Learning for the Long Run: Children Increasingly Prioritize Generalizable Knowledge in Tool Selection

Children are ecological learners; they adapt their information search strategies to fit the structure of their immediate learning environment. In two studies, we explore whether this search behavior is impacted by the potential for later information gain. In Experiment 1, 60 5-9-year-olds were presented with a population of different colored balls with a hidden property and a sample that was either representative or unrepresentative of that population. Children were asked to select a tool that would allow them to learn about the hidden property: one tool yielded more information from the local set of sampled items, while the other yielded more information about the global population. We find that, regardless of age, children consistently choose the most effective tool to support global information gain. In Experiment 2, we assessed whether this choice was truly motivated by the tool's potential to support learning that goes beyond the immediate sample (i.e., global yield), or due to children's reliance on another strategy (i.e., tool versatility or set ratios). To test this, we included 151 7-14-year-olds and 34 adults. In all conditions, participants were presented with a choice between learning tools that pitted global yield against an alternate strategy. We found that both children and adults consider global yield when selecting a tool, a tendency that increases with age. Children also consider the versatility of the tool when planning for downstream learning opportunities. We discuss the implications of these findings for understanding how learners develop strategies for prioritizing the acquisition of generalizable knowledge. Summary: Children choose information-gathering tools by balancing immediate usefulness with the potential for broadly informative learning. By age five, children use sample representativeness to guide prospective choices about how to learn. Younger children often rely on tool versatility, whereas older learners increasingly prioritize tools that support broader inference. Development reflects shifts in epistemic strategy use rather than convergence on a single optimal approach to information search.