Do children match described probabilities? The sampling hypothesis applied to repeated risky choice

Abstract

• We examined children’s repeated choices when outcome probabilities are described. • We connect research on the development of causal learning and adults’ risky choice. • Hypotheses from causal learning did not extend to children’s repeated risky choices. • Children’s choices were best described by wide exploration and option-switching. • We highlight the importance of the learning format in children’s repeated choices. One way in which children can learn about probabilities of different outcomes before making a decision is from description, for instance, by observing graphical representations of frequency distributions. But how do repeated risky choices develop in early childhood when outcome probabilities are learned from description? Integrating previous findings from children’s sampling processes in causal learning and adults’ repeated choice behavior, we investigated repeated choices from 201 children aged 3 to 7 years and 100 adults in a child-friendly risky choice task. We expected young children to probability match and predicted that the perceived dependency between choices would shape the underlying choice process. However, the assumed cognitive processes derived from the causal learning and risky choice literature did not generalize to children’s or adults’ repeated risky choices when outcome probabilities were learned from graphical representations prior to making a decision. Moreover, choice behavior did not differ as a function of the perceived dependency between guesses. Instead, children broadly diversified choices, and switching between options dominated older children’s choice behavior. Our results contribute to increasing evidence of childhood as a phase for heightened exploration and highlight the importance of considering the learning format when studying repeated choice across development.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 1 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.