Adaptive learning and Generative AI: the winning combination for online training!
The approaches of Adaptive learning, or Adaptative learning in French, aim to modify for each learner the parameters of their course in order to improve their pedagogical effectiveness, their power of engagement or to prevent dropping out. Affected parameters may be:
This adaptation is based on different types of data, collected throughout the course of the training by the learner. Examples include the mastery score, the rate of interaction with interactive resources, or the time spent viewing instructional content.
Several adaptive learning approaches are currently used by the market, perhaps you have already tested one of them. We will see which ones and how generative AIs are challenging this status quo for the benefit of the learner.
This first approach uses advanced machine learning technologies such as deep learning applied to large cohorts of learners. They generally rely on models of the learner's level that they will update as and when observations are obtained about the learner, such as the answers to the exercises. These sophisticated models create a detailed representation of learner skills and calculate predictions to determine what content (or content parameter) could optimize learner progress.
This very specific personalization offers excellent results, superior to those of other approaches if we believe the studies on this subject, as well as valuable data for educational monitoring such as maps representing the distribution of proficiency.
Nevertheless, This approach requires large amounts of data to offer reliable and meaningful representations, raising issues of data availability and quality. In addition, on an LMS, the available content was created using different authoring tools: the data they produce (the mastery score, for example) is generally different, which requires homogenization work. Without that, algorithms can't use them to adapt learning experiences.
On the other side of the spectrum, another approach focuses on simplicity, using a set of pre-established rules to modulate learning. For example, “if the learner succeeds 60% of the first module, module 2 is unlocked. If not, we offer him the revision module.”
This method, although less efficient than machine learning algorithms, is already producing benefits. Its advantages: easy implementation, more sparing in terms of the resources needed.Its disadvantage: designers must go into the details ofeducational engineering to specify the rules in question, and define in each case, the mastery score, the path to be followed, create as many branches... which takes time and requires a certain expertise at the same time.
Generative AI is creating a new approach. It combines the power of the first approach with the simplicity of the second. And where previous approaches simply adapted the list of contents and the order in which they were addressed, this new technology allows you to create specific content on the fly, tailored to the individual needs of each learner.
A first step in this direction that is already visible in e-learning platforms is the chatbot. No more learners left alone in the face of their e-learning or condemned to navigate in an opaque forum: they can now ask their questions as they are trained to a chatbot powered by generative AI that will answer them in a few seconds. The challenge for the providers of these chatbots is on the one hand to select the right information and to provide access to it in a secure manner, and on the other hand to guide the AI so that it gives an answer that is easily assimilated by the learner.
But simply waiting for learners to ask a question is far from covering the potential of generative AI in terms of adapting learning. On the one hand, because all Learners are not necessarily aware of their needs. And on the other hand, because answering a question, however clear and digestible it may be, is not enough to learn: it is also necessary - and above all - for the learner to remobilize the information and use it in a situation.
This is why Didask is developing a set of interactions powered by generative AI that transform learners' practices. For example, just after answering a question, our AI will offer the learner to train through a practical case that it generates on the fly according to his knowledge and the needs of the learner. We also developed A coach able to analyze any of a learner's production (a chemical formulation, a commercial pitch, the application of a safety procedure, etc.) and to make him in-depth feedback, as a private trainer would do in an individual session. We have also expanded the use of our AI learning experiences directly into employee work tools so that the learner is trained as soon as they need it, wherever they are.
This set of innovations will gradually transform the way in which adaptive learning is considered and deployed. Where we used to be content with adapting ready-to-use learning sequences, we are now able to create learning experiences from scratch that meet the needs of learners. This adaptive learning, which is based on generative AIs, thus offers a finer and more dynamic personalization of training courses. We are only at the beginning and the developments that are to come in our sector over the coming months and years promise to be very exciting!
To go further on adaptive learning, you can consult our articles:
Prenez directement rendez-vous avec nos experts du eLearning pour une démo ou tout simplement davantage d'informations.
Cognitive sciences & pedagogy
Cognitive sciences & pedagogy
Cognitive sciences & pedagogy
