UC Berkeley

In the study of early childhood development, the “active child” is an enduring theme that has inspired, motivated, and puzzled developmental psychologists over the decades. Despite a large body of evidence demonstrating that young children can allocate their attention and exploration in a non-random manner, we still do not have clear answers for how being active influences children’s development. How does the “active child” fit into the story of development? It is often thought that activity is relevant to development because it contributes to learning. This idea is consistent with the perspective that children may be active learners. However, to date, there is very little empirical work showing that young children can successfully acquire specific pieces of knowledge through the voluntary allocation of attention, self-driven exploration/free play, or question-asking. At the same time, the definition of active learning itself has only grown more divergent. It has been taken to mean physical activity while engaging in a task, a learning method that emphasizes higher-order thinking, elaborative cognitive processes such as generating explanations, interactive or collaborative learning, etc.

The current dissertation seeks to understand whether learning can successfully occur through children’s own actions, as well as the processes that support such learning. Our work suggests that active learning is not a single unitary process—instead, it may be better described as multiple processes working in concert so that learning can actually occur. First, learners will need to detect situations when there is something to be learnt. Chapter 2 presents a series of experiments demonstrating that the looking times of 8-month-old infants appear to be driven by the evidence an observed event provides for a set of alternative hypotheses over the currently favored hypothesis. In other words, young infants can successfully detect situations where their current understanding is inaccurate or incomplete, and there may be a better explanation for the events that they observed. Second, learners will need to selectively attend to or approach potential sources of information. Chapter 3 provides evidence that 13-month-old infants preferentially approach and explore sources of unexpected events, which are great opportunities for obtaining information relevant for theory revision. Third, learners would have to generate evidence that is relevant to the learning goal. Chapter 4 presents empirical findings that preschool children can systematically generate data to learn about categories, and the systematicity of their information search strategy is correlated with their classification performance. Chapter 5 demonstrates that preschool children judge the effectiveness of presented questions in a way that is consistent with maximizing information gain, suggesting that the computational foundations for developing effective information search strategies may be in place by an early age. Finally, learners would have to actually learn from the self-generated data, incorporating the observed outcomes into one’s knowledge. Chapter 6 shows that children as young as 2-years-old can acquire higher-order generalizations based on self-generated evidence through the course of free play.

Taken together, the experiments presented in this dissertation demonstrate that children are active learners, and the component processes that support active learning is present by early childhood. However, these processes may emerge at different time points during development, and the capacities also continue to develop, enabling children to become more adept at active learning over time.