When Reem Tarazi and her husband moved from Philadelphia to the city’s suburbs, they prioritized finding the right school district for their two young children. Tarazi, a pediatric neuropsychologist, was concerned about class size, student-teacher ratios and extracurricular activities. They focused their search on Cheltenham, a diverse school district of 4,300 students just north of the city.
Choosing a school is one of the most important decisions a parent can make. But getting informed is often no simple task. While states and districts typically publish school report cards, the information can be confusing and hard to follow. Too often, parents are left to hunt for information and make sense of the data on their own.
The team at the Learning Engineering Virtual Institute (LEVI) Applied AI Lab wanted to make hard choices for parents like Tarazi a little easier. They created the School Comparison Tool, a pilot demo built around the Cheltenham School District, which does not publish school-by-school information on its website.
“We tried to build something that was data-driven but also parent-friendly,” says Perpetual Baffour, research director at The Learning Agency, which helped build the pilot. “Instead of asking parents to sift through number-heavy dashboards and reports, we created a chatbot that’s interactive, so parents can feel like they're in a conversation.”
The initiative is supported by the Walton Family Foundation, part of our ongoing effort to look for new ideas in education and support innovators with creative solutions.
The demo tool runs on ChatGPT and pulls statistics from the Future Ready PA Index, a statewide database of school progress measures. Parents can select from a menu of topics, including test scores, attendance rates, school enrollment and teacher quality. The chatbot’s responses come with helpful context and offer questions for further exploration.
For example, if a parent wants to compare test scores between schools, the pilot chatbot might suggest looking at which school performed better over the last three years to provide more insight into student progress over time. “These suggestions also give parents a little bit of inspiration about how they can approach chatting with the tool,” Baffour says.
The chatbot provided scaffolding and guidance which, for a non-educator like me, was exceptionally helpful.”
Tarazi had a chance to pilot the demo last month. “I’m not the most technologically savvy person,” she says. “So, what stood out to me was how user-friendly it is.” Tarazi liked the simple interface, but most of all, she appreciated the guidance offered by the chatbot.
“What’s really great about the tool is that it educates the user in a way that other websites don't,” Tarazi says. “The program didn’t just give me statistical data—it helped me interpret it. And I wasn’t restricted only to the questions I could come up with. The chatbot provided scaffolding and guidance which, for a non-educator like me, was exceptionally helpful.”
Tarazi recalls the process of trying to evaluate schools before moving to Cheltenham. Back then, she didn’t think to look at teacher tenure, but the pilot tool prompted her to consider how long teachers at a certain school had taught within the district. “I think that does tell you something about the quality of the district,” she says, “the contentment of teachers, and how those numbers may be linked to reading performance or other outcomes.”
That sort of deeper understanding is precisely what LEVI engineers hope to foster.
“Knowing that 95% of teachers in this school have certain credentials, or that 15% of students tend to be chronically absent—these are data points school district administrators use when they’re making policy decisions but aren’t always readily available to parents,” Baffour says. “We believe that kind of information can give parents a broad picture of what’s happening at a school beyond just academic performance.”
Tarazi hopes a future version of the tool will compare schools across districts. As the parent of biracial kids, she also wants to know demographic information about teachers and students at certain schools.
Baffour is confident that the next generation of the chatbot can be tailored to meet a diverse range of needs. “The basic prototype is out there. It's ready to be used and tinkered with,” she says. “We would love for schools or school districts or nonprofits to build and layer on it even more.” The code is published on GitHub and can be easily adapted to different data sources. A school district or state could customize the program and get it up and running within a day.
Baffour believes some of these customizations for the tool can include multilingual capabilities and the ability to display visuals like charts and graphs. As the underlying generative AI model advances, the tool will only get better.
“It’s intended to be an informational tool,” Tarazi says. “But, for me, it also functions as an educational tool. It helped me broaden the kinds of questions I asked and gave me input on how to understand the data. I would 100% recommend the tool to parents who are trying to find the right school for their child.”
