Predict, reflect, revise: Moderators and mechanisms of conceptual learning in childhood

Abstract

Conceptual learning in childhood is challenging, particularly because children’s intuitive theories often conflict with scientifically accurate concepts, and these beliefs rarely revise spontaneously. Prior research has largely focused on structured instructional settings, leaving uncertainty about how learning unfolds in open, inquiry-based contexts. It is also unclear how developmental improvements in conflict monitoring support belief revision during learning, and conceptual learning outcomes have rarely been assessed beyond accuracy-based measures. The overarching goal of this dissertation was to investigate how conceptual learning in childhood is supported by both stable individual differences and state-level cognitive mechanisms, and how these processes can be leveraged through inquiry-based learning. It examined whether trait-level cognitive reflection and state-level conflict monitoring predicted learning in inquiry-based contexts, and how developmental differences in conflict monitoring relate to successful belief revision. It also evaluated whether generating predictions enhanced learning in complex, multidimensional tasks, with immediate and longer-term outcomes quantified using Bayesian computational modeling. Children aged 5 to 10 completed tasks in which they generated predictions or post-hoc evaluations across simple and complex science domains. Behaviorally and physiologically, incorrect predictions elicited signatures of conflict reflected in slower responses and increased pupil dilation. Conflict monitoring and belief revision efficiency improved with age, and trait-level cognitive reflection predicted deeper conceptual learning. While generating predictions facilitated belief revision in simple and complex tasks, its benefits did not translate to deeper conceptual understanding or long-term retention, revealing boundary conditions for conflict-enhancing strategies. The findings indicate that conceptual learning is facilitated by heightened conflict monitoring catalyzed by actively generating predictions, but that cognitive conflict alone is insufficient for deeper and durable conceptual learning. By integrating trait-level moderators, state-level mediators, and cognitive conflict-enhancing learning strategies, this dissertation provides an integrative account for supporting effective conceptual learning in childhood.