Last September, I had the chance to present an idea on predictive learning at Moodle Moot Africa, held in Johannesburg (https://ilite.africa/blog/predictive-learning-presentation/). While the presentation itself may not have been my strongest performance, I take pride in my determination, which often compensates for any gaps in presentation skills.
Recently, I had the opportunity to present the concept of predictive learning to Dr. Yangama Jokwiro, an associate professor at La Trobe University and co-founder of the Vaka Health Foundation. Vaka Health is dedicated to driving digital health transformation across Africa by developing innovative solutions aimed at improving healthcare access, education, and system efficiency. Their initiatives blend technology, policy, and education to advance nursing practices and global healthcare delivery.
As part of their educational efforts, they utilise a Moodle platform (https://nursesguild.africa/), and it is within this innovative environment that iLite endeavours to contribute meaningfully.
The Nurses Guild Moodle site (https://nursesguild.africa/) serves as a central learning hub for nursing professionals across the continent, fostering growth and collaboration in the field. iLite’s Predictive Learning Tool, Sherpa, has been carefully designed to enhance this platform by ensuring that every piece of content—whether courses, PDFs, images, or videos—is discoverable, indexed, and summarised for straightforward access and optimal usability.
In addition to Sherpa’s capabilities, iLite is actively working on integrating cutting-edge AI solutions into its clients’ existing Moodle environments. This initiative aims to elevate the overall learning experience by leveraging artificial intelligence to create smarter, more adaptive systems. Through these efforts, iLite seeks to provide learners with highly personalised pathways to knowledge, enabling them to access curated content that aligns with their unique needs and goals. By blending innovative technology with the robust infrastructure of Moodle, iLite continues to redefine digital learning for nursing professionals and beyond.
The predictive learning plugin aims to make curated content discoverable, thereby supporting the learning process.
This contributes to a greater purpose, enabling nurses to fully develop their skills by reflecting on their progress, identifying skill gaps, and finding courses that support their development. Through the integration of tools like Sherpa, along with AI-driven enhancements, the learning environment becomes more dynamic. These technologies not only ensure accessibility to diverse educational resources but also cultivate a tailored learning experience. When supported by visual aids or interactive features, such as those showcased in training modules or learning dashboards, learners can achieve deeper engagement and understanding, bridging the gap between theoretical knowledge and practical application.
Book to Bot
However, that is only part of the story. Dr. Jokwiro authored a pharmacology book, available on Amazon (https://www.amazon.com/Learn-Pharmacology-Classes-Clinical-Placement-ebook/dp/B0CPSBC9MX). During our discussions about predictive learning, Dr. Jokwiro pointed out that students were increasingly turning to AI rather than relying on his book. This observation prompted the idea of training an AI bot specifically using the content of his publication.
Dr. Jokwiro kindly provided us with a Word copy of his book. We used an open weights model from Huggingface to perform the chunking and store the results in a vector DB. In truth, this did take a little trial and error to dial in the chunk size and overlap. I am not convinced that we have it right, but we certainly have a working model. Not that I have a lot of experience, but enough to know it’s not just a tweak in one area that brings the most gains.
I then tested a few small models that are efficient for question-and-answer tasks. While the initial results are promising, there remains a fair bit of room for improvement. I think a lot more effort will be spent on finding the right model for our purpose and budget.
But for now, the current system is functional and meets basic objectives, laying a foundation for further refinement. Collaborating closely with Dr. Jokwiro ensures that we align academic rigour with technical advancements, striving toward a solution that is both reliable and precise.
Looking to the future, it is possible that all content within Moodle will be used to train the bot, enabling answers to be derived not only from the book but also from the entire site.
A final thought: The UI is very far from finished, but the aim for this phase was to prove that we could integrate the predictive learning tool into The Nurses Guild Moodle site and create a bot trained on some of our own content.