The Influence of Child-directed Teaching and Discovery on Children's Generalization of a New Rule

• In Event: Poster Session 07
  In Poster Session: PS 07 Topic - Cognitive Processes

Abstract

Child-directed teaching (CDT) has historically been regarded as caregivers’ primary mode of instruction. As such, previous work has investigated how this form of teaching influences a host of learning outcomes including children’s exploration, imitation, and acquisition of causal rules. Additionally, some theories claim that CDT automatically induces children’s generalization, or extension of newly acquired information to new cases, because of children’s recognition of ostensive cues characteristic of pedagogical interactions (Csibra & Gergely, 2009). However, evidence from the domains of mathematics and causal reasoning (Kapur, 2014; Sobel & Sommerville, 2010) suggest that discovery learning, as opposed to CDT, better supports the initial acquisition of conceptual information which is critical to a learner’s ability to generalize. Thus, we hypothesize that discovery learning, not CDT, will better support children’s generalization of new information.

To test our hypothesis, we have designed an online game in which approximately 50 children (ages 4-6) will learn to activate a virtual machine (i.e., cause the machine to light up and play music) using blocks which vary on several perceptual features (e.g., shape & color). Children are initially presented with example blocks, some causal and some inert, to learn which features are critical to a block’s ability to activate the machine. Half of participants will learn about the example blocks through an experimenter’s child-directed demonstration and the other half will explore and test the example blocks without any demonstration, enabling these children to “discover” the critical block features themselves. Afterwards, children will be tested with a series of novel blocks and will be asked to use evidence about the examples to predict which novel blocks will activate the machine. Thus, our measure of generalization is based on how successfully children can use newly acquired evidence to predict the functionality of novel blocks they have not explored or tested previously. Our prediction is that children who discover evidence about the example blocks, rather than receive instruction, will more accurately predict which novel blocks are functional. We have conducted in-person pilots of this methodology which have heavily informed the design of this online game.

In our analysis, we plan to compare the average performance (i.e., number of correct predictions) of participants in the directed and discovery conditions using independent sample t-tests. In the case that children’s performance scores are not normally distributed, we would use Wilcoxon-Mann-Whitney tests instead. Given our average data collection rate for online studies (about 5 participants per week) and that our measure of interest is easily calculated, we believe it is highly feasible for data collection and analysis to be completed before April’s meeting.

Arguably, the ability to generalize knowledge to new situations is the goal of all learning. Thus, we believe our study has potential implications for how children are instructed, especially in domains with high amounts of conceptual knowledge, such as mathematics. Moreover, our findings will contribute to new lines of work that have shifted away from the Western-centric emphasis on child-directed teaching to investigate other ways in which children learn about the world.

Authors

• Carol Medina, University of Chicago
  Presenting Author

• Amanda Woodward, University of Chicago
  Non-Presenting Author