If you know the training world at all, you have almost certainly heard of adaptive learning. Its promise? Training that adapts in real time to each learner’s needs, to give them the most effective learning experience possible.
Appealing, isn’t it?
So: an unavoidable evolution of digital learning, a passing fashion, a utopia, or a reality today or tomorrow? Read on for a clear view
In their report on the digital transformation of continuing professional training (March 2017), Nicolas Amar and Anne Burstin, members of the French general inspectorate of social affairs, define adaptive learning as follows:
“This method consists in adapting the learning to the learner’s profile, in a logic of ultra-personalisation. Drawing on big data, cognitive science and algorithms, it adjusts the learning scenario to the individual learner’s needs continuously throughout the course, taking into account the number of clicks, the individual’s preference for one tool or another (video, written or audio material, subsequently offered in preference) or the difficulties encountered on the first questions. Adaptive learning also helps maintain the learner’s motivation and attention through suitable learning arrangements or through digital (chatbot) or human support.”
In traditional digital training, every learner is offered the same route to a learning objective. Adaptive learning promises personalised training that adapts to each learner — their level, their experience, their pace, their preferred ways of learning — to offer them a tailored progression, tailored content and tailored activities. The idea is to reach their objectives as quickly and as effectively as possible.
Adaptive learning therefore personalises some of the fundamentals of learning:
Adaptive learning rests on the combination of 3 key components:
the data available about the learner, collected before and during the course,
to understand how the brain works when learning,
complex algorithms able to analyse and exploit the data collected, in order to make predictions and generate content
The visible mechanics are fairly simple: the user usually takes a positioning test, so their existing knowledge can be assessed as precisely as possible and what remains to be learned can be identified.
Based on the learner’s results, the learning management system (LMS) presents only the sequences best suited to their needs. On the design side, this means the content has to be broken down finely enough for the LMS to work as a decision tree with many branches. The finer the branching, the more personalisation is possible.
The promise of micro adaptive learning is more spectacular still, with granularity in the content itself. Content would be created to match each person’s needs and way of learning as closely as possible. On the design side, the work would be titanic: an enormous tree structure (to cover a wide range of topics) and a gigantic volume of content (to explain each topic in a way personalised to every individual). Let us be honest — that work is close to impossible for a human being. But not for an artificial intelligence.
Are we already at the point of having an AI create the content? As far as we know, micro adaptive learning is still in its infancy — but AI is advancing so fast that micro adaptive learning will very probably arrive in a more or less near future.
For adaptive learning to work in its full version, a great deal of data about the learner has to be collected (age, number of clicks, response times, time connected, quiz success rates, courses taken and so on).
Based on that data and its analysis, the AI builds a detailed profile of the learner and identifies the points to work on first in order to reach the objective.
The AI then creates a bespoke programme, generating both the content and the best format for working through it, according to the learner’s profile.
The more data there is, the better the AI can read the learner and offer the most relevant and effective course.
As you would expect, the benefits on offer are many, and significant.
No more learners clicking through screens out of boredom, or feeling swamped by content beyond their reach. With a tailored progression, activities pitched at their level and media that suit their preferred way of learning (reading, audio, video and so on), attention and engagement both go up.
By adapting to how someone learns, and by personalising their pathway so the right information arrives in the best way at the right moment, with as many repetitions as needed, adaptive learning makes retention easier and helps learners embed what they have studied for the long term.
By taking account of the learner’s experience, their skills and their real needs, adaptive learning makes for more effective learning and, for some, a real saving of time. No need to dwell on what the learner already knows: they can concentrate solely on what they need to take in or consolidate.
Better efficiency and retention also mean, for companies, better trained people, trained faster. An investment worth making.
Micro adaptive learning is in its infancy and has to grow alongside AI. It still needs a few years of development before it can show its full power and deliver on its promises.
Macro adaptive learning, on the other hand, has been with us for years. AI complements and partly automates what instructional designers were already doing very well. It is the familiar: “if you get questions 1, 3 and 7 wrong, then you need to study sequence A. If you get questions 2, 4 and 6 right, then sequence B counts as acquired.”
Adaptive learning is perfect for some disciplines — those that rest on memorisation or on acquiring automatic responses — but it is, for now, less suited to subjects that are harder to break into sequences, such as those aimed at building skills in argument, reflection, or spoken and written expression.
The importance of social interaction in learning needs no demonstration. That is one of adaptive learning’s main limits: effective training, certainly, but dehumanised. Which is why it is essential to place adaptive learning within a broader approach that brings people back in. The parallel here is blended learning (combining distance and classroom training), which is far more effective than elearning alone.
Adaptive learning is built on collecting a great deal of data. Is there a risk that this data will be used outside the training context? What does it reveal about our private lives? What could someone with bad intentions do with it?
Questions about data protection and privacy can and should be asked.
Adaptive learning is full of fine promises which, with the rapid development of AI, are on the way to being kept — in a future near enough to be worth paying attention to now.
Adaptive learning should be seen as a tool, not an end in itself: to get the most from it, use it with a clear view of its strengths and its limits, and take care to combine it with formats that bring people back into the picture, such as blended learning.