Planned Protest Modeling in News and Social Media
نویسندگان
چکیده
Civil unrest (protests, strikes, and “occupy” events) is a common occurrence in both democracies and authoritarian regimes. The study of civil unrest is a key topic for political scientists as it helps capture an important mechanism by which citizenry express themselves. In countries where civil unrest is lawful, qualitative analysis has revealed that more than 75% of the protests are planned, organized, and/or announced in advance; therefore detecting future time mentions in relevant news and social media is a direct way to develop a protest forecasting system. We develop such a system in this paper, using a combination of key phrase learning to identify what to look for, probabilistic soft logic to reason about location occurrences in extracted results, and time normalization to resolve future tense mentions. We illustrate the application of our system to 10 countries in Latin America, viz. Argentina, Brazil, Chile, Colombia, Ecuador, El Salvador, Mexico, Paraguay, Uruguay, and Venezuela. Results demonstrate our successes in capturing significant societal unrest in these countries with an average lead time of 4.08 days. We also study the selective superiorities of news media versus social media (Twitter, Facebook) to identify relevant tradeoffs. Civil unrest (protests, strikes, and “occupy” events) is a common happening in both democracies and authoritarian regimes. Although we typically associate civil unrest with disruptions and instability, for a social scientist civil unrest reflects the democratic process by which citizenry communicate their views and preferences to those in authority. The advent of social media has afforded citizenry new mechanisms for organization and mobilization, and online news sources and social networking sites like Facebook and Twitter can provide a window into civil unrest happenings in a particular country. Why study and forecast protests? Our region of interest is Latin America and protest is an important topic of study here, as many countries here are democracies struggling to Copyright c © 2015, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. consolidate themselves. The combination of weak channels of communication between citizen and government, and a citizenry that still has not grasped the desirability of elections as the means to affect politics means that public protest will be an especially attractive option. To illustrate the power of protest in Latin America we need only recall that between 1985 and 2011, 17 presidents resigned or were impeached under pressure from demonstrations, usually violent, in the streets. Protests have also resulted in the rollback of price increases for public services, such as during the ‘Brazilian Spring’ of June 2013. Forecasting protests is an important capability in many domains. For the tourism industry, forecasting protests can support the issuance of travel warnings. For law enforcement, forecasting protests can aid in preparedness. For the social scientist, protests forecasts will provide insight into how citizens express themselves. For the government, a protest forecasting system can help prioritize citizen grievances. Finally, protests can have a debilitating effect on multiple industries (esp. those that rely on worldwide supply chain management) and thus a protest forecasting system can aid in planning and design of alternative travel and shipping routes. Planned protests Our basic hypothesis is that protests that are larger will be more disruptive and communicate support for its cause better than smaller protests. Mobilizing large numbers of people is more likely to occur if a protest is organized and the time and place announced in advance. Because protest is costly and more likely to succeed if it is large, we should expect planned, rather than spontaneous, protests to be the norm. Indeed, in a sample of 288 events from our study selected for qualitative review of their antecedents, for 225 we located communications regarding the upcoming occurrence of the event in media, and only 49 could be classified as spontaneous (we could not determine whether communications had or had not occurred in the remaining 14 cases). EMBERS We are an industry-university partnership charged with developing an automatic protest forecasting system for 10 countries in Latin America. Our system, called EMBERS, has been deployed since Nov 2012 and has been generating forecasts (called warnings or alerts) automatically, without a human-in-the-loop. These forecasts are emailed to a third party (MITRE) for evaluation. Analysts at MITRE organize a reference database of protests (called the Gold Standard Report, or GSR) by surveying newspapers for reportings of protests, and compare our warnings against the GSR to generate a scoring report (evaluation criteria described later). The full EMBERS system has been described elsewhere (Ramakrishnan et al. 2014), including the overall system architecture, data sources used for analysis, and the various forecasting models in EMBERS. EMBERS adopts a multi-model