The Biological Approach to Adaptive Learning
Every brain is different; therefore, learning must be personalized to meet each learner’s needs. But how can this be accomplished in an e-learning environment? Our answer, developed and refined over the past two decades, is a biological approach to adaptive learning.
The biological model differs significantly from the more traditional “inference models.” Until recently, the inference construct has been the primary model for the more advanced variants of e-learning, often delivering less than satisfactory results.
Inference models take an “engineering” approach. They assume it’s possible to build an interconnected network of learning objectives based on a series of events, essentially drawing “arrows” between them. When something goes wrong—i.e., a learner makes mistakes that show a lack of comprehension of a certain learning objective—the system backtracks through the “arrows” to find the last place where the learner comprehended the material.
In our view, however, learning does not happen this way, and even the more sophisticated, complex inference-type models have significant short comings. In fact, we believe learning occurs in a far more chaotic way. That’s why, based on our expertise and research in brain function and how learning actually occurs, we decided very early on to abandon the inference construct and, instead, to develop a biological model.
The biological model embraces the fact that there are countless ways to accomplish a series of learning objectives. Of course, there must be a general theme around what needs to be learned. But very often, what the learner grasps first—whether it’s one learning objective or another—makes no difference at all. For some learners, it appears learning is better promoted by skipping ahead to something they cannot grasp at all, then reverting to something they more easily comprehend, as they work toward what they struggled to understand in the first place. Others learn from a more linear approach, one step at a time, systematically building on top of pre-existing knowledge. We think that choice and ability to handle both extremes and everything in between is key to adapting learning sufficiently and effectively.
What matters most in providing the needed support for vastly different preferences and personalized paths is understanding how each individual progresses through his or her learning, while keeping track of what has been learned and the content that is most closely related. For the latter, our biological model utilizes “zones of proximity”—that is, those things that are somewhat related. Proximity provides insights, based on each learner’s specific experiences and results, into where he or she might struggle. As we have found, when weaknesses are detected, there is a higher likelihood of weaknesses occurring within related areas or zones of proximity.
Think of it as Amazon’s “other customers also bought this” promotion, which relies primarily on statistics and datamining. Similarly, biological models in learning have both a strong semantic structure and massive data to support these predictions.
Developing a biological approach to adaptive learning has excited us. Drawing from our backgrounds in human and medical sciences, including neuroscience, we have embraced the challenge of building systems that are “plastic” enough to accommodate differences among learners’ brains.