“Orange”, she Wrote: Children’s Ability to Make Inferences About the Credibility of Text Based on an Author’s Knowledge Access

• In Event: 3-183 - When is Knowledge Justified? Investigating Children’s Epistemic Understanding

Integrative Statement

An extensive body of literature has established that in face-to-face interactions, young children are able to evaluate the credibility of verbal information, or testimony, by evaluating the credibility of the source (see Harris, Koenig, Corriveau & Jaswal, 2017; Mills, 2013 for a review). Less is known, however, about what children do to evaluate the credibility of second-order sources— such as books or web pages— that do not readily allow for the evaluation of their author. Some research indicates that preschool children prefer to trust written information over other sources, e.g., preferring to trust a person reading a text label over a person presenting oral testimony (e.g., Corriveau, Einav, Robinson, & Harris, 2014; Einav et al., 2018). Little research has explored whether children always display a preference for written information, or if they modify their trust based on epistemic information about the author (but see Vanderbilt, Ochoa, & Heilbrun, 2018 for children’s ability to use author accuracy). Here, we ask if children are able to use information about author knowledge access to make inferences about the credibility of written text.
Sixty-nine children (33 female) ages four through six were shown three trials where a colored ball (e.g, red vs. yellow) was hidden in a box, and an author wrote down the color they believed was in the box. Half of the children saw a knowledgeable author (i.e. looked inside the box before writing), whereas half saw an ignorant author (did not look inside). Next, children saw two informants who offered contrasting opinions either sourced from the author’s text (text informant; “This says that it is…”) or from their own knowledge (oral informant; “I say it is…”) and children were invited to endorse one of the two informants and explicitly indicate which informant was “better” at stating the color of the ball (at no point did children ever receive accuracy information). Results indicate that children in the knowledgeable author condition endorsed the text informant significantly above chance (t(33) = 6.09, p < 0.001), whereas children in the ignorant author condition did not (t(36) = 0.442, p = 0.66). We also conducted a linear regression on the total number of trials in which children endorsed the color indicated by the text (i.e., the author’s testimony) with children’s explicit judgment of which informant was “better” and age as the predictors. Results indicated that condition (B = 0.60, S.E. = 0.193, p = 0.003) children’s explicit judgment (B = 0.80, S.E. = 0.21, p < 0.0001) were significant predictors, though age (B = -0.06, S.E. = 0.16, p = 0.74) did not. Overall, the model accounted for 30.70% of the variance in test trials, F (3, 65) =9.60, p < 0.0001, see Figure 1. Taken together, our findings indicate that young children use author knowledge access to determine whether or not a text is trustworthy. We discuss implications of these findings for future exploration of children’s evaluation of second-order information sources.

Authors

• Ian L. Campbell, Boston University
  Presenting Author

• Kathleen Helen Corriveau, Boston University
  Non-Presenting Author