This research explores how making explicit predictions improves science learning in primary school children by examining the role of executive functions, specifically inhibition and switching. Through a water displacement experiment, the study compares children who predict outcomes with those who evaluate them after the fact. Bayesian computational models reveal that while children with lower executive function scores struggle in post-event evaluation, the act of predicting “equalises” performance by promoting greater metacognitive flexibility. These results suggest that prediction scaffolds the belief revision process, allowing children to update their internal models regardless of their baseline cognitive control. Consequently, the authors propose that prediction serves as a vital learning strategy to offload the cognitive demands of scientific reasoning in developing minds.