Teaching three-and-a-half-year-olds to reason about ambiguous evidence

Abstract

Previous research suggests that three-year-olds fail to learn from statistical data when their prior beliefs conflict with the evidence. Are young children’s causal beliefs are entrenched in their folk theories or can young preschoolers rationally update their beliefs with evidence? Motivated by Bayesian accounts of rational inference suggesting that both statistical evidence and children’s prior beliefs should affect learning, we conducted a training study to investigate this question. Children (mean: 45 months) were included in the study if they failed to endorse the statistically more probable cause given ambiguous evidence (evidence in the form ABE, CAE, ADE, etc.) where the recurring cause, A, violated children’s prior beliefs. (A was a psychological cause; the remaining variables were biological.) The children were assigned to a Statistical Reasoning training condition, one of two Prior Belief training conditions (Baserates or Mechanisms), or a Control condition. Relative to the Control condition, children in the test conditions were more likely to endorse the a priori unlikely variable on a free-explanation task. Critically, children in the Statistical Reasoning condition passed the free-explanation task, even though their only information about the belief violating variable came from the ambiguous evidence. This suggests that teaching children statistical reasoning improves their ability to learn even from data inconsistent with their prior beliefs.