Jadon Geathers, a doctoral student in the field of information science, has taken many foreign language classes – French, then Mandarin and Spanish – but never felt fully engaged with a language until he started chatting in Mandarin with his fellow graduate students.

“The moment I started speaking with my friends, my proficiency increased substantially,” Geathers said. “The key was getting to engage in these authentic speaking opportunities in an environment that made me feel comfortable.”

Geathers realized that most language students lack similar opportunities, which led him to develop ChitterChatter. The foreign language conversation tool uses AI to give students a low-stress way to practice their speaking skills. The education technology (or “edtech”) tool can simulate conversations with different accents and dialects and gives feedback to users on their grammar and pronunciation. Unlike most language learning apps, ChitterChatter is designed for language educators to align speaking practices with the curriculum and track their students’ progress.

Geathers will present ChitterChatter on July 22 at the ACM Learning @ Scale 2025 conference in Palermo, Italy.

“It’s the next best thing to finding a native speaker,” said Rene Kizilcec, associate professor of information science at the Cornell Ann S. Bowers College of Computing and Information Science and the project’s faculty adviser. The tool won the 2025 Bits on Our Mind (BOOM) award at the annual Cornell Bowers CIS technology showcase, held in April.

Geathers and Kizilcec, who directs the Future of Learning Lab, hope the edtech tool will improve access to conversation opportunities for students everywhere, especially those who can’t find – or can’t afford – someone to practice with them.

“Quality language learning shouldn’t be a luxury. It is something that should be available to everyone,” Kizilcec said. “Language teachers recognize the importance of regular speaking practice, but they lack scalable ways to offer practice opportunities to their students.”

While the first pilot study included students learning Spanish, Kizilcec and Geathers are now recruiting college language instructors internationally as collaborators – so teachers can use the tool in their classroom and give feedback for further development. They also plan to expand the tool to include a wide range of languages.

Geathers developed ChitterChatter using OpenAI’s Realtime API, which enables AI-powered speech-to-speech interactions. The tool currently offers natural conversations with eight voices that convey emotion and sound like speakers from different regions and cultural backgrounds.

“Teachers can determine what they want students to learn and practice, and students can see the immediate connection between what they are doing in class and for their homework,” Kizilcec said.

Geathers conducted the pilot study with students in three college-level Spanish classes. In interviews, students overall reported they felt much more comfortable practicing with the AI conversation partner than with another person. Being able to converse without feeling self-conscious or judged was especially important to the students, who felt the tool would be most helpful as an ungraded part of the curriculum.

Now that they have a prototype, Geathers and Kizilcec are refining the platform using feedback from language instructors and students. Geathers wants to add more context to the interaction, such as an AI-generated image of the speaker, and background noise to set the scene, such as at a café or park. He is also working on improving the feedback the tool gives to students and enhancing the interface for language teachers to streamline the creation of new conversation modules.

Through ChitterChatter, they hope to put an end to the common phenomenon of people spending years studying a language without having the confidence to hold a real conversation.

“We want to give people plenty of opportunities to practice speaking,” Kizilcec said, “because that’s what makes languages beautiful and useful.”

AJ Alvero, an assistant research professor in information science at Cornell Bowers, also contributed to the development of ChitterChatter.

Patricia Waldron is a writer for the Cornell Ann S. Bowers College of Computing and Information Science.