Children’s Motivations and Strategies in a Repeated Prisoner’s Dilemma

• In Event: 3-210 - Game Theory in the Playground: The Development of Cooperative Problem-Solving in Children

Integrative Statement

Reciprocity is a key mechanism for the evolution of cooperation (Axelrod & Hamilton, 1981) and a foundation of social interactions (Gouldner, 1960). Children engage in reciprocal cooperative interactions as toddlers (Dunfield & Kuhlmeier, 2010) and preschoolers (Sebastian-Enesco & Warneken, 2015), but little is know about their motivations for cooperation in these contexts and what they believe about the partner’s motivations. In this study, we investigated children’s motivations and behavior using a Repeated Prisoner’s Dilemma (RPD) game, a standard tool from game theory.
We tested 9-11-year-olds (N=167) using a visual RPD interface which has been used successfully with this age range (Blake et al., 2015). Children were recruited from schools and a university family database. Each participant played a ten-round RPD game with three pre-programmed partners which used different strategies. In each round, the child and the partner decided to either cooperate (C) or defect (D) at the same time, creating four possible payoffs in points: CC = 3 points each; DD = 1 point each; and CD and DC = 0 points for the cooperator and 4 points for the defector. Players thus have an incentive to defect in order to maximize their points and protect themselves against getting zero. After playing all partners, children were asked open ended questions about why they and the partner cooperated and defected in the game. Responses were coded into three non-exclusive categories (Cohen’s kappa for inter-rate agreement was .72): Strategic Self-interest (gain most points), Interpersonal (paying back), and Prosocial (to be nice).
The majority of children (81%) believed that the partner was motivated by Strategic Self-interest, but only 45% said this was the reason for their own decisions. Interpersonal motivations were common for both the partner (37%) and self (35%). Interestingly, 50% of the children attributed Prosocial motivations to the partner, but only 10% said this motivated their own decisions. We next added these codes as dummy variables to multi-level logistic regression models to determine which motivations predicted cooperation across all rounds (Figure 1). For the child’s own motivations, Interpersonal justifications predicted increased cooperation (B = 0.95, p < .001) and Strategic Self-interest predicted decreased cooperation (B = -0.57, p = .001). Attributing Interpersonal motivations to the partner also predicted increased cooperation (B = 0.61, p = .001), but the other motivations had no effect.
We also assessed whether children’s decisions approximated several standard game theoretic strategies: TFT cooperates on the first round and then mirrors the partners last decision; GRIM cooperates until the partner defects and then defects continuously; WSLS repeats the prior decision if the player earns a high payout (CC and DC) or else switches to the opposite decision. All three strategies predicted children’s decisions in the game: TFT: B = 1.20, p < .001; GRIM: B = 0.49, p < .001; WSLS: B = 0.81, p < .001.
Combined these results demonstrate that, by about 9 years of age, children attribute sophisticated motives to interactive partners in the RPD and use reciprocal strategies to resolve this cooperative dilemma.

Authors

• Meia Chita-Tegmark, Boston University
  Non-Presenting Author

• Peter R Blake, Boston University
  Presenting Author