This is a cool article from Jonah Lehrer (The Frontal Cortex) posted at Wired Science. The research Lehrer is discussing seeks to understand causal reasoning in young children (mean age of 54 months). Below the excerpt from his post (which serves as a good overview) is the link to the original research article and a couple of paragraphs from the introduction. I have also included links to a couple of books by Alison Gopnik, one of the leading researchers in this field.
Read the whole article.
- By Jonah Lehrer
- September 28, 2011Pablo Picasso once declared: “Every child is an artist. The problem is how to remain an artist once we grow up.” Well, something similar can be said about scientists. According to a new study in Cognition led by Claire Cook at MIT, every child is a natural scientist. The problem is how to remain a scientist once we grow up.The psychologists conducted their experiments on four and five-year-olds, so they had to be pretty simple. Sixty kids were shown a boxy toy that played music when beads were placed on it. Half of the children saw a version of the toy in which the toy was only activated after four beads were exactingly placed, one at a time, on the top of the toy. This was the “unambiguous condition,” since it implied every bead is equally capable of activating the device. However, other children were randomly assigned to an “ambiguous condition,” in which only two of the four beads activated the toy. (The other two beads did nothing.) In both conditions, the researchers ended their demo with a question: “Wow, look at that. I wonder what makes the machine go?”Next came the exploratory phase of the study. The children were given two pairs of new beads. One of the pairs was fixed together permanently. The other pair could be snapped apart. They had one minute to play.Here’s where the ambiguity made all the difference.
The original research article is available online.
Where science starts: Spontaneous experiments in preschoolers’ exploratory playHere are a few paragraphs from the introduction that looks at the history of causal reasoning in children.
Claire Cook a, Noah D. Goodman b, Laura E. Schulz
Probabilistic models of expected information gain require integrating prior knowledge about causal hypotheses with knowledge about possible actions that might generate data relevant to those hypotheses. Here we looked at whether preschoolers (mean: 54 months) recognize ‘‘action possibilities’’ (affordances) in the environment that allow them to isolate variables when there is information to be gained. By manipulating the physical properties of the stimuli, we were able to affect the degree to which candidate variables could be isolated; by manipulating the base rate of candidate causes, we were able to affect the potential for information gain. Children’s exploratory play was sensitive to both manipulations: given unambiguous evidence children played indiscriminately and rarely tried to isolate candidate causes; given ambiguous evidence, children both selected (Experiment 1) and designed (Experiment 2) informative interventions.
The ‘‘child as scientist’’ account would seem to predict that an additional functional feature of theories – the ability to support informative exploration – should also emerge in early childhood. However, evidence for this seemingly fundamental point of comparison between science and cognitive development, the dynamic by which new knowledge is acquired, has been strikingly mixed. Indeed, education research looking at the relationship between self-guided exploration and science learning has found evidence against the claim that children ‘‘learn by doing.’’ Studies suggest that students have a poor metacognitive understanding of principles of experimental design, difficulty designing controlled interventions, and difficulty anticipating the type of evidence that would support or undermine causal hypotheses (Inhelder, Piaget, 1958; Klahr, Nigam, 2004; Kuhn, 1989; Kuhn, Amsel, O’Laughlin, 1988; Koslowski, 1996; Masnick, Klahr, 2003).One of the authors the paper cites is Alison Gopnik. A couple of her books on this topic include Causal Learning: Psychology, Philosophy, and Computation by Alison Gopnik and Laura Schulz (more academic) and The Scientist in the Crib: What Early Learning Tells Us About the Mind by Alison Gopnik, Andrew N. Meltzoff, and Patricia K. Kuhl (for more mainstream audiences).
Research in science education however, typically investigates students’ understanding of real world phenomena (e.g., density, balance relations, etc.). In such contexts, children’s reliance on domain-specific prior beliefs may mask their formal reasoning abilities (Koslowski, 1996; Kuhn, 1989; Kushnir, Gopnik, 2005; Schulz, Bonawitz, Griffiths, 2007; Schulz, Gopnik, 2004; Sobel, Munro, 2009). Additionally, students are often tested on relatively complex, multivariate problems (e.g., Kuhn, 1989; Masnick & Klahr, 2003). Such problems are appropriate for investigating factors that could affect classroom performance but may underestimate children’s causal reasoning in simpler contexts.
Developmental studies provide stronger grounds for optimism about children’s ability to design informative interventions. Work in fields ranging from perception to motor learning to industrial design (e.g., Adolph, Eppler, & Gibson, 1993; Berger, Adolph, Lobo, 2005; Brown, 1990; Lockman, 2000; Norman, 1988, 1999) suggests that learners discover action possibilities or affordances (Gibson, 1977) in the environment through exploration. Research suggests for instance that toddlers inspect the length and ends of rakes when they need a tool to reach a distant object (Brown, 1990), and the rigidity of handrails when they need to cross narrow bridges (Berger et al., 2005). Similarly, when access to a toy or food is obstructed, toddlers, non-human primates, and even corvids can perform novel interventions to gain information and achieve their goals (Brauer, Kaminski, Reidel, Call, Tomasello, 2006; Emery, Clayton, 2004; Hood, Carey, Prasada, 2000; Mendes, Hanus, Call, 2007; Stulp, Emery, Verhulst, Clayton, 2009). However, children can learn object functions without designing experiments; the ability to intervene on physical features of the environment to gain information does not necessarily entail the ability to intervene when information is unknown because of formal properties of the evidence (e.g., because causal variables are confounded).
The strongest evidence that children may understand some formal principles underlying experimental design comes from research looking at children’s causal reasoning. Studies suggest, for instance, that preschoolers understand patterns of co-variation well enough to distinguish genuine causes from spurious associations: if two variables together generate an effect but only one variable generates the effect independently, children conclude that the other variable is not a cause (Gopnik, Sobel, Schulz, Glymour, 2001; Kushnir, Gopnik, 2005, 2007; Schulz, Gopnik, 2004). Children’s causal judgments are also sensitive to the base rate of candidate causes. When the status of a causal variable is ambiguous, preschoolers are more likely to believe it is causal when causes are common than when they are rare (Sobel, Tenenbaum, Gopnik, 2004). Moreover, preschoolers can draw accurate inferences not only from observed evidence but also from evidence they generate (by chance) in exploratory play (Schulz, Gopnik, Glymour, 2007). Finally, two recent studies (Gweon, Schulz, 2008; Schulz, Bonawitz, 2007) suggest that children’s exploratory play is affected by the ambiguity of the evidence they observe; given confounded or un-confounded evidence about which of two variables controls which of two effects, preschoolers’ selectively explore confounded evidence. Critically however, selective exploration of confounded evidence is advantageous even if children explore randomly (with no understanding of how to isolate variables): the more different actions children perform, the better their odds of generating informative data.