I had the opportunity to participate as a subject matter expert and contributor in the creation of the “Learning to Adapt” white paper. As more institutions, Higher Education or K12, adopt adaptive learning technologies and systems this series can be a powerful and informative starting point. The report (one of two parts), while not a complete census, gives readers a great foundation for the discussions with providers.
Providing institutional stakeholders with frameworks and resources to facilitate conversations regarding evaluation and adoption of adaptive learning solutions. This differentiated analysis is especially important because the range of solutions provided by adaptive learning suppliers varies substantially in appropriateness for different institutional environments and implementation scenarios.
In addition to working with a fantastic team, we also had the opportunity to interview a vast cross-section of providers. It was interesting how adaptivity and personalized learning tended to overlap and in some cases were synonymous with the various providers. Fundamentally the pillars of adaptive systems were being executed at varying levels of maturity but all demonstrated that learner adaptivity is made up of learner, progression or content adaptivity.
Though focused on higher education application, easily systematized courses fair best as candidates for adaptivity. Small cohort courses with a high degree of educator/mentor engagement wouldn't likely be a fit. Courses with high n-counts, delegated delivery or even large-scale virtual programs (MOOCs) can apply adaptive learning assuming:
- Providers and/or institutions have identified the scope of the course or even how it fits into a larger curriculum.
- Course content is then deconstructed to lower (ideally lowest) levels of granular addressability relative to learning objectives. These are typically the feature vectors (key derived data points) for an adaptive algorithm.
- Progressions or concept maps link these objectives with multiple pathways. This is different than the structure that is typically encountered via a syllabus or curriculum map. Learner profiles are then applied and adapted based on their individual pathways. These can be seeded in advance as points of abstraction (starter profiles or archetypes) based on the normal or most repeatable path through the materials.
- The provider or institution should have multiple learning objects to draw from for each learning objective. This is what the algorithm needs to pick a learner-appropriate approach, try alternatives or even scaffold the learner to a more difficult item, object or conceptual approach.
- While adaptive learning approaches are innovative they aren’t the only way to ensure learning is happening. If we consider these traditional approaches as the baseline and then apply the process relative to a large pool of learners, monitor and record their paths through the progressions, concept maps and outcomes this will offer continuous updates to the original models.
This is what educators have been doing since Socrates.
Adaptive systems can apply Bayesian or other data mining/statistical methods to segment the learner population, making recommendations and predictions about paths and learning objects based on prior efficacy for each new student.
While I participated on the larger of the two documents, I recommend them as a set. The first providing a great introduction and the second as a look into the adaptive learning marketplace.
Check out the report here: Learning to Adapt: Understanding the Adaptive Learning Supplier Landscape