Challenges of Twitter-based Predictions of Civil Unrest in Latin America

Social media is believed to be responsible for facilitating critical communication often required to fuel momentum preceding the events of civil unrest. The information technology revolution has brought with it an explosion of novel data sources data such as Twitter, Facebook, news/blogs, Wikipedia that didn’t previously exist or were not accessible to researchers. This paper focuses on an ongoing research project that involves tracking open social indicators such as social media and Web searches in order to predict critical societal events (protests, strikes) in targeted Latin American countries before they were reported by local news media [1]. The project requires analyzing billions of pieces of information available through social media, such as tweets, news feeds, and web queries in order to develop models that could generate a number of warnings that will be evaluated on their lead-time; the accuracy of the warning, such as the where/when/what of the alert; and the probability associated with the alert. We explore the ability of Twitter data to act as a predictive signal of civil unrest developing network-based and volume-based statistical models.