AI to Enhance and Support Children’s Learning
When ChatGTP was released in December 2022, people jumped on and started asking it questions and getting answers. This can be really helpful in many contexts, but in learning, we ideally want the model to help lead the student to getting to the answer themselves.
One of the key challenges in education is that classrooms are made up of individual students who are at different levels of learning. Teachers consistently say that working to target learners where they are while at the same time covering the grade level content is a big challenge. The COVID experience has further increased the variability in levels in classrooms and, therefore the need for individual experiences. At the same time, there are big new developments in the AI space. It's possible that work in AI can help solve this big education challenge.
The microcomputer revolution of the late 1970s and early 1980s set in motion the development of intelligent tutors to apply a psychology-based approach to learning. The key goals for new systems were to be able to teach a task as well as perform it, adapting dynamically to the learner. However, these intelligent tutors struggled to ask and process questions that lead to deep learning, like, “Why do you think that?” “What would happen if…” “How might we…”
The launch of large language models opens new possibilities with AI that can generate text and pictures. With the right prompts (instructions that tell the models how to act), the models interact with students like a socratic tutor, refraining from giving the answer to learners and instead asking questions that prompt learners to explain their thinking, elaborate on answers, and go deeper in their responses.
The models are able to do more than tutor. They can: simulate characters in books that students can chat with, debate with learners, and co-write stories, to name a few. They can also write essays that mimic student work. We should not ignore the potential misuses of the technology. Education-specific applications built on top of the models are often designed to address the issues specific to school settings.
More research needs to be done to see how to best use the tools. We also need more work to help parents, students, and teachers learn how to interact with them. A good way to join the conversation about AI is to jump in and try working with some AI tools. Khanmigo, Khan Academy’s AI tutor for students and assistant for teachers is one of these tools you can try. Only with people using and testing them will we see if they deepen student understanding and improve learning outcomes.