In the Mood for Crime: Twitter and the Ambient Population

• In Event: Innovative Techniques for Examining Routine Activities and Situational Perspectives

Abstract

Spatio-temporal shifts in population and their relation to crime patterns are well-documented. In particular, recent studies have shown that ambient (dynamic) population measures are more strongly associated with risks of victimization than residential (static) population measures. This presentation aims to explore a potential use of social media data, in addition to providing estimates of the size of population at-risk: to examine variations of “the mood” of individuals across time and space. Detailed police-recorded crime data for a large Eastern Canadian city was obtained, for a whole year (2016). This data allows hourly counts of infractions, at census tract level. Other studies have used social media data to quantify the ambient population; here, the content of messages sent through one of these platforms (Twitter) is analyzed and put in relation to crime. A classification of messages is presented, followed by the analysis of relationships with geographical variations of crime.

Authors

• Rémi Boivin, Université de Montréal

• Francis Fortin, Université de Montréal