What is adaptive learning?

What is adaptive learning?

And how can adaptive technology create personalized learning experiences for students?

Adaptive  learning is not a new concept, but its implementation is now finally  becoming a reality and those in the education sector should know what it  is so they can be ready to benefit from it.

So what is adaptive learning?

Adaptive  learning is the use of technology to create a personalized learning  experience for each individual student — essentially replicating the  role of a personal tutor.

Let’s break down what we mean by, “the role of a personal tutor.” There are four aspects to a personal tutor’s role:

  1. To be aware of all the concepts within an area of learning and understand how they are linked
  2. To identify which concepts a learner has already mastered
  3. To construct an ordered learning path of the concepts the learner has yet to master
  4. To support the learner as they attempt to master new concepts

For  an adaptive learning system to be truly successful, each of these steps  requires automating, so let’s think about what that might look like.

Firstly,  let’s look at the knowledge of all the concepts within an area of  learning, which we will refer to as a knowledge map. A knowledge map for  an adaptive learning system must link all the concepts within a topic  area and all the topic areas within a subject. Let’s use a very basic  example to show just how complex this process is: to understand the  number 5, a learner must understand its cardinality (that it can  represent an amount e.g. five eggs), that it can be ordinal (the 5th  item in a list) and that it can be nominal (referring to a particular  object e.g. a player in a football team or a house number).

If  all of this is built into understanding just a number, imagine how much  knowledge is required to understand a complex mathematical concept like  factorisation of a quadratic equation!

What the knowledge map for the Early Fractions (blue) and Fractions (orange) looks like in Mathspace. Each node refers to a concept, e.g. node 3569 at the bottom is, “Multiply and divide mixed numbers”

Once a  knowledge map has been created, the next step in building a functioning  adaptive learning system is to create a method by which the system can  identify which concepts are already mastered and which are not.

The  way a teacher or tutor does this is to present the learner with  questions and this is the logical method for an adaptive learning system  to use too. However, the standard method for answering assessable  questions on a computer is to use multiple choice as these questions can  easily be coded as right or wrong. The problem with this method is that  it misses out on all the steps and calculations that lead to a final  answer and the learning that comes with all that working.

To  really enable a learning program to identify what concepts the learner  has mastered (and which they haven’t) the questions must be structured  in a way that allows the learner to show their thought process, to enter  their working. This process gives the learning program the ability to  assess the learner’s knowledge and understanding. It also opens up more  opportunities to provide feedback, and to help them master new concepts.

(We’ll talk a bit more about this in a minute!)

By entering working, learners can identify mistakes and self-correct. It also allows significantly more data to be collected by the system creating a better picture of the learner’s level of understanding.

If this  can be automated, then the third step is to automate the construction  of an ordered learning path that helps the learner master the remaining  concepts.

To  do this the system must combine the information it holds from both the  knowledge map, and the data on the learner’s current level of  understanding. It is then a simple process of closing gaps in an order  from most basic to most complex concept — an ordered learning path.

To  clarify what we mean by an ordered learning path, we like to use the  analogy of building a house. You can’t start decorating a house until it  has a roof on it; you can’t put the roof of a house before building the  walls; you can’t build the walls of the house without first laying the  foundations. Just as building a house must start with laying the  foundations, complex mathematical concepts require you to first master  all the simpler concepts on which they build.

Once  we have constructed an ordered learning path for the learner, we then  need to help them to master new concepts. Clearly, mastering new  concepts requires more than just understanding any prerequisite  concepts; it requires teaching of the new concept and practice applying  the concept to consolidate understanding. Answering questions, even when  showing working, is not enough. A personal tutor does not simply sit by  whilst a learner practices, they monitor the learner, let them get  stuck and puzzle it out when they can and intervene when required. Their  intervention can be in the form of pointing out where a student has  gone wrong, helping them understand what the next step might be or even  showing them the next step. This is the final process that must be  automated for an adaptive learning system to successfully replicate the  role of a personal tutor.

Whilst  the other steps are challenging, it is this last step that is the most  difficult. It requires an awful lot of work. Each and every question  relating to every reasonable concept must be considered and every method  of solving that question must be identified. Relevant question-level  feedback, such as hints and next steps, must be produced for all the  different ways a learner may use, so that when a learner gets stuck and  requires help, the system can provide the most appropriate feedback.  With the advancement of artificial intelligence, no doubt this aspect of  the process will also be automated in the future. But for now it still  requires a team of hard-working humans. (You can see how our team of  teachers at Mathspace achieves this in this video.)

This learner’s work on Mathspace is showing where they have entered correct steps (yellow ticks) and incorrect steps (grey cross), asked for hints (blue question marks), skipped steps (red arrow) and reached the final answer (green tick).

At Mathspace,  we refer to this question-level feedback as the inner loop, it forms  the first layer of the adaptive system, helping learners to solve  individual questions pertaining to a specific topic. In addition to this  inner loop, there is a middle and an outer loop. The middle loop  presents the next relevant question within  a concept to match the learner’s level of understanding — increasing or  decreasing question complexity as a learner demonstrates understanding  or a lack of it. The overarching outer loop presents the next relevant concept based on the concepts already mastered — the personalised path through the knowledge map.

Adaptive learning is in its relative infancy as an educational tool, but the huge progress that Mathspace has made in the area shows the potential that technology can offer. No  doubt adaptive learning technology will continue to advance and what’s  possible in this space will continue to develop too. It’s certainly an  exciting time to be working in education!