Computer programs may help predict students’ grades in school as well as determine successful pathways for completing assignments, finds a new Stanford University study.
In four experiments across two studies, researchers looked at the performance of 370 undergraduate and graduate students in an introductory coding class.
They examined how students approached tasks—the size and frequency of the student’s code or the change in coding methods, for example—and compared the students’ processes with their later course grades.
The researchers, led by Paulo Blikstein, an associate professor of computer science at Stanford, developed computer models based on more than 154,000 snapshots of student code taken while the students worked through their assignments.
While not all the tested variables correlated with final grades, the computer-based prediction models that focused on how students approached problem-solving proved more effective in predicting final grades than midterm test scores did. The researchers also identified common ways students accomplish assignments, which the Stanford researchers concluded could be useful for issues like offering widespread help to students struggling to complete open-ended tasks.
The project highlights potential for computer-based, large-scale evaluation of student performance, especially in areas such as project-based learning where determining students’ successes and places where they struggle is difficult or time-consuming.
The models “could potentially save a lot of time and energy for both teachers and students,” Blikstein said.
A version of this news article first appeared in the Inside School Research blog.