Teacher’s aide: BabbleBot assists educators with reading instruction

Fremont Elementary School kindergarteners have a new friend to help them learn how to read.

On April 8, a team from Trine University's computer and information technology (CSIT) senior design class introduced the students in Leanne Lies’ class to BabbleBot, an online application featuring a character (also named BabbleBot) who generates customized stories and pictures using phonemes and sight words they need to practice.

As the kindergarteners read the stories out loud, the application records and transcribes their reading for Lies to evaluate.

Science of Reading

BabbleBot is based on Science of Reading (SoR) principles, and Fremont Elementary School principal Eric Bryan said the project came out of SoR sessions Trine University’s Franks School of Education hosted last summer, funded by a Lilly Endowment Inc. grant.

“In those conversations, (while we were) brainstorming ideas of what can we do, the idea of building an app came up and I volunteered to be a guinea pig,” Bryan said. “The idea behind it is to streamline processes and gather data that’s going to help us drive instruction. Any way we can streamline that process, gives us more instructional time during the day.”

Lies said Science of Reading is based on teaching to read with sound combinations rather than pictures.

“Once the text gets more complex, they can’t rely on those pictures, so the students need to decode and read and sound out those words instead of using the picture for help,” she said.

One challenge Lies faces with implementing Science of Reading in her classroom is that none of her current reading books are decodable using SoR principles. She said the BabbleBot project gives an opportunity for her students to practice those skills.

Helping teachers, developing skills

The design team of CSIT majors Connor Heitman of Milford, Michigan, Kyczar Aalbregtse of Lafayette, Indiana, and Aaron Wrost of Brooklyn, Ohio, was tasked with creating an application that would implement SoR principles.

Each has family members who work in education.

“Knowing how hard it can be to help kids be engaged in the classroom, I wanted to be part of a team that makes learning fun,” Wrost said. “Also, being able to look over the work for all of the students in the classroom at once would be amazing for teachers, taking off a whole workload on their busy schedules."

“This (project) allowed me to understand the need for building an inclusive platform that takes into account the fact that all students learn differently, and that an app of this nature needs to encourage students to learn, rather than discourage,” Heitman said. “It also provided the opportunity to expand my skillset by developing my back-end development skills.”

Implementing AI

After considering multiple options to generate stories and pictures, the team decided to implement AI in BabbleBot.

The app uses OpenAI's GPT-4 language model to generate stories and the organization’s DALLE-3 image generation model for pictures. The stories incorporate the character and setting students choose from options available, as well as phonemes and sight words specified for a particular week’s lessons.

“There are not many AI-based projects being used in classrooms right now,” said Heitman. “It’s only a matter of time before AI is in every industry, similar to the internet or technology as a whole, and we wanted to be on the forefront of AI being used in an educational environment.”

None of the team had worked with AI in an application previously, and Heitman and Wrost were in CSIT’s cybersecurity track as opposed to web development, so there was a learning curve throughout the development process.

The group developed story templates and complex filtering algorithms to make sure content generated not only covered desired concepts but was age-appropriate.

“With AI being so new to many consumers, including parents, we believe that being transparent in how we use AI is so important, since without teacher/parent approval, no classroom will use BabbleBot,” Heitman said.

The team worked closely with literacy expert Alecia Pfefferkorn, assistant professor in the Franks School of Education throughout the design process, who facilitated further communication between the team and Lies.

“The students came in (the week of April 1) and showed us their design,” Lies said. “We gave them some input and they changed a few things.”

BabbleBot sends a recording and transcript of the student reading for the teacher to evaluate. The student moves on to a new set of phonemes and sight words once they pass four lessons, the amount Fremont teachers felt reflects mastery.

“The idea is that once a student learns a phoneme or sight word, they add it to their toolbox that will help them read more complex words in the future,” Heitman said.

Babbling onward

Team members hope to continue developing and adding new features to BabbleBot, even after their coursework is done.

"I would love to continue to work on BabbleBot, commercializing it and implementing it into as many classrooms as possible,” said Heitman. “I feel it is a platform with high potential and want to see it help bridge the literacy gap we face as a nation."

The pilot version of BabbleBot uses phonemes and sight words provided by Fremont Elementary School. Future iterations will allow teachers to import their own.

The team also hopes to design an algorithm that will allow AI to accurately measure how well a student says each word.

Whatever path BabbleBot travels, Bryan says there is a critical need for new tools to assist in the teaching of reading.

“Our reading skills are low across the state, and the question is what we can do to build a better foundation of reading to help our kids be more successful,” said Bryan. “There are a lot of statistics and data that show if you cannot be a fluent reader by third grade, the likelihood of you graduating drops drastically.”

The Trine students hope they have been able to develop a tool that will eventually help kids succeed in school and beyond.

"I didn't know what it would turn into, but I am very glad I made the decision to work on this," Aalbregtse said.