33 Pages Posted: 1 Feb 2017 Last revised: 30 Aug 2017
Date Written: August 29, 2017
A large body of laboratory-based research suggests that people display substantial in-group bias even based on trivial groupings. We use the release of a popular augmented reality game Pokémon Go to study this phenomenon in a hybrid lab-field experiment. We analyze the behavior of over 900 Pokémon Go players randomly matched with other Pokémon Go players to participate in a 1-shot Prisoner's Dilemma game. We find that participants are more cooperative when their partner is from the same Pokémon Go team, demonstrating an ecologically valid occurrence of bias based on a trivial grouping. We also use transformed outcome lasso regressions to look for heterogeneity in treatment effects. Machine learning, rather than unprincipled manual data mining, minimizes overfitting and reduces susceptibility to multiple comparison issues and researcher degrees of freedom. Our approach suggests an important moderator of the in-group bias effect: the salience of Pokémon Go. As people’s interest in playing wanes, so does the size of the in-group bias in our experiments. Thus our full set of results suggest that real-world minimal group bias is quick to arise but also potentially fragile.
Keywords: behavioral economics, in group bias, machine learning
Suggested Citation: Suggested Citation
Peysakhovich, Alexander and Rand, David G., In-Group Favoritism Caused by Pokémon Go and the Use of Machine Learning for Principled Investigation of Potential Moderators (August 29, 2017). Available at SSRN: https://ssrn.com/abstract=2908978