The idea is that you can test users and, based on their answers, automatically assign them digital learning content that addresses exactly the things they don't understand. No more, no less. The benefits are enormous; users don't have to spend time relating to information they already know and can focus on learning new things all the time.
I first encountered adaptive learning in 2003 in London, where I attended a demonstration of it in connection with a product called Aspen LCMS from Sumtotal Systems. I saw an e-learning module that started with a test, a so-called 'pre-assessment test', that was supposed to uncover the gaps in the recipient's knowledge. After the test, you moved on to the part of the module that conveyed knowledge. The areas you already had a handle on were left out. Finally, you were presented with a final test or exam, where you as the designer had the opportunity to choose whether the users should only answer questions regarding the content they had just reviewed, or whether it should cover all the content in the module.
Potentially a really smart and time-saving way to learn.
That raises a good question. If the technology has been available since at least 2003, why isn't it more widespread?
Let's look for some answers. Adaptive learning consists of two elements: testing and dissemination.
The test part can be a simple test that asks users a series of questions in order to find any gaps in their knowledge. It can also be an adaptive test that continuously adjusts which questions are asked based on the answers given. Adaptive testing is used, for example, in connection with national tests in primary schools. The tests are designed so that the difficulty of the questions is increased if the questions are answered correctly, and lowered if the answers are incorrect. In theory, it should be possible to relatively accurately identify what a student can and cannot do. In practice, this is a different matter, as can be seen from the recent criticism of the reading test in the national tests in primary schools.
Let's assume that it is possible to develop an adaptive test that can uncover any gaps in the user's knowledge. The idea is that the software will then automatically make recommendations on what material the user should review.
Where I think the chain breaks down is in the notion that, based on a test, you can significantly reduce the amount of material that is communicated and do so in a way that makes sense to the recipient.
We humans are not learning machines that can simply be fed information about the parts of a topic we don't understand. We need the context to understand the content.
As I recall, the content of the e-learning module in the module I saw in London 2003 was quite “demo-like” and not a real course. A bit like what you see in layout, where the text often consists of “lorem ipsum” filler text. Adaptive learning works great as a concept, where the examples used are dummy data. But it fails when it hits reality, and real content needs to be on the field.
In principle, one could well imagine that one could design smaller pieces of information that could be served to users who have gaps in their knowledge, but in practice it is difficult and time-consuming to design learning material that precisely delivers the necessary information - and nothing else.
Danish Area9 has apparently had success with adaptive learning. In any case, they have been sold to the American McGraw-Hill for Education for an estimated 1.5 billion. In January 2018, the state's growth fund invested 180 million in the subsidiary Area9 Lyceum, and they are generally pretty hyped with quite a lot of media coverage. I am of course curious to see if they will achieve the goal of making good adaptive learning, so I have visited their website and looked at two different courses, both of which use their software and method. A course that deals with Corporate Social Responsibility (CSR) with two sub-topics; business and human rights, and anti-corruption, and an app where you can learn something about wine.
I must say that it does not change my overall impression of adaptive learning: It sounds good in theory, but the practice is lacking. I think that you spend a relatively long time answering the questions, some of which are quite bad: unclear or too easy. If you answer the questions incorrectly, you are asked to review some material, typically some boring static slides with some definitions on them, after which you can move on. At some point you get the same question again, and if you had not learned the answer by looking at the boring slides, you can probably remember that you answered incorrectly the first time and refrain from ticking the wrong answer options. But if you answer incorrectly again, it is the SAME information as the first time(!), which is supposed to pull you out of your ignorance. So there is not much adaptive about the presentation.
If you were presented with this kind of content as regular e-learning, you would die of boredom. Of course, you can say that if that is the level of ambition on the content side, then it is of course good that you are not presented with more content than is absolutely necessary.
To be fair, I think the app where you could learn something about wine was somewhat better than the course on CSR, although I still don't think the wine app meets the high expectations that are set for adaptive learning. What is better about the wine app is not the adaptive element, by the way, but that the graphics are more inviting and the questions are better.
To give you an idea of the communication part of the wine app, I will describe a typical content slide for you.
It's about the Zinfandel grape, so it says Zinfandel in the headline. There's a photo of a typical wine made from the Zinfandel grape, and then the text says that Zinfandel wine is full bodied, fruit forward, high acid, and medium tannin. That's it. So there's not much meat to it.
Area9 has also updated their website so that the examples mentioned are no longer available. The website does of course have cases where they show that users are highly satisfied with the material, but I would have liked to see a real scientific comparison between their adaptive, boring content and a non-adaptive e-learning module where instead of spending the money on licenses for Area9, they had spent the money on creating some engaging content.
My conclusion regarding adaptive learning for now is that, not surprisingly, it is not automatically good just because it is adaptive. It still requires well-thought-out, well-presented, relevant content for people to learn anything. And adaptive learning sounds better in theory than it does in practice. I have yet to see a solution that actually works with this approach when we get beyond very basic skills training (math, grammar, etc.).
I'd love to hear about good examples of good adaptive learning if you know of any, I just haven't seen it yet.
Do you want to? learn more?
If you want to know more about digital learning and e-learning, you can start with our E-learning FAQ
Here are some articles about different types of digital learning that you might find interesting.
- What is elearning and digital learning?
- What programs should you use to create elearning?
- Does gamification work?
- What is microlearning?
If you are interested in reading more about motivation and learning, these articles may interest you.
- Self-determination theory. The most important theory you need to know about learning.
- Coercive design – how not to design e-learning .
- Which learning methods best support motivation for learning?
- Your brain is NOT a computer - Predictive Coding
- Flick 2 learn. Why Interactive elearning is NOT always exciting elearning
... and are you looking for help developing exciting e-learning , or would you like to learn how to create exciting e-learning yourself - then give us a shout.