The process by which children learn native languages is markedly different from the process of learning a second, or non-native, language. Children are typically immersed in their native languages. They receive input from the adults and other children surrounding them, based on immediate need and interaction, during every waking hour.
Second language learners are exposed to input from the new language in very different ways, most commonly in a classroom setting. The second language learner relies heavily on memory skills with sparse interaction, in contrast to the first language learner that can rely on environmental reinforcement and social interaction to learn words.
Social robots have the potential to drastically improve on this paradigm, making the second-language learning experience more like the experience of learning a native language by engaging the child in a rich, interactive exposure to the target language, especially aspects not typically covered by traditional technological solutions, such as prosody, fundamental phonetics, common linguistic structures, etc.
Our project explores how to design child-robot interactions that encourage child-driven language learning, that adapt and personalize each child’s learning experience. We incorporate game design and machine learning into the child-robot interaction design. The child and robot play through a suite of educational games together. Using real-time sensor data and gameplay features, the robot constructs a model of each child's learning and emotional trajectory, then uses these models to inform its own decision making during the game. Thus, the robot's behaviors become personalized to individual children based on their learning style, personality and knowledge/emotional states during gameplay.