Testing Unpredictability with Multiple Worlds

Cornell University, Ithaca NY

Investigators

Abstract

This project uses a "multiple worlds" experimental paradigm to test the prediction that divisions of opinion between political or demographic groups may be arbitrary and unpredictable as cultural fads have been shown to be. Recent research shows that the popularity of cultural artifacts, such as books, music, and movies, may be contingent on initial opinions. These are often arbitrary and unpredictable, yet subsequently can drive huge opinion cascades. A similar dynamic may explain the divisions that emerge between political partisans or demographic groups (e.g. young and old, men and women). That is, when an initial group of visible partisans disagree, subsequent opinion generators from each side may take this a cue to how they should act, furthering the divide between the groups in a cascading fashion. Findings will serve the national welfare by shedding light on arbitrary social processes that may amplify societal discord and so offering the possibility of ameliorating their effects. The hypothesis of this project is that division between groups can be driven by a dynamic similar to that found by Salganik, Dodds, and Watts (2006) in their investigation of the fickleness of cultural success. They created an experimental music download site, with multiple worlds in which users could only see the downloads of those in their same experimental world. Findings were that the opinions converged more in the worlds where users could see the activity of others around them. To test the hypothesis of this project, a website will be created where experiment participants will be asked to register their political and demographic affiliations. They then will be divided into different experimental worlds, including one independent world (a control). In all worlds, users will vote on various issues. In experimental conditions, users will be able to see the opinions for all in-group and out-group members in their world. By contrast, those in the control condition will vote without seeing other people's opinions. The prediction to be tested is that initial opinion differences between groups will be arbitrary, such that the direction of the association between opinions and groups will vary between worlds, but that the group influence will amplify these differences, leading variations to be stronger than expected by chance. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.