Mastery inspires awe. In the arts, sports and in the workplace, masters and expert practitioners are exceptions, outliers. If aware, intellectually, of the science of expertise, we yet prefer, emotionally and psychologically, the mystique of mastery.

For instance, when my son sits down to play Rachmaninoff, his grandparents tear up in awe of their extraordinary grandson. And he is extraordinary, especially if you accept that no one is ordinary. Mostly, he has developed a daily habit of rising at 5:30 a.m. to practice and receives guidance from a private tutor on a weekly basis. He is no prodigy, if that means that he’s endowed with innate qualities or gifts. When listening to a prodigy, we would never assume that level of proficiency would be within our grasp. It would be foolish or arrogant to think, “I could do that.” In “Pride and Prejudice,” Lady Catherine boasts of her unproven but inevitable genius for the piano, “If I had ever learnt, I should have been a great proficient.” What if she is not as foolish as she seems?

Certainly, high levels of mastery have not been the goal of corporate learning. Most would avoid claims that digital learning facilitates true mastery. And, generally speaking, innovation in digital learning over the past 20 years has led to broader distribution of instructional content; it has not optimized learning.

So, what makes good learning? Many would answer engagement, which could mean minimally interesting or even somewhat experiential. We aim for a high rate of course completion paired with high marks for engagement. The types of business challenges successfully addressed through such a learning model are not insignificant, but they are rarely transformative. Transformation calls for something else, something more than minimally prepared knowledge workers. It calls for mastery and higher levels of expert performance.

The What and Why of Adaptive Learning

The goal of adaptive learning, in contrast to that of other digital learning solutions, is not subjectively defined engagement; at its best, that goal becomes deliberate practice. It takes effort, specific guidance and spaced repetition. Decades of research evidences the efficacy of one-on-one tutoring. Yet, for the most part, digital learning is modeled on classrooms and textbooks. Adaptive learning, in contrast, recognizes the individuality of learner needs.

One way to better understand adaptive learning, at least how the term is used today within the context of learning and development (L&D), is to note what it is not:

    • It isn’t a solution that accommodates the largely debunked concept of “learning styles.”
    • It isn’t a strategy for increasing the accessibility of learning (for those with sensory impairments).
    • And it isn’t synonymous with personalization, as there are several ways to personalize a learning experience that may not align directly to the goal of individual mastery. For example, showing or hiding content depending on the learner’s role may be described as personalization, but unless that showing or hiding is triggered by learner performance (and not just by the learner profile), it isn’t adaptive.

What does constitute adaptive learning, on the other hand, is being shaped by two key factors — learning science and artificial intelligence (AI). Today’s adaptive learning, within the most sophisticated educational technology (edtech), automatically calibrates to the needs of individuals, recreating at scale the efficacy of a tutor using real-time data analysis. Consider three types of adaptivity in learning:

    • Curriculum adaptivity matches a learner’s needs to available offerings. This automates the recommendations provided by the system.
    • Adaptivity of the learning sequence within a course enables dynamic pathing based on learner performance.
    • Finally, adaptivity of the practice and learning activities ensures that learners receive the content that will be most effective in building their individual proficiency.

AI-enhanced learning has been likened to the smart recommendations of video streaming services. This comparison really doesn’t explain adaptive learning. Netflix, after all, isn’t giving users an individual cut. An adaptive movie might, hypothetically, deliver a unique combination of car chases, sword fights, love scenes, musical interludes, etc. A user might have a very long or very short movie, based on detailed analytics of response cues.

A better metaphor may be a GPS or self-driving car. We can use a GPS to help us identify a destination, but its unique value is helping with the route, which we could liken to learning sequence, and with the changing conditions, which we can liken to the adaptivity of the practice and learning activities.

eLearning and Adaptive Learning

Most eLearning offers roughly the same experience for all learners. A contrast to adaptive learning begins to suggest a new benchmark.

    • eLearning targets the average user with a standard course design while adaptive learning creates a unique learner experience for each individual.
    • eLearning is often a linear, content-first approach while adaptive learning can be non-linear with smart practice in place of content presentation.
    • eLearning assessment strategies often include quick checks and a final (summative) assessment; adaptive learning uses learner-specific practice (formative assessment).

Learners come with relatively low expectations for eLearning. They expect to review information, click through content, answer a question or two, and receive their completion. When they encounter an adaptive solution for the first time, they may feel like they’re being tested. They may be worried about failing, or they may even be frustrated by the fact that adaptive learning requires more effort and attention. The shift requires effective change management. Learners need to experience an increased reward (proficiency and retention) before giving up the ease and convenience of the informational slides that they click through to completion.

Applied Research

The research behind adaptive learning arguably predates the affordances of AI to enable it at scale. AI isn’t needed to create adaptivity, but it is needed to achieve scale. True learning adaptivity, with or without technology, factors in a complex set of learning variables. The research into these variables is diverse and spans decades. It goes beyond Bloom’s taxonomy to include his research into the efficacy of one-on-one tutoring over classroom instruction. It is informed by the psychological states in learning progression, the zone of proximal development, the value of interleaving, the criticality of self-efficacy and the inverse logic of the Dunning-Kruger effect. It embraces the science of expertise and the science of individuality.

With adaptive learning, the ability to target individual learners supports the best of learning science. It should also help us to evaluate and measure effectiveness. The data from standard eLearning consists of completion and maybe a score. Mature adaptive learning provides detailed, micro-level analytics:

    • Rather than duration or seat-time of a learning experience, adaptive learning can measure time-to-proficiency.
    • Instead of ambiguously targeting engagement, an adaptive learning strategy measures it, capturing where learners put their focus, how much effort it takes, how confident they feel about their performance, etc.
    • In addition to gathering learner perception of the content, an adaptive learning platform provides content analytics indicating accuracy, difficulty, etc.

Learning Engineering and Data-informed Design

This type of information enables an entirely new way of approaching learning design. Rather than letting courses die a slow death, as we post them to the learning management system (LMS) and move on, the content analytics make possible an iterative, data-driven design. When content is rapidly developed, a minimally viable module provides data on the content and on the learner experience, giving insight for real-time improvement.

The sub-specialization that focuses on using learning and content analytics to inform their design process is referred to as learning engineering. But, before audibly groaning about having yet another term for a training professional, consider the term’s origins in cognitive psychology. Next-generation adaptive learning affords opportunities to apply this broader body of research that is more interdisciplinary and dynamic. Yes, it further disrupts the industry’s collective commitment to stagnated but performative models like ADDIE and other, still-popular norms of instructional design, but it also legitimizes and refreshes the work. We don’t need to envy or mimic advances in marketing automation — personalization within the context of learning and powered by AI-driven adaptive learning becomes not a tool for increasing page views but one for predicting the type of unique practice, rehearsal, guidance and pathing to achieve mastery.

Closing Thoughts

The next wave of adaptive learning solutions will be even more ambitious, integrating a variety of additional approaches to support knowledge, skills, character and meta-learning. While the promise of advancing our various business interests is compelling, the possibility of advancing that most central element of our humanity, the capacity to learn, instills a depth of purpose that is perpetually fresh, resilient to inevitable setbacks.