one user to another and even if you have the same conversation again as they are based on prompts provided by the user.
Some of the feedback provided at the Durham event suggested that using NLMs can have limited use as users can get to a point where the model is stuck in a loop and can’t end the “conversation”.
Another challenge with NLMs is their knowledge base. For example, Copilot’s knowledge base was last updated in October 2023; however, it has the ability to search the live-web for more up to date information. By contrast, Edinburgh’s ELM was last updated in December 2023 but doesn’t have the ability to search outwith its own knowledge base.
Depending on the subject matter, this might not be an issue. But in fields where rapid changes happen such as medicine, it means that the model you use needs to be selected carefully.
I have only briefly touched on two areas around the topic, but others that can be explored include flipped classrooms, adaptive learning software and using conditions within LMSs that can track student progress highlighting where instructors may need to intervene to ensure the student is coping with the course.
Conclusions
What do I think about personalised learning after the Durham event? First of all, it is a much larger area than I realised but I have started to get an appreciation for what it entails. With the growth of AI it could help to drive more personalised learning across a broad range of areas and potentially remove some of the work involved in setting up personalised learning paths for students. Of course, that will depend on ensuring the data available through those models are accurate and up to date.
Personalised learning can be an effective way for building up someone’s knowledge on topics and providing a formative framework for students; however, in subjects that require professional accreditation like medicine, where there is a much more rigid curriculum, I believe that it is a significant challenge to provide a more personalised experience.
The Durham event has developed how I think about personalised learning, and I am sure with the rate technology such as AI is developing I will be revisiting that again.
Guidance for students and staff on AI use is available on the university website.
Share
