Improving Education Access with AI

We have excellent secondary educational institutions in the US. However, not everyone can afford the cost of these transformative experiences. The issue of education inequality in the U.S. is a significant one, especially for children from low-income families. Many schools in underresourced areas struggle to retain seasoned teachers, impacting the quality of education these kids receive.

AI has the potential to be a game-changer here, especially for enhancing one-on-one tutoring used in these schools to support classroom learning. Imagine a scenario where an AI tool taps into the knowledge of experienced teachers, effectively bringing that expertise into virtual tutoring sessions. This is precisely the goal behind a system developed by researchers at Stanford University called Tutor CoPilot. Built on OpenAI’s GPT-4, it’s integrated with FEV Tutor, a platform that connects students with tutors online. Here’s how it works: Tutors and students communicate through a chat, and if a tutor needs help guiding a student, they can press a button to get suggestions from Tutor CoPilot.

The system wasn’t built overnight. Stanford researchers trained GPT-4 on a rich dataset of 700 actual tutoring sessions. These sessions were one-on-one, with experienced teachers working with students from first to fifth grade on math problems, guiding them to recognize errors and understand the underlying concepts. Now, Tutor CoPilot generates tailored responses, which tutors can refine to help students learn effectively.

As Rose Wang, a Stanford PhD student involved in the project, put it, “I’m really excited about the future of human-AI collaboration systems... This technology is a huge enabler, but only if it’s designed well.” The goal isn’t for AI to take over teaching. Instead, it acts as a support tool, helping tutors prompt students towards the right answers and deeper understanding. For example, Tutor CoPilot might suggest that a tutor ask how the student arrived at a certain answer or offer hints to encourage alternative problem-solving methods.

When testing the system’s impact, researchers looked at 900 tutors virtually teaching math to nearly 1,800 students from underserved communities across the Southern U.S. In this trial, half of the tutors had access to Tutor CoPilot, while the others did not. Results showed that students with Tutor CoPilot support were 4% more likely to pass an exit ticket assessment (with pass rates at 66% vs. 62%).

For now, Tutor CoPilot excels at assisting with foundational math concepts, says Simon Frieder, a machine-learning researcher at the University of Oxford who wasn’t involved in the project. "You couldn’t really do a study with much more advanced mathematics at this current point in time," he notes.

Not only does this tool bring experienced teaching strategies to newer tutors, but it’s also cost-effective, costing around $20 per tutor per year. This is a fraction of what it would cost for in-person training.

Moreover, Tutor CoPilot can strengthen the connection between tutors and students by training new tutors to adopt the problem-solving approaches of seasoned educators. Mina Lee, an assistant professor of computer science at the University of Chicago who reviewed the project, sees this as a positive development. “This work demonstrates that the tool actually does work in real settings,” she says, highlighting that AI can augment, rather than replace, meaningful human interactions.

Moving forward, the Stanford team plans to explore how well new tutors retain the methods they learn through Tutor CoPilot, giving insights into how lasting these AI-supported skills might be. They’re also looking into expanding the system’s scope to other subjects and age groups.

“There’s a lot of substantial ways in which the underlying technology can get better,” Wang says. “But we’re not deploying an AI technology willy-nilly without pre-validating it—we want to be sure we’re able to rigorously evaluate it before we actually send it out into the wild. For me, the worst fear is that we’re wasting the students’ time.

What are your thoughts about how to address this important issue towards an improved standard of living for all our citizens?