Adults and Children Rationally Account for Future Numerical Decisions When Choosing to Gather More Information

• In Event: The Dynamics of Children’s Information Search

Abstract

Rather than passively absorb all the information that is available, humans and animals actively explore and selectively attend to aspects of the world (Gottlieb, Oudeyer, Lopes, & Baranes, 2013). What determines when we want to explore and what we choose to attend? Inspired by information theories, recent research in cognitive, developmental, and neuroscience suggest that subjective uncertainty influences an observer’s curiosity and motivation to gather information (e.g., Gottlieb et al., 2013; Kidd, Piantadosi, & Aslin, 2012; Loewenstein, 1994). This source of uncertainty has been described as “a gap in knowledge” that can be resolved by information seeking behaviors. Most of the past work on the role of information gain in driving exploratory choice tends to pit which options a learner chooses to investigate in a problem and considers only the informative utility of action given the current problem (e.g., Schulz & Bonawitz, 2007). However, at a higher level, the observer also needs to decide which problems to solve in the first place – when should they keep exploring, and when should they move on to a new problem? What influences this “meta-decision” process?
To address this question, we developed a novel numerical comparison task that gives participants the option to collect numerical information multiple times before deciding on the largest quantity. If participants only consider current decisions, they should seek more information when the trials are more difficult to decide. However, if participants are able to consider future decisions as well, they may try to balance the cost and value of additional information for the current decision, and potentially conserve cognitive effort for future decisions.
In three experiments in adults (N = 96) and preschool children (N = 60, mean age = 4.79 years), we found that the amount of active information seeking did not simply increase as the decision became more difficult. Instead, there seems to be an inverted U-shaped relationship between trial difficulty and how much one chooses to seek information. Adults and children alike explored the most when the trials were of moderate uncertainty, and explored less when the trials were easier (ps < .001). Importantly, they also explored less when trials were too difficult to solve (ps < .046 with Bonferroni correction).
Interestingly, children’s exploration seemed to recover when the trials were the easiest (p = .039), while adults did not show such recovery. It is possible that in addition to exploring to reduce existing uncertainty, children also explore to confirm their decisions when the trials were of minimal uncertainty. Additionally, prior ANS precision (p < .001), instead of age (p = .34), modulated the effect of trial difficulty on children’s information seeking behavior, suggesting that information seeking is driven by subjective uncertainty.
These results suggest that even preschool-aged children are selecting actions in a way that considers longer-term utilities.

Authors

• Jinjing Jenny Wang, Rutgers University - New Brunswick
  Presenting Author

• Elizabeth Bonawitz, Rutgers University - Newark
  Non-Presenting Author