HNDS-R: Understanding Expectation-Driven Learning in Early Childhood: An Experimental and Computational Investigation

The fields of developmental psychology and early childhood education both grapple with one critically important question: What factors best promote preschoolers’ memory and learning over time? Both fields point to two potential factors: 1) children’s prior knowledge and expectations about the world, which might help scaffold the learning of new information, and 2) children’s affinity toward surprising and expectation-violating information, which might help children allocate their limited attention and identify opportunities for new learning. Yet, exactly how these factors work together to shape the kinds of information preschoolers learn and maintain long-term is not well understood. Therefore, this project explores the kinds of information preschoolers learn and remember from events that either match with or violate their prior expectations. It also examines the individual cognitive abilities that allow young learners to leverage surprise to support memory and learning. The research has important implications for the development of effective AI tutoring systems that can be employed in early educational contexts. The proposed training plan builds the research infrastructure at a U.S. institution by training students in computer programming and computational methods, important skills for engaging today’s field of cognitive science and artificial intelligence. Given children’s limited cognitive capacities, an important question for developmental and cognitive science is what mechanisms best facilitate children’s learning and acquisition of new information? Past research points to two synergistic mechanisms: 1) children’s prior knowledge and expectations, which might be leveraged to scaffold new learning, and 2) surprise, the cognitive-affective state generated from experiences that violate prior expectations. It is likely that the combination of prior knowledge and attention to expectation-violating information helps children efficiently allocate their limited capacities to become robust learners. Yet, little research to date has directly investigated how young learners leverage their prior knowledge and affinity for surprising events to bolster episodic memory and subsequent learning. To this end, the current project 1) assesses the kinds of episodic details children learn from expectation-related and surprising events, 2) explores the cognitive capacities (i.e., executive functioning) and task-based factors that support learning in these contexts and 3) leverages computational modeling to evaluate the robustness of children’s expectation-driven learning over time, and the role of prior knowledge. This project is jointly funded by Perception, Action, and Cognition, Human Networks and Data Science, Developmental Sciences, Science of Learning and Augmented Intelligence, and the SBE Postdoctoral Fellowship Program. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.