Preschool children learn about causal structure from conditional interventions

The conditional intervention principle is a formal principle that relates patterns of interventions and outcomes to causal structure. It is a central assumption of experimental design and the causal Bayes net formalism. Two studies suggest that preschoolers can use the conditional intervention principle to distinguish causal chains, common cause and interactive causal structures even in the absence of differential spatiotemporal cues and specific mechanism knowledge. Children were also able to use knowledge of causal structure to predict the patterns of evidence that would result from interventions. A third study suggests that children's spontaneous play can generate evidence that would support such accurate causal learning.     

Inferring causal networks from observations and interventions

Information about the structure of a causal system can come in the form of observational data—random samples of the system's autonomous behavior—or interventional data—samples conditioned on the particular values of one or more variables that have been experimentally manipulated. Here we study people's ability to infer causal structure from both observation and intervention, and to choose informative interventions on the basis of observational data. In three causal inference tasks, participants were to some degree capable of distinguishing between competing causal hypotheses on the basis of purely observational data. Performance improved substantially when participants were allowed to observe the effects of interventions that they performed on the systems. We develop computational models of how people infer causal structure from data and how they plan intervention experiments, based on the representational framework of causal graphical models and the inferential principles of optimal Bayesian decision‐making and maximizing expected information gain. These analyses suggest that people can make rational causal inferences, subject to psychologically reasonable representational assumptions and computationally reasonable processing constraints.     

What children infer from social categories

Children hold the belief that social categories have essences. We investigated what kinds of properties children feel licensed to infer about a person based on social category membership. Seventy-two 4-6-year-olds were introduced to novel social categories defined as having one internal - psychological or biological - and one external - behavioral or physical - property. For half of the participants, the internal property was described as causing the external one; for the others, no causal relationship between properties was mentioned. Children were asked to choose as a novel exemplar of a category one with only the internal or only the external property. Children inferred that exemplars had a psychological property irrespective of causal status, but they inferred the presence of a biological property only when described as causal. Children did not draw systematic inferences regarding either of the two external properties. These findings indicate that children treat psychological and causal properties as central - and perhaps essential - to human kinds.     

Error probabilities for inference of causal directions

A main message from the causal modelling literature in the last several decades is that under some plausible assumptions, there can be statistically consistent procedures for inferring (features of) the causal structure of a set of random variables from observational data. But whether we can control the error probabilities with a finite sample size depends on the kind of consistency the procedures can achieve. It has been shown that in general, under the standard causal Markov and Faithfulness assumptions, the procedures can only be pointwise but not uniformly consistent without substantial background knowledge. This implies the impossibility of choosing a finite sample size to control the worst case error probabilities. In this paper, I consider the simpler task of inferring causal directions when the skeleton of the causal structure is known, and establish a similarly negative result concerning the possibility of controlling error probabilities. Although the result is negative in form, it has an interesting positive implication for causal discovery methods.     

Causal learning mechanisms in very young children: two-, three-, and four-year-olds infer causal relations from patterns of variation and covariation

Three studies investigated whether young children make accurate causal inferences on the basis of patterns of variation and covariation. Children were presented with a new causal relation by means of a machine called the "blicket detector." Some objects, but not others, made the machine light up and play music. In the first 2 experiments, children were told that "blickets make the machine go" and were then asked to identify which objects were "blickets." Two-, 3-, and 4-year-old children were shown various patterns of variation and covariation between two different objects and the activation of the machine. All 3 age groups took this information into account in their causal judgments about which objects were blickets. In a 3rd experiment, 3- and 4-year-old children used the information when they were asked to make the machine stop. These results are related to Bayes-net causal graphical models of causal learning.     

The importance of decision making in causal learning from interventions

Recent research has focused on how interventions benefit causal learning. This research suggests that the main benefit of interventions is in the temporal and conditional probability information that interventions provide a learner. But when one generates interventions, one must also decide what interventions to generate. In three experiments, we investigated the importance of these decision demands to causal learning. Experiment 1 demonstrated that learners were better at learning causal models when they observed intervention data that they had generated, as opposed to observing data generated by another learner. Experiment 2 demonstrated the same effect between self-generated interventions and interventions learners were forced to make. Experiment 3 demonstrated that when learners observed a sequence of interventions such that the decision-making process that generated those interventions was more readily available, learning was less impaired. These data suggest that decision making may be an important part of causal learning from interventions.     

