CMU Researchers Help Launch Project To Improve Education

Researchers at Carnegie Mellon University's School of Computer Science (SCS) are part of a nationwide project harnessing data from virtual tutoring services to advance teaching and learning.

School districts across the country invested in tutoring programs during the pandemic to supplement learning while students were at home. As a result, these virtual tutoring platforms collected large amounts of data about effective teaching practices.

Led by Cornell University, researchers from CMU, the Massachusetts Institute of Technology and the nonprofit research organization RAND launched the National Tutoring Observatory (NTO) to create the Million Tutoring Moves open-access dataset. It will be the largest dataset of its kind on effective teaching practices, and will support students while also aiding researchers and developers studying teaching and learning.

The NTO received $5 million in grants from the Gates Foundation and the Chan Zuckerberg Initiative to leverage artificial intelligence and transform that data into insights that can accelerate the science of teaching and learning.

Ken Koedinger, the Hillman Professor in CMU's Human-Computer Interaction Institute (HCII) and director of its Master of Educational Technology and Applied Learning Sciences (METALS) program, likened the NTO's impact to the invention of the telescope.

"Like any science, collecting data on what's happening drives our understanding and improves applications," he said. "For example, astronomers wanted to understand how planets, stars and galaxies move, so they built telescopes. For tutoring, it's critical that technique should lead technology. We want to discover the ideal techniques for supporting learning through tutoring and especially in supporting student engagement."

The NTO will bring together school districts, tutoring platforms, educators and researchers to collect virtual tutoring data that will be used to create an open-access dataset on effective teaching practices.

While tutoring happens in every school, Koedinger said there's little known about what teaching methods are effective. Researchers can use the data from the NTO to improve teaching and develop better AI-powered tutoring tools.

"Tutoring is happening all over the place in many different ways. The NTO is going to have a broad impact by figuring out what works in teaching and when and why," Koedinger said. "Some of these insights will be automated and some of them will be used in training. One-on-one tutoring can be super helpful for learning, but tutors can be good or bad. We don't really understand that difference well enough, and this project is an opportunity to dig into the question of what makes tutoring effective. How can we scale it? So many more students can benefit from it."

Koedinger has been in classrooms since the 1990s, working to understand human learning and create educational technologies that increase student achievement. He co-founded Carnegie Learning, which provides K-12 education technology, and is a principal investigator at PLUS, a platform that pairs technology with virtual tutors to improve students' math skills. Both of these organizations and others will partner with the NTO, which Koedinger says is still in its early stages and focused on setting up its infrastructure.

"CMU is uniquely positioned because we're on all sides of the NTO," he said. "We're a provider of virtual learning data, a user of the analytics within the infrastructure and a builder of the NTO infrastructure. There's this idea of, if they build it, they'll come. Well, we're already there, and we're trying to make the dreams come true for ourselves and other tutoring providers."

Along with Koedinger, the NTO's leadership team includes Rene Kizilcec, associate professor of information science at Cornell; Justin Reich, associate professor of comparative media studies and writing at MIT and director of MIT's Teaching Systems Lab; Rachel Slama, director of the labor and workforce development program and a senior policy researcher at RAND; Doug Pietrzak, CEO of Freshcognate; and Danielle Thomas, a systems scientist in the HCII.