Visibility around which students have understood a concept and which have not often is limited for educators. A few students respond to questions. Some appear engaged. Many remain silent. When teachers assume understanding, rather than actively verify it, a structural blind spot is created. And once a session ends, that assumption becomes the foundation for further learning.
For students whose understanding is incomplete, subsequent learning can be more difficult. Gaps that remain unaddressed can accumulate over time. Students may struggle to apply concepts, lose confidence or disengage. By the time these issues become visible – through end of semester or end of year assessment results – the opportunity for simple intervention has often passed.
Perhaps what students need is just-in-time support – short, targeted bursts of explanation, practice or feedback delivered when confusion or cognitive overload occurs. This is where AI could be transformative. Instead of pushing students to the next topic before they are ready, lecturers can use AI to produce personalised learning pathways. A series of checkpoints within the course could deliver incremental learning. This would ideally encourage students to consolidate the topic according to their own learning time before progressing further.
Busy lecturers can use AI to create, within minutes, bite-sized learning resources that students can work through, revisit and recombine to consolidate their understanding as needed. These can range from short readings or a brief case study to two-minute explainer videos or podcasts. Lecturers could introduce these elements of micro-learning after each topic or bi-weekly to validate the students’ learning during the course.
In my own teaching practice, I have seen the benefits of intervening just in time, before confusion sets in. A short diagnostic quiz developed using AI assessed student learning. Depending on their score or learning level, students were directed to individual learning pathways. Students who had previously disengaged in lectures and had fallen behind began re-entering the learning process. They described the short content as “manageable”, “less overwhelming” and “easier to fit into their busy life”.
The difference was apparent.
AI can enable lecturers to take adaptive learning to a new level. I have developed a toolkit, a just-in-time adaptive learning (JAL) model, which includes five primary areas of action. Rather than focusing only on content delivery, it identifies and responds to variation in student understanding as it emerges and looks at ways to provide timely feedback.
The model brings together:
- Checkpoint insight: Identify moments of uncertainty where gaps in knowledge emerge. Lecturers could develop bi-weekly AI-generated quizzes to assess student learning.
- Pedagogy: Depending on the students’ quiz score, students can be instructed to re-explain or reinforce learning through relevant bite-size resources that address different learning styles. This can be achieved via modular content.
- Modular content: Provide short, targeted AI-created content such as video tutorials, case studies and simulations. For busy lecturers, AI is a gift that can create materials based on lectures and reading material.
- Differentiated learning pathways: Lecturers can design learning pathways based on the extent to which students demonstrate their knowledge. This allows for personalised learning, suited to level of proficiency. The advantages of multiple pathways are that an advanced pathway may challenge, engage and motivate progressive learners to discover new concepts and ideas, whereas students at emerging levels can progress at their own pace on a recovery pathway.
- Immediate feedback: Individual, incremental feedback is time-consuming and challenging. I found AI agents provided timely feedback to students using them as a learning tool. Their feedback on problem-solving tasks was student-friendly and interactive. Lecturers can use these AI tools to develop learning activities ring-fenced with their lecture notes and selected readings.
Together, these elements create a more responsive learning process – one that recognises that understanding does not develop uniformly.
This approach does not depend entirely on advanced systems. It is possible to implement learning pathways using tools such as Copilot, ChatGPT and Claude. AI systems such as Synthesia or HeyGen will create video content from written content. Most of these tools are supported by university learning management systems. Tools are also available that convert speech to text and voice to text, support gamification and game-based learning that allow difficulty to be adjusted, and use Google Assistant and Apple Siri.
Students and teachers can use AI to plan, organise and summarise complex reading into flashcards, and assignment briefs into manageable chunks. It also places extra demands on teachers. Supporting adaptive learning and developing AI to create individual student pathways requires new skills and time to learn. However, once learned, the skills will save time in content creation and can reach varied learners and produce reusable resources. Institutions should provide training, tools and workload recognition for this work.
These are just-in-time strategies that educators can apply immediately:
- Build in frequent, low-stakes checkpoints of understanding.
- Use AI to generate learning material to target follow-up activities in different modes, with multiple explanations and examples of concepts for independent learning for a diverse range of students.
- Create opportunities for students to reflect on their own learning and uncertainty early.
- Provide a toolkit to initiate their own independent recovery to learning.
These changes shift the focus from assuming understanding to actively supporting it. AI can be used effectively here, and it gives institutions an opportunity to explore the potential of just-in-time adaptive learning.
Zabin Visram is a professor of hospitality at Glion Institute of Higher Education London.
Leave a Reply