Elizabeth Bonawitz

The Double-edged Sword of Pedagogy: Instruction limits spontaneous exploration and discovery

Motivated by computational analyses, we look at how teaching affects exploration and discovery. In Experiment 1, we investigated childrens exploratory play after an adult pedagogically demonstrated a function of a toy, after an interrupted pedagogical demonstration, after a naïve adult demonstrated the function, and at baseline. Preschoolers in the pedagogical condition focused almost exclusively on the target function; by contrast, children in the other conditions explored broadly. In Experiment 2, we show that children restrict their exploration both after direct instruction to themselves and after overhearing direct instruction given to another child; they do not show this constraint after observing direct instruction given to an adult or after observing a non-pedagogical intentional action. We discuss these findings as the result of rational inductive biases. In pedagogical contexts, a teachers failure to provide evidence for additional functions provides evidence for their absence; such contexts generalize from child to child (because children are likely to have comparable states of knowledge) but not from adult to child. Thus, pedagogy promotes efficient learning but at a cost: children are less likely to perform potentially irrelevant actions but also less likely to discover novel information.

Serious Fun: Preschoolers Engage in More Exploratory Play When Evidence Is Confounded

Researchers, educators, and parents have long believed that children learn cause and effect relationships through exploratory play. However, previous research suggests that children are poor at designing informative experiments; children fail to control relevant variables and tend to alter multiple variables simultaneously. Thus, little is known about how childrens spontaneous exploration might support accurate causal inferences. Here the authors suggest that childrens exploratory play is affected by the quality of the evidence they observe. Using a novel free-play paradigm, the authors show that preschoolers (mean age: 57 months) distinguish confounded and unconfounded evidence, preferentially explore causally confounded (but not matched unconfounded) toys rather than novel toys, and spontaneously disambiguate confounded variables in the course of free play.

Can Being Scared Cause Tummy Aches? Naive Theories, Ambiguous Evidence, and Preschoolers' Causal Inferences.

Causal learning requires integrating constraints provided by domain-specific theories with domain-general statistical learning. In order to investigate the interaction between these factors, the authors presented preschoolers with stories pitting their existing theories against statistical evidence. Each child heard 2 stories in which 2 candidate causes co-occurred with an effect. Evidence was presented in the form: AB?E; CA?E; AD?E; and so forth. In 1 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. Whereas 3.5-year-olds learned only from the within-domain evidence, 4- and 5-year-olds learned from the cross-domain evidence and were able to transfer their new expectations about psychosomatic causality to a novel task.

Children balance theories and evidence in exploration, explanation, and learning

We look at the effect of evidence and prior beliefs on exploration, explanation and learning. In Experiment 1, we tested children both with and without differential prior beliefs about balance relationships (Center Theorists, mean: 82 months; Mass Theorists, mean: 89 months; No Theory children, mean: 62 months). Center and Mass Theory children who observed identical evidence explored the block differently depending on their beliefs. When the block was balanced at its geometric center (belief-violating to a Mass Theorist, but belief-consistent to a Center Theorist), Mass Theory children explored the block more, and Center Theory children showed the standard novelty preference; when the block was balanced at the center of mass, the pattern of results reversed. The No Theory children showed a novelty preference regardless of evidence. In Experiments 2 and 3, we follow-up on these findings, showing that both Mass and Center Theorists selectively and differentially appeal to auxiliary variables (e.g., a magnet) to explain evidence only when their beliefs are violated. We also show that children use the data to revise their predictions in the absence of the explanatory auxiliary variable but not in its presence. Taken together, these results suggest that childrens learning is at once conservative and flexible; children integrate evidence, prior beliefs, and competing causal hypotheses in their exploration, explanation, and learning.

Occam's Rattle: Children's Use of Simplicity and Probability to Constrain Inference.

A growing literature suggests that generating and evaluating explanations is a key mechanism for learning and inference, but little is known about how children generate and select competing explanations. This study investigates whether young children prefer explanations that are simple, where simplicity is quantified as the number of causes invoked in an explanation, and how this preference is reconciled with probability information. Both preschool-aged children and adults were asked to explain an event that could be generated by 1 or 2 causes, where the probabilities of the causes varied across conditions. In 2 experiments, it was found that children preferred explanations involving 1 cause over 2 but were also sensitive to the probability of competing explanations. Adults, in contrast, responded on the basis of probability alone. These data suggest that children employ a principle of parsimony like Occams razor as an inductive constraint and that this constraint is employed when more reliable bases for inference are unavailable.

