Poster #196 - Preschoolers’ causal inferences from conflicting testimony and observations reflect informant accuracy but not self-knowledge

• In Event: 3-106 - Poster Session 11
  In Poster Session: PS 11 Section - Social Cognition

Integrative Statement

How do children incorporate information from testimony and statistical evidence when making causal inferences and evaluating informant reliability? Previous studies suggest children are sensitive to informants’ past accuracy and certainty, but that unlike adults, they are not sensitive to self-knowledge -- how well informants’ communicated certainty reflects their actual knowledge (Bridgers et al., 2016; Tenney et al., 2011). However, in this previous work, children were only exposed to one informant with either low self-knowledge or high self-knowledge. Four-year-old children are better able to demonstrate selective trust when informants with different levels of reliability are directly contrasted, and they are able to choose from whom to learn directly among them (e.g., Vanderbilt, Heyman & Liu, 2014). Would directly contrasting informants with low and high self-knowledge make this cue more salient?

In this study, we presented 4-year-olds (N = 49) with two informants who differed in self-knowledge. The informants both endorsed the same block as being better at making a machine play music, but one informant was certain while the other was uncertain. The informants’ endorsements were then contradicted by deterministic physical evidence demonstrating that the unendorsed block was in fact better (conflict phase), suggesting that while both informants were wrong, the uncertain informant had better self-knowledge (because she was aware she did not know which block was better, while the certain informant was confident about the wrong block). Children were then asked to identify which block was more effective at making the machine play music. All children chose the block that was demonstrated to work, and none chose the block that the informants’ endorsed, suggesting that they recognized that both informants were incorrect (exact binomial test, p < 0.0001). Children’s tendency to select the unendorsed block as better also significantly differed from children’s performance in a control condition (N = 32) where children were only given the informants endorsements (two-tailed Fisher’s exact test, p < 0.0001), and a majority (26/32) chose the endorsed block.

Next, both informants were certain about two new blocks but endorsed opposite blocks (generalization phase). Children were no more likely to trust the previously uncertain informant than the previously certain informant (two-tailed binomial test, p = 0.57). An additional control condition (N = 32) confirmed that children understood the difference in certainty between the two informants. When a certain informant endorsed one block and an uncertain informant endorsed the other block (and there was no conflicting data), children chose to place the certain informant’s block on the machine significantly more often than the uncertain informant’s block (23/32, p = 0.02). Our results are consistent with single informant studies showing that children do not preferentially trust informants with high self-knowledge over informants with low self-knowledge. These results deepen our understanding of the cues children use to evaluate informant reliability and how children integrate information from social and physical sources to inform their learning.

Authors

• Sophie Bridgers, Stanford University
  Non-Presenting Author

• Nafisa Bhuiyan, University of Toronto
  Non-Presenting Author

• Jing Yi Yvonne Wang, University of Toronto
  Presenting Author

• Daphna Buchsbaum, University of Toronto
  Non-Presenting Author