approach, wherein different models are leveraged for their selective superiorities to generate a fused set of alerts. Arguably, one of the best performing models in EMBERS is the planned protest model that detects ongoing organizational activity and generates warnings accordingly. This paper is the first to present this model in detail, including the research issues involved, and how we addressed them in EMBERS. Capturing mentions of protest planning and organization is not as easy as it might appear. First, articles of interest are written in different languages (Spanish, Portugese, French, Dutch, and English). Second, multiple locations are often mentioned in a given article, leading to (natural language) ambiguity about the intended location of the event. Significant reasoning is required to discern the correct protest location. Finally, dates are often described in relative terms, e.g., ‘Sunday’ and thus these vague references need to be resolved into absolute temporal information. Our detection approach combines shallow linguistic analysis to identify a corpus of relevant documents (articles, tweets) which are then subject to targeted deep semantic analysis. Despite its simplicity, we are able to detect indicators of event planning with surprisingly high accuracy. Our contributions are: 1. We present a protest forecasting system that couples three key technical ideas: key phrase learning to identify what to look for, probabilistic soft logic to reason about location occurrences in extracted results, and date normalization to resolve future tense mentions. We demonstrate how the integration of these ideas achieves objectives in precision, recall, and quality (accuracy). 2. We illustrate the application of our system to 10 countries in Latin America, viz. Argentina, Brazil, Chile, Colombia, Ecuador, El Salvador, Mexico, Paraguay, Uruguay, and Venezuela. Our system predicts the when of the protest as well as where of the protest (down to a city level granularity). We conduct ablation studies to identify the relative contributions of news media (news + blogs) versus social media (Twitter, Facebook) to identify future happenings of civil unrest. Through these studies we illustrate the selective superiorities of different sources for specific countries. 3. Unlike many studies of retrospective forecasting of protests, our system has been deployed and in operation for nearly two years. The end consumers of our alerts are analysts studying Latin America. Our results demonstrate that we are able to capture significant societal unrest in the above countries with an average lead time of 4.08 days.
منابع مشابه
Triggering the Protest Paradigm: Examining Factors Affecting News Coverage of Protests
Communication scholars have articulated the concept of protest paradigm to capture the news media’s tendency to portray social protests as deviant, threatening, or impotent. Developments in the media and social environment have led to more diversified media representations of protests, however. Correspondingly, scholars have started to treat the protest paradigm as a variable. Extending this li...
متن کاملThe Dynamics of Protest Diffusion: Movement Organizations, Social Networks, and News Media in the 1960 Sit-Ins
American South in the spring of 1960 transformed the struggle for racial equality. Sociological investigation began within months of the f irst protest (Laue [1966] 1989; Oppenheimer 1963; Searles and Williams 1962; Wehr 1960), and the sit-ins have become an exemplary case in the literature on social movements (Killian 1984; McAdam 1982, 1983; McAdam and Sewell 2001; Morris 1981, 1984; Oberscha...
متن کاملMedia Ethics in the Field of carrying News and Information with an Emphasis on Religious Teachings
Nowadays, as a social reference, news media has a unique contribution in supplying information needs of the human society so that all of the social institutions of life such as economics, politics and culture institutions of a society will be affected directly or indirectly by this media. Therefore, news and information media are very important in respect to ethics, and ignoring the ethical app...
متن کاملA Comparative Review of Hijab Discovery News Coverage in News Media
Purpose: News media play an important role in attitude towards various issues including hijab and hijab discovery. As a result, the purpose of this research was comparative review of hijab discovery news coverage in news media. Methodology: This study in terms of purpose was applied and in terms of implementation method was quantitative. The research population was the hijab discovery news in ...
متن کاملVahid Online: Post-2009 Iran and the Politics of Citizen Media Convergence
An attempt is made to study the social network site, Vahid Online, pseudonym of a leading Iranian activist who has the largest social media followership online. Vahid Online is Iran’s leading distributor of information about social and political news about Iran, a source of information used by citizens and journalists. Similar to Twitter, Vahid Online posts, shares, and communicates news in sho...
متن کامل