Controlling the message: preschoolers’ use of information to teach and deceive others

Effective communication entails the strategic presentation of information; good communicators present representative information to their listeners—information that is both consistent with the concept being communicated and also unlikely to support another concept a listener might consider. The present study examined whether preschool-age children effectively select information to manipulate others’ semantic knowledge, by testing how children choose information to teach or deceive their listeners. Results indicate that preschoolers indeed effectively select information to meet some specific communicative goals. When asked to teach others, children selected information that effectively spanned the concept of interest and avoided overly restrictive or overly general information; when asked to deceive others, they selected information consistent with the intended deceptive messages under some circumstances. Thus, preschool children possess remarkable abilities to select the best information to manipulate what others believe.

Bayesian models of child development

Bayesian models have been applied to many areas of cognitive science including vision, language, and motor learning. We discuss the implications of this framework for cognitive development. We first present a brief introduction to the Bayesian framework. Bayesian models make assumptions about representation explicit, and provide a detailed account of learning. Furthermore, they can provide an account of developmental transitions and other phenomena in development, such as curiosity and exploration. Drawing on recent work bridging empirical developmental data and modeling, we show that these features of the Bayesian approach provide solutions to problems that elude traditional accounts of learning and raise new areas of investigation.

Mind the gap: investigating toddlers' sensitivity to contact relations in predictive events.

Toddlers readily learn predictive relations between events (e.g., that event A predicts event B). However, they intervene on A to try to cause B only in a few contexts: When a dispositional agent initiates the event or when the event is described with causal language. The current studies look at whether toddlers’ failures are due merely to the difficulty of initiating interventions or to more general constraints on the kinds of events they represent as causal. Toddlers saw a block slide towards a base, but an occluder prevented them from seeing whether the block contacted the base; after the block disappeared behind the occluder, a toy connected to the base did or did not activate. We hypothesized that if toddlers construed the events as causal, they would be sensitive to the contact relations between the participants in the predictive event. In Experiment 1, the block either moved spontaneously (no dispositional agent) or emerged already in motion (a dispositional agent was potentially present). Toddlers were sensitive to the contact relations only when a dispositional agent was potentially present. Experiment 2 confirmed that toddlers inferred a hidden agent was present when the block emerged in motion. In Experiment 3, the block moved spontaneously, but the events were described either with non-causal (“here’s my block”) or causal (“the block can make it go”) language. Toddlers were sensitive to the contact relations only when given causal language. These findings suggest that dispositional agency and causal language facilitate toddlers’ ability to represent causal relationships.

Rational randomness: the role of sampling in an algorithmic account of preschooler's causal learning.

Probabilistic models of cognitive development indicate the ideal solutions to computational problems that children face as they try to make sense of their environment. Under this approach, childrens beliefs change as the result of a single process: observing new data and drawing the appropriate conclusions from those data via Bayesian inference. However, such models typically leave open the question of what cognitive mechanisms might allow the finite minds of human children to perform the complex computations required by Bayesian inference. In this chapter, we highlight one potential mechanism: sampling from probability distributions. We introduce the idea of approximating Bayesian inference via Monte Carlo methods, outline the key ideas behind such methods, and review the evidence that human children have the cognitive prerequisites for using these methods. As a result, we identify a second factor that should be taken into account in explaining human cognitive development--the nature of the mechanisms that are used in belief revision.

Sticking to the Evidence? A computational and behavioral case study of micro-theory change in the domain of magnetism

An intuitive theory is a system of abstract concepts and laws relating those concepts that together provide a framework for explaining some domain of phenomena. Constructing an intuitive theory based on observing the world, as in building a scientific theory from data, confronts learners with a “chicken-and-egg” problem: the laws can only be expressed in terms of the theorys core concepts, but these concepts are only meaningful in terms of the role they play in the theorys laws; how is a learner to discover appropriate concepts and laws simultaneously, knowing neither to begin with? Even knowing the number of categories in a theory does not resolve this problem: without knowing how individuals should be sorted (which categories each belongs to), a the causal relationships between categories cannot be resolved. We explore how children can solve this chicken-and-egg problem in the domain of magnetism, drawing on perspectives from history of science, computational modeling, and behavioral experiments. We present preschoolers with a simplified magnet learning task and show how our empirical results can be explained as rational inferences within a Bayesian computational framework.