Learning From Approach-Avoid Decisions: Children Explore and Learn More Than Adults.

• In Event: 2-193 - Information Gain as a Guide for Decision-Making in Early Exploratory Learning

Integrative Statement

How do children and adults decide to explore, and how does their exploration constrain what they learn? In two experiments, we investigated approach-avoid decision-making in a causal learning task.

In Experiment 1, 62 4- to 7-year-olds and 37 adults completed an approach-avoid learning task (Rich & Gureckis, 2015). On each of 16 trials, participants chose whether or not to test a block on a machine (to approach/avoid). The decision to approach provided information about the block’s effect (positive/negative), but the learner also received a positive or negative reward on the basis of this effect. The decision to avoid allowed the learner to avoid potentially negative consequences, but also precluded information gain (see Figure 1). Blocks varied along two dimensions – pattern and color – and produced positive/negative outcomes when placed on the machine on the basis of a two-dimensional rule (e.g., white striped blocks negative, and all others positive). Positive effects resulted in a one-sticker reward ($0.02 for adults) and negative effects incurred a two-sticker cost ($0.04 for adults). After the learning task, participants completed a test that assessed whether they learned the valence of each block.

Results reveal developmental differences in exploration and learning. Age predicted both the avoidance of negative blocks, F(2, 96) = 42.69, p < .001, and the approach of positive blocks F(2, 96) = 32.30, p < .001. For both positive and negative blocks, children were more likely than adults to approach, thus choosing to pursue information. As a result of this exploration, children were more likely than adults to learn the correct two-dimensional rule, while adults were more likely to learn an incorrect one-dimensional rule, χ2(2) = 24.40, p < .001 (see Figure 2). These results suggest that children may be more exploratory than adults, and as a result can learn about causal outcomes more accurately.

In Experiment 2 (data collection in progress), we disentangle two competing hypotheses for younger children’s higher level of exploration relative to older children and adults. Under one hypothesis, young children are slower learners than adults: while adults learn a one-dimensional rule quickly and subsequently avoid blocks they expect to be negative, children require more data to learn a rule and explore more due to limited expectations about blocks’ effects. Under a second hypothesis, young children are more information seeking than adults: while adults choose to avoid blocks they expect to be negative, children sometimes choose to approach these blocks for information (despite similar expectations about valence). To test these hypotheses, we replicate Experiment 1 but ask participants to predict the valence of the first four blocks, after choosing whether to approach a block but before observing the block’s effect. If children are slower learners, we would expect their predictions to diverge significantly from adults. If children are instead more information seeking, children’s predictions will align with adults despite divergence in their approach-avoid decision-making.

Together, the results from these experiments provide deeper insight into exploratory decision-making across the lifespan, as well as the consequences different exploratory decisions may have for learning.

Authors

• Emily G. Liquin, Princeton University
  Presenting Author

• Alison Gopnik, University of California - Berkeley
  Non-Presenting Author