A Narrative-based Intelligent Tutoring System for Children
The Smart Primer is an interactive, narrative-based tutor that combines some of the best features of both physical- and electronic-based learning methods. The personalized narratives are based on each child’s physical surroundings as well as their current abilities. This system, inspired by Neal Stephenson’s groundbreaking science-fiction novel The Diamond Age, turns a tablet computer into a smart personal tutor.
In the Smart Primer, the narrative is interspersed with interactive problems that are tied directly into the story. The underlying tutoring algorithm will understand the child’s skill level, reactions, and emotions, and use this to give the child the right problems to solve at the right time. These activities are often driven by the location of the child (e.g., in their backyard) or the availability of other objects in their environment (e.g., fruit from their home). We also use NLP-based chatbots to scale up interaction and feedback for more personalized interaction. More importantly, the narrative tutor will leverage constructive learning techniques to build knowledge based on exercises that take place in the physical world, not simply behind a screen. Our initial focus will be on creating stories that help students develop skills in science, computing, mathematics, reading, and writing.
Extending our past work on narrative-driven, chatbot-supported learning, the Smart Primer research team has been pushing in two major directions over the past few years (2021-2024): (1) exploring how to blend real-world, hands-on educational activities with a technology-mediated learning experience, using augmented reality; and (2) investigating how to leverage large language models (LLMs) and other foundation models to personalize each child's learning journey.
Our research has resulted in several published projects (links available below) that have made progress toward the ultimate vision of Smart Primer. We have successfully created story-driven educational mobile apps that engage children with the world around them, personalize the experience to the learner, and allow the learner to express themselves and influence the story. Our projects have also shown promising findings that show children are highly engaged and more motivated to learn science after using our apps. Furthermore, our most recent project (in submission) found that a novel LLM architecture we developed led to higher learning gains.
The application revolves around a story in which the user and an adorable chipmunk, named Tangerine, are the main characters. The two of them are famous detectives in the magical world Diamondoid. From time to time, they are contacted by residents of Diamondoid who are in need of their help. In order to help them, the user needs to solve tasks that requires the use of math, science, and English. To solve the tasks, the user and Tangerine communicate and work together through a chat interface.
Our goal is to eventually build a smart and adaptive system in which the story and tasks adapt to the user’s skill set and preferences.