Husker researchers developing AI-based app to boost STEM grades

(From left) Bilal Khan, Mohammad Hasan and Neeta Kantamneni | Craig Chandler, UComm
(From left) Bilal Khan, Mohammad Hasan and Neeta Kantamneni | Craig Chandler, UComm

A University of Nebraska-Lincoln computer scientist is harnessing the power of artificial intelligence to help undergraduate STEM students improve their academic performance. His project will strengthen the pipeline of college graduates prepared for STEM jobs, the number of which in the United States is expected to expand by about 10% by 2030.

With a three-year, $600,000 grant from the National Science Foundation, Mohammad Hasan is developing a machine-learning-based app, called Messages from a Future You, aimed at providing students with targeted, real-time interventions that boost their performance in STEM courses. Using the app, students can engage in dialogue with their “future self” — an avatar derived from the student’s photograph — about how to boost their grade.

The app would be the first that uses an artificial agent to provide tailored interventions that account for the myriad factors impacting a student’s final grade.

“The existing approaches mainly target academic improvement based on just academic performance alone,” said Hasan, assistant professor of big data and artificial intelligence in the Department of Electrical and Computer Engineering. “But end-of-semester performance is not just influenced by academic activities during the semester. It is shaped by other things: family background, socioeconomic status, peer interactions, interactions with the instructor, science identity and more.”

The app would be a cost-effective, portable tool to combat a national attrition rate from STEM majors that hovers around 50%, driven largely by students’ poor academic performance. From his seven years of experience teaching large introductory courses at Nebraska, Hasan had first-hand knowledge of why undergraduates may struggle early on in a STEM major.

“Helping students in large classes is not easy, because you can’t really talk to every person throughout the semester, and students would not come to you unless they are in real trouble,” he said. “And usually when they come, it’s at the very end of the semester, when it’s not really possible to meaningfully help them.”

Hasan started brainstorming ways he could boost student performance. Although institutional-level change may hold the most power to move the needle, Hasan believed there were smaller-scale, less expensive ways to help students.

He recognized the power of machine learning to power an app that would help students succeed: The app could “learn” about the student’s behavior and background, then use that data to predict future performance and provide advice for changing that trajectory.

Read more: