Peers are powerful socializing agents in the lives of adolescents, that is, youth between the ages of 11 to 17 years. Decades of sociological, psychological, and criminological literature have found that youth aggression (e.g., bullying, fighting) and delinquency (e.g., truancy, vandalism, alcohol and drug use) are predominately determined by the behaviors of youth in one's primary friendship. Computer involvement in adolescent networks is growing in recent years, giving rise to new group dynamics and new opportunities to study group interactions and individual preferences.
The work proposed here will develop algorithms and methodologies for inferring adolescent network structure from partial observations about individuals. It will take information collected from small middle schools to infer information about larger groups of students. The PIs will create a set of games for use in the classroom that would collect data for this research while also providing information about class dynamics. The data sets collected through those games will be compared with previous approaches.
This research has potential impact on a wide range of disciplines. Innovations in identifying adolescent network structures will allow better estimation of the role of peers in the rising rates of risky adolescent behaviors. The intellectual partnership between an expert in computer science and one in psychology represents an important step in bringing innovative computational thinking to better understand issues of real-world significance such as youth aggression and delinquency.
An on-line social computer game has been developed to aid social scientists in observing, in a non-intrusive way, the behavior of children and their roles within their peer group. The purpose of this research was to study the capability of such computer games to observe real world interactions and to gather useful information for the identification of bullies. The game's participants solve a collaborative and an adversarial task, and are allowed to communicate only through a chat system that is monitored and logged for later analysis. We found that observable data from the game, such as the amount of messages sent and received (in a chatroom interface accompanying the game), and points transactions, correlates well with questionnaire data while providing more detailed information about the participants interactions. The result is a new tool that alleviates the cost of obtaining data and considerably reduces the fatigue of the subjects while providing sound results. Preliminary results with this tool suggest that a careful further development of this tool will enable finding real-world bullies and victims from simple online interactions. Such a development would have an important effect in informing teachers, reseachers, and decision makers on ways to improve education and the workforce. It is yet to be determined if the tool has statistically significant predictive value, and the research team is hoping to develop the social-science and computer-science theories supporting such a tool in its future research.
