Children's Differential belief revision about preferences and normative rules

• In Event: Poster Session 04
  In Poster Session: PS 04 Topic - Social Cognition

Abstract

As adults, we are often resistant to changing our minds; we don’t always revise our explicitly stated beliefs when presented with evidence that contradicts them (e.g., Skita, 2010). In childhood, on the other hand, belief revision is a part of learning: studies have shown that children are rational learners: given the appropriate evidence, children revise their beliefs about physical, psychological and social causes (e.g., Kimura & Gopnik, 2018; Bonawitz et. al., 2012; Kushnir & Gopnik, 2007) often more “open-mindedly” than adults (Lucas et. al. 2013). But studies of children’s rational learning do not generally require explicit statements of hypotheses about evidence. In a recent study, Macris & Sobel (2017) found that stating a physical-causal belief explicitly – even if just a guess – changes the revision process for children. Once stated, it takes more than mere counterevidence to override it, often requiring some form of causal explanation or explicit contradictory testimony.
We follow up on this finding by investigating whether the same is true of social beliefs.
We examine two types of social beliefs that children can reason about explicitly: norms and preferences. In the ongoing study, moderated via Zoom, preschool-aged children (n = 26, Mage = 59.8mos, SD = 5.8mos, Range = 49.9-71.8mos) are randomly assigned to the Norm or the Preference conditions. Children are taught about a new creature wearing an item of clothing with two salient features - shape and pattern (e.g., triangular, striped hat). In the Norm condition, participants were given a rule for Daxes, “they must wear hats that look like this” while those in the Preference condition were told a rule about an individual Dax, “he likes hats that look like this.” Children are asked to explicitly endorse one of two mutually exclusive hypotheses supported by the initial data (triangular or striped). Following their endorsement, children are presented with three pieces of conflicting counterevidence (for example, if they endorsed stripes, they see x3 evidence supporting triangles, not stripes). Children then have the option to revise or maintain the initial endorsement. This procedure is repeated across two trials. For each trial, children received a score of 1 if they revised their original hypothesis and a 0 if they did not.
Across conditions, there was a low rate of revision, X2(2) =5.077, p=.079. This is consistent with findings on children’s revision of an explicitly stated hypothesis. Further, there is a trend showing different rates of revision across domains. Hypotheses about norms (Mscore=1.08) may be more readily revised than those about preferences (Mscore=0.64). This is consistent with findings on the obligatory nature of norms - when a norm is in place, one ought not act otherwise thus children may feel it necessary to revise an initial hypothesis given new evidence. On the other hand, it is acceptable to hold multiple preferences, or even to hold a preference but act otherwise. Because our results are preliminary, we suggest future empirical work to explore domain differences, and to contrast explicit belief revision about social causes to more implicit forms social learning.

Authors

• Zoe Finiasz, Cornell University
  Presenting Author

• Dana Asal Karami, Cornell University
  Non-Presenting Author

• Esha Sheth, Cornell University
  Non-Presenting Author

• Tamar Kushnir, Duke Univeristy
  Non-Presenting Author