A compelling plea for improved support for students, particularly in their initial semesters, is underscored by the fact that STEM college students in the U.S. graduate at a rate approximately 20% lower than their non-STEM counterparts. While there is a recognized need for substantial, enduring changes—ranging from advocating for evidence-based teaching methods—the entrenched nature of academia can sometimes hinder their implementation.
Certain scholars and specialists have taken steps to nudge these pivotal transformations forward. Interventions designed to aid struggling students in navigating their academic journey represent a promising approach.
Significance
Mohammad Hasan and former Husker Bilal Khan, hailing from Nebraska, delved into the realm of machine learning—a branch of AI that discerns patterns in data to make predictive analyses. Initially, Hasan and Khan developed a model leveraging the final outcomes, survey responses, and research findings of 537 Husker students enrolled in computer science courses from 2015 to 2018.
Subsequently, they tested this model on a cohort of 65 graduates from the same course. Among them, 32 students received automated notifications via the university’s course management system at three intervals during the semester—after six days, nine weeks, and twelve weeks. These messages conveyed the model’s forecast regarding the students’ likelihood of passing the course, categorized as “Good,” “Fair,” “Prone to Risk,” or “At Chance.” Conversely, the remaining 33 students in the control group consistently received the message “Unwilling to Make a Prediction.”
Of the 33 students in the control group, 24, or roughly 73%, successfully completed the course. In stark contrast, those who received personalized projections based on their performance trajectories exhibited significantly higher success rates: 29 out of the 32 students passed, equating to an impressive rate of nearly 91%. Moreover, 86% of the surveyed students who reported vigilantly monitoring their progress admitted to intensifying their efforts after being apprised of the forecasts.
Future Prospects
Although preliminary, the findings underscore the value of integrating AI-driven initiatives into STEM education, as posited by Hasan and Khan. The duo aims to conduct a more extensive, longitudinal study to explore whether factors beyond academic scores—such as student behaviors within courses, attitudes towards science, and demographic variables—could facilitate the broader application of such interventions across diverse courses.
