Combining the art of teaching with learning science
Adaptive learning is getting smarter thanks in part to powerful data analysis and software engineering. But adaptive learning works best when it combines statistical modeling techniques with something much more fundamental — good instructional design and what we know about how students learn.
In other words, today’s adaptive learning works best by blending computer science with learning science and, by extension, good teaching. These are some of the basic principles driving today’s education technology.
- Course design should begin with measurable, student-centered learning objectives and align those with content, assessment, and activities.
- Goal-directed practice and targeted, timely feedback are critical conditions for effective learning.
- Learning mastery involves developing component skills and knowledge and then synthesizing and applying them appropriately.
- Self-directed learning requires that students monitor, evaluate, and adjust their approaches
What’s under the hood?
The merger of breakthrough data science techniques and learning science to produce a smarter, adaptive learning methodology is in part based on 12 years of research from the Open Learning Initiative at Carnegie Mellon University. The goal of OLI was to develop scientifically based, classroom-tested online course materials that helped enact instruction online in the absence of a formal instructor.
OLI showed that adaptive learning, when it is informed by sound instructional design, works best when it leverages powerful statistical computational models that combine data from practice activities. By analyzing a steady stream of student activity data, adaptive learning software can infer whether or not a student is progressing toward, or likely to achieve, a particular learning objective. Based on this continual and dynamic analysis, the software personalizes the practice activities for each individual learner.
This type of real-time statistical modeling of individual learning performance produces highly accurate estimates about student learning progress against a set of outcomes. The resulting predictive analytics function is the game-changer in today’s adaptive learning. Periodic quizzes and assessments provide a rear-view look at what has already happened, while predictive learning analytics enables advanced insight into what’s likely to happen in the future.
Beyond reports to actionable insight
Lots of elearning products, particularly learning management systems, have robust reporting functions. They take categories that instructors establish, and they filter raw data through those categories to give an informative synthesis about what students have done in the course. With these reports in hand, faculty can start to draw their own inferences to apply and test next semester. The data produced by today’s smart adaptive learning technology, on the other hand, can generate inferences that lead to immediate interventions — interventions that may have the direct effect of keeping a student from failing or withdrawing from a course.
After monitoring a learning dashboard that showed student progress on activities, measured against learning estimates and learning objectives, the professor realized the students mostly understood the basic material that his next lecture was going to cover but were struggling with other concepts. He needed to plan a different lecture, Jerome reports the professor saying, and “to come up with better examples in his slides.”
Formative assessments drove instructional improvements
The really fascinating change is what came next: The professor shared the analytics with his class to explain why he was adjusting the lecture. After that, he said, he had “never seen so many students in such a large lecture actually paying attention and being engaged.”
In this case, students not only know how they did on an assessment, but also how they are doing relative to what’s important — the intended outcomes of the course. They are getting timely feedback on what the course design has established as the priority.
In short, analytics extracts possible insights from the raw data rather than just leaving instructors with a pile of reports to grapple with, probably in some future semester. Nor does it leave students with a summative assessment against goals they never fully absorbed. It helps students and instructors act on the data right now and realize its full value.
The learning technology adapts but, more powerfully, it enables and empowers the adaptability we hope to see among students and faculty in the ideal learning environment.