PopBots : leveraging social robots to aid preschool children's artificial intelligence education

Abstract

Today's children are growing up with artificial intelligence (AI) devices such as voice personal assistants, home robots, and internet connected "smart" toys. In previous research, we have seen that children lack understanding of how modern Al devices work, making it difficult for them to engage in reflective and constructive interactions with the Al-enabled technology (Druga, Williams, Breazeal, & Resnick, 2017). This thesis explores how young children explore and create with Al, and how such activities influence children's perceptions of Al and their attitudes about themselves as engineers. First, I discuss the design of PopBots -- the first hands-on toolkit developed for children ages 4-6 to explore and learn about Al. The social robot serves as both a programmable artifact as well as a window into understanding the machine learning algorithms. Accompanying this toolkit, I also developed a novel, developmentally-appropriate Preschool-Oriented Programming (POP) curriculum. The PopBots curriculum expands existing computational thinking curriculums by using creative learning activities to teach children three core Al concepts: rule-based systems, generative Al, and supervised machine learning. Next, I evaluated the PopBots toolkit and curriculum with 80 pre-K and Kindergarten aged children from local schools. I found that young children can understand most of the Al concepts presented in the toolkit, but sometimes developmental factors like grade and Theory of Mind skills made a difference. After completing the PopBots curriculum, children developed an understanding of robots as "learning" machines. They also gained confidence in their ability to build their own robots. Overall, this work provided a highly engaging opportunity for children to explore robotics, Al and programming -- and ultimately see Al-based technology as something they can play a role in not just using but also creating.