Mariel K. Goddu 1, Joshua S. Rule 2, Elizabeth Bonawitz 1, Alison Gopnik 2, Tomer Ullman 1 1 Harvard University 2 University of California, Berkeley
The well-known “explore-exploit” dilemma trades off behavior with two distinct functions. Exploring optimizes for gaining information, and exploiting optimizes for gaining a reward. But, what are children optimizing for when they play? One idea is that play supports “COMPETENCE-BASED CURIOSITY.” If playing optimizes for gaining information about one’s own abilities, children should choose to make a game harder when “playing for fun” vs. “playing to win.” We introduced n=37 participants aged 4-10 (M = 7.5 years, SD = 2.1) to a novel bowling game. We asked children forced-choice questions about how to structure the activity when playing for fun versus when playing to meet a win-criterion. Questions concerned either “goal-relevant” factors that affect the likelihood of winning stickers (which should differ between play-mode vs. win-mode), or “goal-irrelevant” factors (which should not). We calculated a weighted average for play and exploit responses, and a difference score between these averages. Scores for goal-relevant variables (M = 0.33, SD = 0.39) differed significantly from goal-irrelevant variables (M = 0.03, SD = 0.26); t(36) = -3.87, p < .001, and indicated a PREFERENCE FOR DIFFICULTY IN PLAY, t(36) = 5.15, p < .001. For goal-irrelevant variables, there was no difference between play-mode and exploit-mode, t(36) = 0.62, p = .54. Age predicted neither goal-relevant variable scores (r = 0.21, p = 0.20) nor goal-irrelevant (r = -0.02, p = 0.91). These results suggest that CHILDREN CHALLENGE THEMSELVES MORE WHEN PLAYING than when they are exploiting. Ongoing investigations probe this tendency in naturalistic behavior.