How Data and Innovation Are Paving the Way for Student Success in STEM

For many aspiring young scientists and engineers, the first year of college is a crossroads. Introductory classes in math and science can determine who pursues careers in STEM and who changes course. These classes can act as bottlenecks and discourage students early on, preventing thousands of students from pursuing careers in math and science.

“Students are going to need to understand how modern technologies work, and STEM gateway courses are the beginning of that process,” said David Yaron(opens in new window), a learning science(opens in new window) researcher and chemistry professor(opens in new window) in Carnegie Mellon University’s Mellon College of Science(opens in new window). “Whether or not students end up in STEM jobs, developing quantitative and technical skills will be important.”

Learning science researchers at CMU are beginning to crack the code on how to use AI enabled technology to help students from all backgrounds get through the math gate successfully.

At a time when there’s a growing need to strengthen the STEM workforce, learning science researchers at CMU have found that the solution may not only come from recruiting more young people into STEM, but also by helping more students get through these challenging intro courses by enhancing traditional learning settings with educational technology.

“Students come in with widely varying preparation,” Yaron said. “Some students are seeing the content in an intro math or science course for the second or third time, and some are seeing it for the first time. Our data is showing that, with our courseware, students who start behind often learn just as effectively, sometimes even at a faster rate than their better prepared peers. The challenge is that they begin farther back, so they need more time and sustained practice to close the gap.”

Improving the calculus for the next generation 

These learning gaps can feel daunting. Approximately one in three students enrolled in higher education across the U.S. are derailed by these gateway courses, with introductory math being the greatest barrier, according to the Mathematical Association of America. 

To expand support for students at this critical educational phase, CMU launched Learnvia(opens in new window), a nonprofit learning collaborative. Free to students, Learnvia combines AI-enabled courseware, support for evidence-based instruction, and the power of learning research to strengthen student success and course outcomes at colleges and universities nationwide.

“For decades, Carnegie Mellon University has served as a global leader at the intersection of AI, research innovation and learning science,” CMU President Farnam Jahanian(opens in new window) said. “With support from the Gates Foundation, Learnvia is using this expertise to develop cutting-edge, AI-enabled courseware that is designed to improve outcomes at scale and help students across the nation not just endure gateway college courses but thrive in them.”

Learnvia’s national network currently serves students at dozens of higher education institutions — from community colleges to small private universities and large public research universities.

“Economic growth depends on a strong, skilled and adaptable talent pipeline. Learnvia’s integration of learning science and AI-enabled instruction strengthens how students build foundational skills at scale,” said Madeline Pumariega, president of Miami Dade College. “We look forward to working with Carnegie Mellon University as its long-standing leadership in learning science is translated into real benefits for broad-access institutions and the students we serve.”

Calculus I is the first pilot course, and there are plans to add Quantitative Reasoning, Pre-Calculus, Calculus II and Calculus III over the next three years. Students who sign up attend instructor-guided classes and complete homework and tests as they would in in-person classrooms. They also get an AI tutor that gives them useful in-the-moment feedback, reducing confusion and discouragement that can lead to dropping courses.

Online support leads to improvements in the classroom

Learnvia builds on Carnegie Mellon’s legacy of learning science innovation. Researchers at CMU’s Open Learning Initiative have been working to improve student learning outcomes for more than two decades. The initiative has developed a host of courseware in subjects like physics, statistics and biology.

“If we want to serve today’s learners well, innovation has to be grounded in science,” said Norman Bier, executive director of CMU's Open Learning Initiative(opens in new window), The Simon Initiative(opens in new window) and Learnvia. “We treat every course as a hypothesis about how students learn best. We design for interaction, feedback and practice, and then we study the data to continuously improve. After 25 years, the evidence is clear: when instruction is interactive and feedback-rich, students succeed.”

As students learn, they’re given instant feedback on what they’re getting right and wrong in order to stay engaged with the course work. Behind the scenes, researchers study how students interact with the course material and use that data to make continual improvements to the courses themselves. Since the initiative’s creation, researchers have found notable learning outcomes for students. They learned more in less time and got feedback when they needed it most.

Researchers found that tailored learning can also level the playing field for students in introductory courses like chemistry. REAL CHEM(opens in new window), a CMU collaboration with Arizona State University, builds on the successes of the Open Learning Initiative, by combining in-person teaching and chemistry lab work with online instruction to deepen learning and improve student outcomes. 

“A student gets immediate feedback if they get something wrong,” Yaron said. “Students are always working on something that they really need to learn — not something they already know. The materials are designed to do that and they do it by asking the student to do something, and if they don't know how, immediately telling them how to do it and giving them a chance to do it again.”

Gizelle Sherwood(opens in new window), the director of undergraduate studies and a teaching professor in the Department of Chemistry, teaches introductory chemistry courses to first- and second-year students. She said REAL CHEM, of which she is also a contributor, has made a difference in her classes. 

“It didn’t really matter what high school you attended,” Sherwood said. “It didn’t matter whether you did or did not take AP Chemistry. Your past experiences don't matter. I’ve had kids with very little background in chemistry and others with extensive experiences. Assigning REAL CHEM as pre-reading for students has now ensured that all of my students are entering the classroom feeling that they are on equal footing.” 

Sherwood said students who use the platform perform better on exams.

Researchers hope the success students experience through these initiatives will help them stick with STEM long after their coursework ends, significantly impacting whether challenging introductory courses, like Calculus I, encourage or deter young students from pursuing critical STEM careers.

"The evidence is clearly established — we know we can improve outcomes through careful learning science and technology,” Bier said. “Now we are focused on impact at scale. The real opportunity is ensuring that 10 years from today, we’re no longer talking about a gateway math crisis because we’ve made success the norm rather than the exception."