Can being scared make your tummy ache? Naive theories, ambiguous evidence and preschoolers’ causal inferences

Abstract

Causal learning requires integrating constraints provided by domain-specific theories with domain-general statistical learning. In order to investigate the interaction between these factors, preschoolers were presented with stories pitting their existing theories against statistical evidence. Each child heard two stories in which two candidate causes co-occurred with an effect. Evidence was presented in the form: AB àE, AC à E, AD à E, etc. In one story, all variables came from the same domain; in the other, the recurring candidate cause, A, came from a different domain (A was a psychological cause of a biological effect). After receiving this statistical evidence, children were asked to identify the cause of the effect on a new trial. Consistent with the predictions of a Bayesian model, all children were more likely to identify A as the cause within domains than across domains. While three-and-half-year-olds learned only from the withindomain evidence, four-and five-year-olds learned from the cross-domain evidence and were able to transfer their new expectations about psychosomatic causality to a novel task.KEYWORDS: domain-general and domain-specific causal learning, Bayesian models, naïve theories, ambiguous evidence, psychosomatic causes NAÏVE THEORIES, AMBIGUOUS EVIDENCE Page 3Can being scared make your tummy ache? Naive theories, ambiguous evidence and preschoolers' causal inferences By the time children are five years old, they understand causal relationships in a variety of domains (Flavell, Green, & Flavell, 1995;Gelman & Wellman, 1991;Gopnik & Meltzoff, 1997;Inagaki & Hatano, 1993;Kalish, 1996; Perner, 1991; Shultz, 1982; Spelke, Breinlinger, Macomber, & Jacobson, 1992). Many researchers have suggested that children's causal knowledge can be best characterized as a set of naive theories:abstract, coherent representations of causal structure that support prediction, intervention, explanation, and counterfactual claims (Carey, 1985;Harris, German, & Mills, 1996;Hickling & Wellman, 2001;Gopnik, 1988;Gopnik & Meltzoff, 1997;Keil, 1989; Perner, 1991;Sobel, 2004; Wellman, 1990; Wellman, Hickling, & Schult, 1997). The view that children's causal representations resemble scientific theories (the theory theory) suggests both that patterns of evidence should affect children's causal commitments and that children's causal commitments should affect their interpretation of evidence. Indeed, this dynamic relationship between domain-appropriate causal beliefs and evidence has been taken as a defining feature of theories (e.g., Gopnik & Meltzoff, 1997).However, despite the expectation that theory and evidence should interact, developmental psychologists have been largely divided between accounts of causal reasoning emphasizing either domain-specific causal knowledge or domain-general learning from data. Thus some researchers have suggested that children's naive theories might be generated by domain-specific modules (Leslie, 1994; Scholl & Leslie, 1999) or innate concepts in core domains (Carey & Spelke, 1...

Win-Stay, Lose-Sample: A Simple Sequential Algorithm for Approximating Bayesian Inference

“…Gopnik & Meltzoff, 1997;Gopnik & Schulz, 2007;Wellman & Gelman, 1992) and causal knowledge and learning have been a particular focus of formal work in machine learning and philosophy of science (Pearl, 2000, Spirtes, Glymour & Scheines, 2000. Causal learning has also been one of the areas in which Bayesian models have been particularly effective in characterizing human behavior, from adults learning from contingency data (Griffiths & Tenenbaum, 2005) to children reasoning about causal systems and events Bonawitz, Shafto, et al, 2011;Gweon, Schulz, & Tenenbaum, 2010;Kushnir & Gopnik, 2007;Schulz, Bonawitz, & Griffiths, 2007;Griffiths, Sobel, Tenenbaum, & Gopnik, 2011;Seiver, Gopnik, & Goodman, 2012).…”

Teaching three-and-a-half-year-olds to revise their beliefs given ambiguous evidence

“…To our knowledge, however, only one previous study has directly tested whether, consistent with the predictions of a Bayesian inference model, children can use probabilistic evidence to revise their domain-specific beliefs (Schulz, Bonawitz, & Griffiths, 2007). In that study, children were provided with ambiguous data both in contexts where they had strong prior beliefs and ones where they did not.…”

“…In the Schulz et al (2007) study, preschoolers were read two books in which two candidate causes co-occurred with an effect. Evidence was presented in the form ABE; CAE; ADE, etc.…”

Probabilistic Models, Learning Algorithms, Response Variability: Sampling in Cognitive Development

“…But not all children pick the most likely response, and an individual child may change responses from trial to trial [e.g., [5][6][7]. The proportion of times that children select a hypothesis increases as the hypothesis receives more support, but children still sometimes produce alternative hypotheses.…”

“…Computational-level models provide clear descriptions of the problems the learner faces and describe ideal solutions for those problems. Probabilistic models at this computational level can characterize how children infer beliefs from evidence in at least some cases, such as causal learning tasks [4][5][6][7]. In these studies, researchers assess the state of children's prior beliefs, carefully control the evidence they receive, and then examine which hypotheses they endorse.…”