New AI Tool Aims to Analyze and Improve Classroom Discussions
The discussions in grade school math classes aren’t exactly known for being riveting, and U.S. students continue to underperform on math skills relative to the competition. Now, researchers at the University of Colorado Boulder are looking to shake that up with a tactical application of AI.
“There’s a big push to have teachers think deeply about their discourse practices and the conversations they have in math classes with students,” explained Jennifer Jacobs, an associate research professor at the UC Boulder’s Institute of Cognitive Science, in an interview with Nvidia’s Scott Martin. To enable that push, Jacobs and her colleagues (Abhijit Suresh, Vivian Lai, Chenhao Tan, Karla Scornavacco, Wayne Ward, James Martin and Tamara Sumner) have developed a new application called Talk Moves.
Using natural language processing (NLP) models, Talk Moves – named after the eponymous strategies that help teachers spurr constructive discussion in classrooms – generates transcripts of classroom discussions and analyzes those transcripts to evaluate what worked and what didn’t.
The NLP component – based on the BERT NLP model – was trained with over 500 transcripts of K-12 math classes (overall, more than 200,000 sentences), allowing the model to identify six different types of talk moves that had been hand-annotated by specialists in the training data. These talk moves include strategies like “keeping everyone together” and “getting students to relate.”
The training was conducted on Nvidia GPU-powered cloud instances. “We use the GPU parallelization to make sure we can train the model much faster — it’s much faster than running it on CPUs,” said Suresh.
Based on its analysis, Talk Moves can then output classifiers to identify which students are responding (and how often), which the developers hope will both spur engagement and improve equity in classroom settings. “A big goal of accountable talk is equity, because we want all students listening, participating, talking and being a part of that community,” Jacobs said.
The application has already been put into pilot use in two Colorado school districts, where teachers are helping to both test and design the evolving tool. “I’m always trying to improve conversations that happen in my math lessons and to help students have discussions with each other to explain their thinking. The application lets me see how I’m doing at meeting that goal,” said Kristin Holmquist, a fifth grade teacher at Eagleview Elementary School.
There’s still plenty of room for improvement, of course: Talk Moves currently underestimates student conversation when transcribing, and some of the talk moves aren’t identified as well as others. Now, the researchers are continuing to improve those metrics and get Talk Moves ready for primetime.