Here are some results of applying the adaptive learning in one of the University of California, Los Angeles courses:
Median exam scores grew from 53% to 72-80%
Average course attrition decreased four-fold (down from 43.8% to 13.4%)
Women students reached ten-fold attrition decrease ratio (down from 73.1% to 7.4%)
We will provide more detailed statistics and explanations below. For now, let’s try to define what adaptive learning is and how to use it in eLearning and corporate training.
Adaptive learning as an instructional design methodology
Slightly rephrasing Business Dictionary definition, adaptive learning is “a type of learning that concentrates on analysing the previous successes and outcomes of the training and uses these results as a basis for improving training strategies to ensure future successes”.
For example, if Big Data analysis shows that the majority of students watch some video for several times or browse for hints and explanations while watching it, this serves as a sign that this learning nugget must be revised and improved to ensure the next group of students will not face this problem.
However, according to one of eLearning industry most influential authors, Clark Quinn, adaptive learning (as the type of learning) is best used on top of a good course design. Thus said, adaptive learning methodology is not a panacea, yet it can be a powerful boost to the course performance and overall training efficiency if applied correctly.
Adaptive learning helped UCLA increase median exam scores from 53% to 80% and decrease course attrition ratio tenfold - from 73.1% to 7.4%!
Adaptive learning as a technology of educational process improvement
The other definition of adaptive learning is provided in the WIkipedia. According to it, adaptive learning is “an educational method, where computers work as interactive teaching devices, allocating educational and human resources according to the needs of individual users. The instructional content is presented in accordance to the user’s individual needs, based on their answers to questions and responses to previous tasks and experiences”.
This definition more closely aligns with the initial idea of adaptive learning, proposed by behaviorist B.F. Skinner back in 1950’s. He constructed a machine that allowed the students effectively grasp new concepts in learning while staying engaged and motivated, rather than simply memorizing the same information time and time again.
This machine’s modus operandi was the following:
The students answered the questions
If the answer was right, positive feedback and gratification was given
If the answer was wrong (a misconception), a series of small tasks (target problems) were given, helping the student find the right answer
The same approach is applied nowadays, when complex algorithms are used to monitor the student’s progress real time and adjust the content exposition (like providing additional materials) to ensure successful knowledge consumption and increased training efficiency.
The logical question emerges - are there any working examples of applying adaptive learning principles with measurable results?
There is a particularly interesting project we developed together with Christopher J. Lee, professor of Biochemistry, University of California, Los Angeles — Courselets.org, an open-source library of instruction units that can be freely combined into new courses. .
Courselets project incorporates an interactive chat interface for answering target problems, provides a GitHub-like cloud repository allowing the instructors quickly and efficiently collaborate on the course content, and even boasts the integrated system for content A/B testing.
Learner mistakes are the main blockers of knowledge comprehension. The core idea of the Courselets is that a mistake (or blindspot in comprehension) is not a flaw, but a marker of the room for improvement. More detailed explanation is provided in this great article on how to turn mistakes into your most powerful learning tools.
Below are the results of applying adaptive learning principles in practice. The statistics were gathered during a 5-year long study of one of the UCLA Computer Department courses on Bioinformatics/ Probabilistic Modeling:
Average learner had about 20 misconceptions on the course content and required around 40 target problems to identify and solve them.
Doing this verbally in 2009 yielded poor results, as answering 40 spoken questions is quite a tiresome experience.
After introducing Courcelets back in 2011, average student was successfully engaged in answering more than 60 target problems. Such approach even eliminated previously existing disparities among learner groups (bottom 50% vs top performing 10%, men vs women, etc.)
As it turned out, the most common misconceptions (blocking progress for about 30% of the students) were not even directly addressed by the course textbook.
Using Courselets technology to discover such misconceptions proved to be highly efficient, as 4 major error models covered around 70-80% of all student errors on average.
The most important results of using Courselets were described above but we will mention them once more:
Median exam scores grew from 53% to 72-80%! The biggest part of this improvement is due to the worst-performing students improving their results to the levels only the students ABOVE the median were able to score before the switch to Courselets.
Course attrition decreased four-fold as compared to lecture format of the course taught in 2004-2009, down from 48.3% to 13.4%.
Women students reached ten-fold attrition decrease ratio (down from 73.1% to 7.4%).
As you can see, correctly applying the methodology and technology of adaptive learning can lead to significant improvements in instruction quality. This is especially important for eLearning, where the educators do not have direct contact with all of their learners and have to rely on technology in identifying and overcoming the blindspots in their instructional materials.