Structure and strength in causal induction

We present a framework for the rational analysis of elemental causal induction-learning about the existence of a relationship between a single cause and effect-based upon causal graphical models. This framework makes precise the distinction between causal structure and causal strength: the difference between asking whether a causal relationship exists and asking how strong that causal relationship might be. We show that two leading rational models of elemental causal induction, DeltaP and causal power, both estimate causal strength, and we introduce a new rational model, causal support, that assesses causal structure. Causal support predicts several key phenomena of causal induction that cannot be accounted for by other rational models, which we explore through a series of experiments. These phenomena include the complex interaction between DeltaP and the base-rate probability of the effect in the absence of the cause, sample size effects, inferences from incomplete contingency tables, and causal learning from rates. Causal support also provides a better account of a number of existing datasets than either DeltaP or causal power.     

Infants infer intentions from prosody

Two studies were conducted to examine infants’ ability to discern intentions from lexical and prosodic cues. Two groups of 14-18-month-olds participated in these studies. In both studies, infants watched an adult perform a sequence of two-step actions on novel toys that produced an end-result. In the first study actions were marked intentionally with both lexical and prosodic cues. In the second study, the lexical markers of intention were presented in Greek, thus providing infants with prosodic but not lexical cues. In both studies, infants reproduced more intentional than accidental actions, suggesting that infants can infer intentions from prosodic cues.     

Young children use pedagogical cues to modulate the strength of normative inferences

Young children understand pedagogical demonstrations as conveying generic, kind‐relevant information. But, in some contexts, they also see almost any confident, intentional action on a novel artefact as normative and thus generic, regardless of whether this action was pedagogically demonstrated for them. Thus, although pedagogy may not be necessary for inferences to the generic, it may nevertheless be sufficient to produce inductive inferences on which the child relies more strongly. This study addresses this tension by bridging the literature on normative reasoning with that on social learning and inductive inference. Three‐year‐old children learned about a novel artefact from either a pedagogical or non‐pedagogical demonstration, and then, a series of new actors acted on that artefact in novel ways. Although children protested normatively in both conditions (e.g., ‘No, not like that’), they persisted longer in enforcing the learned norms in the face of repeated non‐conformity by the new actors. This finding suggests that not all generic, normative inferences are created equal, but rather they depend – at least for their strength – on the nature of the acquisition process.     

Identifying expectations about the strength of causal relationships

When we try to identify causal relationships, how strong do we expect that relationship to be? Bayesian models of causal induction rely on assumptions regarding people’s a priori beliefs about causal systems, with recent research focusing on people’s expectations about the strength of causes. These expectations are expressed in terms of prior probability distributions. While proposals about the form of such prior distributions have been made previously, many different distributions are possible, making it difficult to test such proposals exhaustively. In Experiment 1 we used iterated learning—a method in which participants make inferences about data generated based on their own responses in previous trials—to estimate participants’ prior beliefs about the strengths of causes. This method produced estimated prior distributions that were quite different from those previously proposed in the literature. Experiment 2 collected a large set of human judgments on the strength of causal relationships to be used as a benchmark for evaluating different models, using stimuli that cover a wider and more systematic set of contingencies than previous research. Using these judgments, we evaluated the predictions of various Bayesian models. The Bayesian model with priors estimated via iterated learning compared favorably against the others. Experiment 3 estimated participants’ prior beliefs concerning different causal systems, revealing key similarities in their expectations across diverse scenarios.     

Young children's use of statistical sampling evidence to infer the subjectivity of preferences

A crucial task in social interaction involves understanding subjective mental states. Here we report two experiments with toddlers exploring whether they can use statistical evidence to infer the subjective nature of preferences. We found that 2-year-olds were likely to interpret another person’s nonrandom sampling behavior as a cue for a preference different from their own. When there was no alternative in the population or if the sampling was random, 2-year-olds did not ascribe a preference and persisted in their initial beliefs that the person would share their own preference. We found similar but weaker patterns of responses in 16-month-olds. These results suggest that the ability to infer the subjectivity of preferences based on sampling information begins to emerge between 16 months and 2 years. Our findings provide some of the first evidence that from early in development, young children can use statistical evidence to make rational inferences about the social world.     

