Preschool-Aged Children Jointly Consider Others’ Surprise and Prior Knowledge to Decide When to Explore

• In Event: Learning From Surprise in Childhood: Empirical, Methodological, and Theoretical Advances

Abstract

The role of prediction error in learning has been a topic of broad interest in developmental science. Recent work suggests that even infants and young children explore more and learn better when they observe events that violate their expectations or predictions (e.g., Stahl & Feigenson, 2019; Bonawitz, van Schijndel, Friel, Schulz, 2012; Sim & Xu, 2017; Brod, Hasselhorn & Bunge, 2018). Yet, opportunities to learn based on learners’ own prediction errors can be rather limited; to identify surprising events, the learner must know enough about the world to generate predictions in the first place. Thus, learners with insufficient prior knowledge – especially young children – may miss out on important learning opportunities.

Building on recent advances in children’s use of others’ emotional expressions for learning (Wu, Muentener, & Schulz, 2017; Wu & Schulz, 2018, 2019), here we ask if humans can use others’ expressions of surprise as vicarious prediction error to learn about the world. In Experiment 1, children (n = 56; 3.0-4.9 years) saw an experimenter discover an interesting function of a novel causal toy. Then the experimenter continued to play with the toy behind an occluder and expressed either surprise or happiness; the child could see her expression but not the toy. Rather than exploiting the observed function, children spent more time exploring other parts of the toy following the experimenter’s surprise (M = 58.7%, SD = 24.0) than her happiness (M = 41.1%, SD = 29.3; t(52.0) = 2.45, p = .017, 95%CI [3.21, 31.94]), suggesting that children inferred the presence of a hidden causal function from others’ surprised expressions.

Experiment 2 tested a more sophisticated form of emotion understanding: interpreting the same surprised expression differently depending on others’ prior knowledge. The procedure was the same as Experiment 1 except that the surprised expression was displayed either by the experimenter or a naive confederate who had no prior knowledge of the toy. Children (n = 56; 3.0-4.9 years) spent more time exploring other parts of the toy following the experimenter’s surprise ((M = 67.7%, SD = 22.3) than the confederate’s surprise (M = 45.3%, SD = 27.7; t(51.7) = 3.32, p = .002, 95%CI [8.83, 35.82]), suggesting that children can jointly consider others’ surprise and prior knowledge to guide their own exploration and discovery.

Taken together, these findings reveal children’s powerful abilities to learn from others’ surprise. In addition to learning based on their own prediction errors, children selectively interpret others’ expressions of surprise as “vicarious prediction error” depending on others’ prior knowledge, and strategically explore more when novel discovery is likely. Such abilities help children learn more from less data, supporting “smart” human social learning.

Authors

• Yang Wu, Stanford University
  Presenting Author

• Hyowon Gweon, Stanford University
  Non-Presenting Author