Supporting students in HEIs

This time, last year we were frantically writing a paper on how a Higher Education Institution (HEI) could use the rich data that collects during students’ academic lifetime in order to identify and support those students who struggle and may be potentially at risk of dropping out from their studies (Chounta,, 2020). Dropouts in Higher Education is hardly a new topic – research has extensively studied reasons behind dropouts (Tinto, 2017), computational algorithms for early detection of students-at-risk (Barber, R., & Sharkey, M., 2012), teacher and student-facing dashboards for support students at-risk (Arnold, K. E., & Pistilli, M. D., 2012). Our work aimed at visiting existing practices and adapting them into context: in other words, to design a machine-learning model that would identify students who could be at risk in the University of Tartu. The paper appeared in the Companion Proceedings of the 10th International Learning Analytics and Knowledge Conference (LAK20) and can be downloaded here.

Once we had that, the question was “What do we do with it?”. So, we embarked on this adventure of designing a dashboard that would communicate the information concerning students who may be struggling with their studies to HEI study specialists. That is to the academic stakeholders who make sure that students fulfill the requirements for acquiring a study degree and who are there to discuss students’ issues, provide solutions (to the extent possible), and support. To do so, we carried out a series of design-workshops with curriculum developers, study specialists, and career counselors of the university as well as discussions with student representatives. These discussions resulted in a set of requirements that drove the design of the dashboard and a set of guidelines regarding the feedback interventions that we could provide (these will be discussed in follow-up publications).

Last summer, Timo Soiunen, our master student at the Computer Science Institute undertook the task of designing the dashboard interface, and Hasan Tanvir, also our master student from the Computer Science Institute undertook the design of the machine-learning model that would support the dashboard. We’ve been busy bees 🙂 While designing the dashboard interface and the model, we did that acknowledging that we do not to predict students-at-risk. We aim to provide awareness to students who may be struggling with their studies and to enable them in seeking support.

Today, we were happy and proud to demonstrate the results of our efforts to the program directors meeting of the University of Tartu, hosted by the Vice-Rector of Academic Affairs, Dr. Aune Valk. There, we showcased the connection between the model and the dashboard, the different views of the dashboard interface, and the feedback interventions we can trigger via the graphical interface. We also discussed the ethical aspects regarding the use of data in this context, how we can ensure fairness, transparency, and accountability, and we received valuable feedback for the future.

Future? Yes! Our next goal is to work with the Study Information System team and integrate the dashboard and student model in the ÕIS2 dashboard. The plan is to be ready for large-scale testing in June. This means: busy bees back on the task!

Arnold, K. E., & Pistilli, M. D. (2012). Course signals at Purdue: Using learning analytics to increase student success. Proceedings of the 2nd International Conference on Learning Analytics and Knowledge, 267–270
Barber, R., & Sharkey, M. (2012). Course correction: Using analytics to predict course success. Proceedings of the 2nd International Conference on Learning Analytics and Knowledge, 259–262
Chounta, I. A., Uiboleht, K., Roosimäe, K., Pedaste, M., Valk, A. (2020). From Data to Intervention: Predicting Students At-Risk in a Higher Education Institution. In Companion Proceedings 10th International Conference on Learning Analytics & Knowledge (LAK20).
Tinto, V. (2017). Through the eyes of students. Journal of College Student Retention: Research, Theory & Practice, 19(3), 254–269