Surprise-induced exploration as a tool for learning: A comparative approach with human infants and non-human primates

Abstract

Upon witnessing a surprising event, young humans will often explore the target of that surprise, appearing to seek an explanation (eg: Bonawitz et al., 2012; Chandler & Lalonde, 1994; Perez & Feigenson, 2022; Stahl & Feigenson, 2015; van Schijndel et al., 2015). Surprise-induced exploration could serve to reveal otherwise opaque causal mechanisms, helping to scaffold causal reasoning so that learners can begin to think abstractly about how entities in the world relate. Humans seem uniquely capable of this type of reasoning. Could it be that humans’ tendency to seek explanations sets them apart from even their closest living relatives? In this thesis, I aim to shed light on this question by conducting comparative research with humans and non-human primates. In Chapter 2, I assess looking time as an indicator of surprise, one of the primary tools that has been used to assess non-verbal individuals’ expectations. I argue that this method, while still extremely valuable and worthwhile, may not be as flexible and robust a measure as it is often taken to be, because neither infants nor monkeys looked longer at events that past research tells us they should find surprising (Baillargeon, 1987, 1995; Spelke et al., 1992). In Chapter 3, I present a paradigm for assessing individuals’ expectations about how objects act and interact in the world, demonstrating that bonobos, one of humans’ closest living relatives, can use principles of object dynamics to locate hidden items. And in Chapter 4, I provide evidence that bonobos do not preferentially explore objects that are unexpected, suggesting that they do not seek explanations for surprising events. The research presented in this thesis ultimately provides evidence that humans and primates do not seem to differ in their underlying expectations about how objects interact in the world, but humans may be unique in their propensity to harness instances where these expectations are violated to scaffold more abstract causal reasoning.