Parents and educators face countless hard decisions when a child struggles to learn to read. Having an effective support plan can mean the difference between a student thriving or falling behind. But developing that plan can mean poring over extensive literacy research to identify the specific strategies that will support students’ unique needs.
To help ease that burden, the Learning Engineering Virtual Institute’s AI Lab created a demo Early Literacy Interventions Tool. The pilot chatbot is trained primarily on the U.S. Department of Education’s Doing What Works Clearinghouse, a top resource for early childhood interventions. The ChatGPT-powered tool can answer questions and quickly give evidence-based recommendations to help design learning plans for students.
“The assumption is that educators and parents know where to find reputable information and have the prior knowledge to understand complex literacy issues,” says Perpetual Baffour, research director at The Learning Agency, that helped create the demo. “In reality, it’s a lot for educators to navigate.”
The LEVI AI Lab team wanted to save users hours of painstaking work without compromising the depth of research. “We know there’s high-quality research and data out there,” Baffour says. “We wanted to streamline getting it into the hands of parents and teachers. The chatbot makes it accessible and exciting because it feels interactive.”
At present, the prototype is best at conversational, low-touch exchanges. It’s adept at answering basic questions (for example, “What are promising strategies for instructing students with dyslexia?”) and summarizing strategies or data in simple language.
While the pilot tool can produce “hallucinations” like most LLMs, it aims to synthesize information and make useful recommendations. When asked to suggest a reading comprehension strategy for a student with dyslexia, the chatbot will propose a set of relevant approaches. The responses also typically come with a citation, so users know the source of the information they’re receiving.
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.
Early feedback from individual testers—a mix of parents, educators and researchers—has been positive. But the LEVI Lab team has its sights set on the future. They hope the tool can one day create detailed Individual Education Plans.
But addressing the needs of a unique child is a far more complex task. “We need to do more work and testing before we’re there,” Baffour says. “Our hope is that this prototype inspires school districts or organizations to see what’s possible. Then they can take the baton in building something more sophisticated.”
LEVI’s chatbot was designed to focus on early literacy interventions. However, the model could be easily adapted to investigate any learning disability. There is plenty of publicly-available research from sources such as the National Center for Learning Disabilities.
Future versions of the tool could be integrated with other databases. The code is available on HuggingFace, an open-source platform, for easy replication or customization.
Baffour points to the steady stream of developments coming from companies like OpenAI and Google. “We’ve just scratched the surface of what these tools can do," she says. "This feels like just the beginning.”
