What insights might be gleaned from an education platform that’s entirely online? Plenty. In a newly published paper on the preprint server Arxiv.org (“Tracking Behavioral Patterns among Students in an Online Educational System“), a trio of scientists at the University of Copenhagen investigated an AI model that ingested data from Clio Online, one of the largest content providers for primary schools in Denmark, to identify relationships among factors like time of day, subject, activity type, complexity, and performance. They say their method allowed for tracking changes in behavior among students over time, as well as trends in the broader educational system.

“How students behave … is an important topic in educational data mining. Knowledge of this behavior in an educational system can help us understand how students learn and help guide the development for optimal learning based on actual use,” wrote the coauthors. “In this paper, we [considered] how … student activity in the system affect[ed] performance.”

As the team explained, Clio Online — which counts 90% of all primary schools in Denmark as customers — hosts modules covering all subjects (except math) and comprising texts, videos, sound clips, exercises, quizzes, and tests students use to self-study and complete school assignments. The researchers looked at a slice of available data for 14,810 students, extracting features like activity during school versus non-school hours, time spent completing exercises and reading texts, time spent working with different topics, average session length, and average quiz score.

The researchers fed the compiled corpus into a cluster analysis algorithm that divvied the features up so that those in each group shared more in common with each other than with those in other groups. Interestingly, the results showed that students working with science subjects spent a disproportionate amount of time reading and that the academic performance of those working mostly with language subjects correlated with the amount of time they spent taking quizzes and working during school hours. Additionally, the researchers found that students who worked mostly from home weren’t the top performers within their peer groups.

“One notable conclusion [from our work] is that students using the Clio Online system during non-school hours … do not seem to gain any significant boost to performance. We also saw how taking quizzes seems to increase the performance of students in languages, more so than in other subjects, where reading texts are of more importance,” wrote the researchers. “This fits the intuition that skills such as grammar need to be trained, in order to be learned.”

The researchers leave to future work incorporating additional data, like whether texts or quizzes are assigned together and whether students read on their own or with classmates. In the future, they hope to deploy a production system that can be used to track class-wide changes over time, so that teachers can tailor lesson plans to individual students.