Poster #7 - 7-year-olds Spontaneously Perform near Optimal Information Search

• In Event: Poster Session 02

Abstract

Asking questions is fundamental to acquiring knowledge (Gopnik et al., 1999; Piaget, 1954). Children know this and frequently use questions to navigate through the ever-increasing stream of information. Sometimes it starts with a hypothesis and sometimes with an open-ended curiosity to learn something new. In other situations, they may face a problem where using questions can help them narrow down the relevant information. One way of assessing the use of questions in a search problem is through the 20-questions game.
Children playing this game can ask two types of questions: Hypothesis-scanning, singling out one particular item, or constraint-seeking, asking about higher-order features shared across items (Mosher & Hornsby, 1966). The latter is often more efficient but can be utilized more or less optimally.
Previous work investigating children’s ability to solve the 20-questions game has demonstrated that children learn to ask the right questions before they know how to use the gained information (Legare et al. 2013), their tendency to use stepwise-optimal strategies (Meder et al. 2019), and that ecological validity aid younger children (Swaboda et al. 2022). However, these studies disclose key game mechanics to participating children, which leads to some uncertainty regarding their spontaneous performance. Therefore, in this study, we specifically investigated children’s spontaneous use of search strategies.

Methods
Participants were 34 children in first grade (M = 93 months, SD = 3.4 months). They played two games, one abstract (Figure 1) and one spatial. Both games had 32 items that could be grouped on various levels based on variations in common features. Half of the children were scaffolded. Arbitrary, non-specified questions were allowed to ensure that we did not influence children’s spontaneous strategy with the instructions. An optimal (split-half) search strategy consistently required 5 questions to solve the task.

Results
On average, 7-year-olds performed close to the optimal 5 questions per game (M = 6.3 questions, Range 4.7 – 14.2, Figure 2) and significantly better than a hypothesis-scanning approach (16 questions), t(-32), p < .001. We found no evidence of performance difference between the abstract and spatial games (p = .27) and no effect of scaffolding (p = .50).

Discussion
A near-optimal performance on the 20-questions game demonstrates that 7-year-olds can spontaneously use constraint-seeking questions. The results add to previous work by showing that children use optimal strategies without guidance or hints (c.f., Legare et al., 2013; Meder et al., 2019; Swaboda et al., 2022), making their performance even more impressive.
In a broader context, these results have implications beyond the scope of optimal search: Near-optimal search performance requires abstraction, pattern recognition, problem decomposition, and an algorithmic sequential solution that relies on recursion: skills recognized as core components in computational thinking, a 21st-century skill (PISA, 2018; Wing, 2006). By extension, our results demonstrate 7-year-olds capacity to spontaneously approach a search problem using the full complexity of computational thinking.

Authors

• Kim Astor, Uppsala universitet
  Presenting Author

• Gustaf Gredebäck, Uppsala universitet

• Agnes Vallberg

• Josefine Rönnlund

• Christine Fawcett, Stockholm University