Developing Intuitions about the Tradeoffs of Information Sharing in Collaborative Learning

• In Event: Children’s Early Understanding of Group Coordination in Social and Epistemic Conflicts.

Abstract

Humans rely on a variety of social learning strategies to aggregate information from multiple sources. On some accounts, the more informants there are, the more reliable their testimony is in the aggregate, as long as each individual’s testimony is statistically independent from the others (Condorcet, 1785; Larrick, Mannes, & Soll; Einav, 2018). However, groups frequently perform better when individuals communicate and influence one another, particularly when questions can be answered through demonstrative reasoning (Laughlin, 2011); moreover, small discussion groups can outperform large crowds of individual respondents. Our experiments test 7-to-10 year old children’s understanding of the tradeoffs between informational independence and group discussion for answering two kinds of questions.
In Experiments 1 and 2, we presented children and adults with four constraint-satisfaction problems (e.g., children’s Sudoku) and four questions about population preferences (e.g., the most popular fruit in the world). We then gave them a choice between two ways of asking for advice.
In Experiment 1, the choice was between asking five people to talk together in order to answer the question, or asking each to give their own independent judgment. As predicted, older children and adults chose group discussion for the reasoning problems, where “truth wins” (Laughlin, 2011), but individual responding for the popularity questions, where social influence can distort individual intuitions (Lorenz, et al., 2011). Younger children showed more sensitivity to informational independence than we predicted, based on past literature (Einav, 2018); they preferred group discussion for reasoning problems, but showed no preference either way for population preference questions.
Soliciting independent judgments from more individuals may be informative when discussion can distort individual intuitions, but not if discussion allows individuals to answer questions they would be unable to answer alone. In Experiment 2, children and adults strongly preferred the 5-person group discussion for reasoning problems, over even 10 times as many informants responding individually. Yet, all ages believed that the larger crowd would be more helpful for answering the population preference questions. However, it could be that the preference for group discussion is driven by the perceived difficulty of the question rather than the tradeoffs between discussion and informational independence.
In Experiment 3 we again asked participants to choose between 5 people talking together and 50 people answering alone, but presented them 4 difficult perceptual discrimination tasks (for which no demonstrative reasoning is possible, as with population preferences) and 4 easy versions of the same reasoning questions used in Experiments 1 and 2. All ages preferred the group discussion for reasoning, and adults and older children distinguished between reasoning and perceptual problems, but only adults preferred the crowd of 50 for the difficult percept problems. Older children showed no preference either way for the percept problems, and younger children actually preferred group discussion for both problem types.
Together, these three studies suggest that while adults understand the tradeoffs of social influence, children may either overestimate the power of group discussion for solving problems, or underestimate the risks of groupthink and herding.

Authors

• Emory Richardson, Yale University
  Presenting Author

• Frank C Keil, Yale University
  Non-Presenting Author