HSE Researchers Develop Method to Improve the Quality of Online Courses

iStock

HSE researchers have developed a model that will improve the quality of online courses. Dmitry Abbakumov, Head of the Centre for Computational Educational Sciences at the HSE eLearning Office, shared the news at the annual International Meeting of the Psychometric Society 2020.

Abbakumov's research ‘combining explanatory IRT and psychological networks for understanding and modelling online learners’ difficulties’ focuses on identifying the connection between errors and learning difficulties and determining the sources of their occurrence. Modelling results allow authors of online courses to improve the quality of the developed study materials.

Dmitry Abbakumov, Head of the Centre for Computational Educational Sciences of HSE eLearning Office

Current psychometric theories consider student errors as independent random events. Psychometricians usually do not inquire how a mistake made in a particular task can be connected to other mistakes – both for a particular student and for the entire course audience. We suggest an alternative approach - to consider errors as potentially connected events, initially assuming that everything is related to everything. To determine real connection, we developed a special model, which is an extension of the Ising model. This extension allows you to filter out insignificant ties and highlight significant ones.

For digital education this discovery can become the basis for a comprehensive work on the quality of online courses. Previously the course authors offered point solutions to guide students who made a mistake, now they see errors as a systemic problem in a course. Thus, teachers and lecturers can create explanatory study materials aimed at solving a particular system problem rather than leading to the correct answer in a separate task.

‘It is important for large-scale digital education, which, in contrast to the classroom format, can’t provide the teacher with an opportunity to ask clarifying questions to all students to identify where the group ‘lost contact’. It turns out that we provide the author of the course a tool, a kind of lens that highlights the learning difficulties of thousands of online groups,’ Abbakumov says.

This is a qualitative leap in digital education

The suggested model demonstrates the ability to predict well the difficulties that students face during tests. The overall accuracy reaches 81%. The model was tested using MOOC data from the Coursera platform (185,000 students) and is currently being used in courses by HSE authors.

The Centre for Computational Educational Sciences at the HSE eLearning Office was the first in Russia to study problems of online learning effectiveness and to develop and apply special models that differ from classical psychometrics. Active digitalization of the education environment increased the necessity and urgency of developing these new models.