Exploring Young Children’s Ability to Integrate Reliability With Probability

• In Event: 1-088 - Children’s Early Ability to Reason in the Face of Uncertainty

Integrative Statement

By 6-months, infants have expectations about rudimentary probability (Denison et al., 2013). Recent work shows that children as young as 4 years can integrate proportional information with other knowledge, such as social stereotypes and a speaker’s testimony to make probabilistic inferences (Gualtieri & Denison, 2018, under review). The present experiment examines whether children can integrate statistical information with a variable physical causal system, to make inferences under uncertainty. We presented 4- to 6-year-olds (N=98) with a choice to play with one of two toy machines in an experiment presented on a laptop. To examine integration of statistical and causal information, we varied the distributions of toys in the machines across trials (i.e., the base-rates of target to non-target toys) and the reliability of the machines at producing a toy on each attempt.

To begin, children were familiarized to two machines filled with toys that were not used on test trials. One machine was unreliable, obtaining a toy on 33% of its attempts (and otherwise “dropping” the toy on the way to the chute); the other was reliable, obtaining a toy every time (see Figure 4). Children then saw test trials in which the same machines were filled with new toys – alien toys and grey balls. They were told that their goal was to choose a machine to play with that would get them an alien toy.

Children saw four problems in a counterbalanced order, which together diagnose whether they integrate the base-rates of aliens:balls and the reliability of the machines at producing toys. The base-rates changed across trials, but the reliability of the machines remained constant (one unreliable, one reliable). Base-rates were manipulated such that, in two problems, children should choose the unreliable claw to have the best chance of obtaining an alien, and in the other two problems they should choose the reliable claw (see Figure 5 for the base-rates and the probability of obtaining the target item when base-rates and reliability are integrated). If children appropriately choose the best machine on all four trials, then this provides evidence of integration.

Children received a score of 1 for choosing the machine that was more likely to produce the target toy on each problem. An ANOVA examining age (4,5,6) and problem type (A,B,C,D) revealed a problem type by age interaction F(4.973, 236.229)=2.409, p=.038, with 4-year-olds performing significantly differently than 5-year-olds (p=.047), and 6-year-olds (p<.001). Further, 4-year-olds failed to perform above chance on every problem type but one (Problem A, arguably the easiest problem). Both 5-year-olds (all p’s < or = .05) and 6-year-olds (all p’s<.05), performed better than chance on all problems. Thus, beginning around age 5, children are able to integrate information about a variable physical causal system with base-rates to make accurate choices under uncertainty. This suggests that young children’s causal and physical reasoning systems are well integrated at 5 years, and that children are flexible in their inferences under uncertainty, choosing a less reliable physical system or a more reliable one, depending on outcome probability

Authors

• Elizabeth Attisano, University of Waterloo
  Presenting Author

• Samantha Gualtieri, University of Waterloo
  Non-Presenting Author

• Stephanie Denison, University of Waterloo
  Non-Presenting Author