Adaptive Learning Platforms
This month in Edtech, we’re delving into the world of Adaptive Learning Platforms (ALP). United States research and development firm Gartner predicts the ‘benefit rating’ of adaptive learning platforms as “transformational” — the highest benefit rating that any new emerging education technology can receive. According to Gartner, ALP will enable new ways of doing business across the education industry, and will result in major shifts in industry dynamics.
A Brief History — Intelligent Tutoring Systems
The development of modern adaptive learning platforms can be traced back to the intelligent tutoring systems developed in the aftermath of WW2. Famous codebreaker, mathematician and computer scientist Alan Turing outlined the now-famous ‘Turing test’, a hypothetical in which a human being poses the same question to two agents; one, a human being, and the other, a computer, and receives two answers back. If the person cannot deduce which of the answers has come from the computer, then the computer has passed the Turing test. This theoretical approach to computer development effectively shaped the development of intelligent tutoring systems for the last fifty years, with early implementations in the education sector being delivered through “programmed instruction”, and the general emergence of “Computer-Assisted Instruction’ as a field of computer science in its own right.
Basic components govern the development of adaptive learning platforms, and it’s not too hard to grasp if you break it down to four basic stages. Here’s how it works.
First, the domain model. Think of it as containing all the possible correct answers that the training subject recognises.
Second, the student model, which traces the student’s interactions with the course through the computer, superimposes those interactions over the domain model, and then tracks all the differences between the student model and the domain model. These differences reflect the gaps in the student’s knowledge and ability that still need to be taught.
Third, and most interestingly, is the tutor model, which receives the information gained about the student’s knowledge gaps and then makes an active choice about what tutoring strategy will most effectively bridge those gaps. Remarkably, every time the student is able to successfully correct a past mistake, the adaptive learning platform will log the learning strategy, how successful it was, and whether it should be reapplied for the next learning outcome. By doing so, the computer begins to create a profile of the student; what she knows, how she learns, what works and what doesn’t work in her learning development.
Fourth is the user interface model, the component that integrates these three different processes so that they can best facilitate a progressive dialogue, using computer knowledge gleaned from large data-sets to develop meaningful communications to the student. Think of this last model as the magic ingredient to create systems that might just be able to pass Turing’s famous test; not just in having the information about how to teach a student, but also being able to communicate that information in a meaningful and articulate manner.
The efficiency that a schema like that can provide should be immediately obvious. Conventional classrooms — involving groups of students all absorbing the same curriculum delivered by the same teachers — can often only progress at the rate of the slowest learner. Those ahead of the class are forced to effectively stall on already mastered material. Those behind risk getting left behind altogether. One-on-one interactions between learner and computer potentially cut-out the wastage of double-learning while eliminating the risk of students dropping out or graduating without the required skill-set for the qualification. Transformational indeed.
Present and Future Applications
Companies such as Acrobatiq have already developed systems such as the Learning Dashboard and the Smart Author that adopt the principles discussed above. It’s only a matter of time before more follow suit. But there are still many questions to be answered before this technology can be properly introduced into the mainstream.
Making the transformation to truly adaptive learning programmes will not be an easy process. Developing the course content and adaptive programming is expensive and complicated. Developers need to focus on writing sophisticated algorithms, and thoroughly test how different data affects the algorithm. Educators need to find solid and peer-reviewed learning theories underlying any adaptive algorithm that is implemented. At present, scholars disagree on whether adaptive learning programs can ever replace, or even supplement, one-on-one human interaction for some subjects. Testing has revealed that adaptive learning is especially powerful in teaching mathematical or scientific studies, but it remains to be seen whether it will find application in humanitarian disciplines such as English or History.
It will also be difficult to reconcile modern privacy law with the new kinds of data that adaptive learning programs can capture. The corporate world will have an interest in obtaining data about how their potential customers think, and there are serious ethical issues in how this information is stored, transferred and commoditised in the 21st century. Policymakers will need to consider exactly what data can be collected and what should be destroyed.
These quibbles aside, there is a clear pathway forward now for developing an integrated adaptive learning platform into the modern education landscape. Gartner recommends that the Chief Information Officers of education and technology providers include adaptive learning regimes into all future Requests for Proposals. This allows a greater range of potential bidders for future projects and should necessarily involve faculty and instructors in the process. Adaptive learning is predicted to decrease the drop-out rate of students, both by increasing student engagement and by decreasing the timeframe for students to achieve a qualification. The latter of these is expected to result in an overall lower cost for the student, which in turn increases access to education for society generally.
