stem learning

Janice Gobert, a professor of educational psychology at the Rutgers Graduate School of Education, has received a four-year, $1.9 million grant from the U.S. Department of Education to contribute to key areas of innovation in STEM instruction, assessment and learning.

Gobert and Mike Sao Pedro of Apprendis, LLC will develop online simulations and artificial intelligence tools to automatically score students' performance in high school physical sciences and mathematics and to provide students with feedback. 

“American students continue to fall behind on both science and math compared to other countries,” said Gobert. “Since math is a key to understanding science, we must improve competencies at math, particularly at the high school level, for which there is more overlap between the two subjects.”

The resulting work will provide students with real-time support, driven by AI-based algorithms, needed for math related to science learning such as making predictions, graphing and writing equations. Teachers will receive real-time reports on students’ competencies, which will help them decide whether to pursue whole-class instruction, differentiated instruction or one-on-one support. To date, no assessment systems or intelligent tutoring systems use machine-learning AI to assess students’ science competencies; that is, how well science can conduct authentic experiments and engage in reasoning, as scientists do.

“Remediation should happen in high school to ensure future college and career readiness since poor math skills are a barrier to STEM college majors,” she said. “To date, the growing number of STEM jobs outpaces the number skilled people needed to fill them, so there is a major need and opportunity for students to learn the competencies more deeply.”

On the research project, the team will design, develop, and test the efficacy of AI-based scaffolding, a method to provide particular support to students in real time as they learn these science and math competencies. 

The researchers will implement the competency-based assessments and scaffolds within Inq-ITS (Inquiry Intelligent Tutoring System), a browser-based learning and assessment system envisioned by Gobert and other collaborators. As students "show what they know,” a virtual teaching agent will provide in-depth support when the system detects that students need help on key science and math competencies. Providing scaffolds to students instantaneously may be more beneficial for long-term learning.

Gobert says teachers are often frustrated because assessment data is neither timely nor adequate to understand students' needs to improve science instruction and learning.

“New technologies can provide more accurate assessment techniques than are typically used in schools such as lab reports or multiple choice answers,” Gobert said. “In turn, these can be used for personalized learning of science by students and better data-driven decision making by teachers, administrators and policymakers. Since the onset of COVID, many schools in all 50 states have used Inq-ITS successfully in remote contexts to support teachers’ instruction and students’ learning of science.

The research for the new project will take place with 2,000 freshmen and sophomores in New Jersey, Massachusetts, California and Kentucky in public high schools.