A theory of causal learning in children: causal maps and Bayes nets

The authors outline a cognitive and computational account of causal learning in children. They propose that children use specialized cognitive systems that allow them to recover an accurate "causal map" of the world: an abstract, coherent, learned representation of the causal relations among events. This kind of knowledge can be perspicuously understood in terms of the formalism of directed graphical causal models, or Bayes nets. Children's causal learning and inference may involve computations similar to those for learning causal Bayes nets and for predicting with them. Experimental results suggest that 2- to 4-year-old children construct new causal maps and that their learning is consistent with the Bayes net formalism.     

The role of learning data in causal reasoning about observations and interventions

Recent studies have shown that people have the capacity to derive interventional predictions for previously unseen actions from observational knowledge, a finding that challenges associative theories of causal learning and reasoning (e.g., Meder, Hagmayer, & Waldmann, 2008). Although some researchers have claimed that such inferences are based mainly on qualitative reasoning about the structure of a causal system (e.g., Sloman, 2005), we propose that people use both the causal structure and its parameters for their inferences. We here employ an observational trial-by-trial learning paradigm to test this prediction. In Experiment 1, the causal strength of the links within a given causal model was varied, whereas in Experiment 2, base rate information was manipulated while keeping the structure of the model constant. The results show that learners’ causal judgments were strongly affected by the observed learning data despite being presented with identical hypotheses about causal structure. The findings show furthermore that participants correctly distinguished between observations and hypothetical interventions. However, they did not adequately differentiate between hypothetical and counterfactual interventions.     

Levels of causal understanding in chimpanzees and children

We compare three levels of causal understanding in chimpanzees and children: (1) causal reasoning, (2) labelling the components (actor, object, and instrument) of a causal sequence, and (3) choosing the correct alternative for an incomplete representation of a causal sequence. We present two tests of causal reasoning, the first requiring chimpanzees to read and use as evidence the emotional state of a conspecific. Despite registering the emotion, they failed to use it as evidence. The second test, comparing children and chimpanzees, required them to infer the location of food eaten by a trainer. Children and, to a lesser extent, chimpanzees succeeded. When given information showing the inference to be unsound - physically impossible - 4-year-old children abandoned the inference but younger children and chimpanzees did not. Children and chimpanzees are both capable of labelling causal sequences and completing incomplete representations of them. The chimpanzee Sarah labelled the components of a causal sequence, and completed incomplete representations of actions involving multiple transformations. We conclude the article with a general discussion of the concept of cause, suggesting that the concept evolved far earlier in the psychological domain than in the physical.     

Young children discriminate improbable from impossible events in fiction

Can young children discriminate impossible events, which cannot happen in reality, from improbable events, which are unfamiliar but could possibly happen in reality? When asked explicitly to categorize these types of events, 4-year-olds (N = 54) tended to report that improbable events were impossible, consistent with prior results (Shtulman & Carey, 2007). But when presented with stories made up of improbable events, children preferred to continue these stories with additional improbable events rather than with impossible events, demonstrating their sensitivity to the difference between the two types of events. Children were indifferent between continuing these stories with additional improbable events or with ordinary, possible events. Children's differential performance on the story and categorization tasks suggests that they possess some knowledge of the distinction between improbable and impossible but find it difficult to express this knowledge without a supportive context.     

Children infer affiliative and status relations from watching others imitate

We investigated whether young children are able to infer affiliative relations and relative status from observing others' imitative interactions. Children watched videos showing one individual imitating another and were asked about the relationship between those individuals. Experiment 1 showed that 5‐year‐olds assume that individuals imitate people they like. Experiment 2 showed that children of the same age assume that an individual who imitates is relatively lower in status. Thus, although there are many advantages to imitating others, there may also be reputational costs. Younger children, 4‐year‐olds, did not reliably make either inference. Taken together, these experiments demonstrate that imitation conveys valuable information about third party relationships and that, at least by the age of 5, children are able to use this information in order to infer who is allied with whom and who is dominant over whom. In doing so, they add a new dimension to our understanding of the role of imitation in human social life.