Computational Modeling of Grievances and Political Instability Through Global Media

Investigators:

Project Details

Abstract:

The relationship between political, economic, and social grievances and political instability has concerned policymakers and scientists for more than a century. To date, research has been limited to comparative analysis of countries and a limited number of social surveys conducted within select countries; such traditional methods are well-established, however they are also expensive, labor intensive, and slow. The events of the ‘Arab Spring’ – the sudden outbreak of mass protests across North Africa and the Middle East that began in 2011 and led to the overthrow of regimes in Egypt, Tunisia, and Libya, a brutal and prolonged civil war in Syria, and severe crack-downs in Bahrain - provide a stark example of the weakness of these conventional research approaches to rapid, dynamic events. The Arab uprisings caught both policymakers and academics by surprise, despite the fact that these events appear to be rooted in large part out of grievances that built over decades of autocratic rule, widespread corruption and economic stagnation. The euphoria of social media activism that came with the uprisings has now evolved into social media realism, as upheaval has not necessarily translated into social or political change. Rather, it has led to diverse outcomes across countries that may or may not have an impetus in social media.

Hundreds of millions of people around the world use social media to communicate, making this technology-enabled forum a major de facto platform for political participation, expression, advocacy, and mobilization: but to what end, for political or social change, or violent collective action? This project aims to exploit the recent availability of worldwide, individual-level data from social media and web-aggregated news sources to assess the possibility of measuring perceptions of “grievances” at the micro-level and in real time as a means of forecasting instability. Utilizing a multidisciplinary approach that incorporates researchers from the fields of computer science, mathematics, and the social sciences, this research will generate theoretical, methodological, and empirical advances. By triangulating measures of inequality, perceptions, and sentiments across social media, surveys, online new sources, and traditional databases, the project evaluates their relative strength in terms of ascertaining and measuring grievances to forecast political instability. In addition, the widespread availability of online news reports offers the ability to collect content from newspapers and other print media worldwide and code for perceived grievances, We seek to develop analytical methods for triangulation with social media data, to improve inference.

Primary Findings:

In the first phase of the project, we intensively analyzed a single case study — the Boston Marathon Bombing — to improve inference about how terrorist attacks impact public attitudes through different media. Preliminary findings on attitudes and sentiments relating to grievance stemmed from a quasi-experimental research design using multiple methodologies/ data sources to examine changing attitudes around one-sided violent events. These analyses yielded different temporal patterns in public response: surveys show the attitudinal changes hold for more than a year, while social media analytics show the impact dissipating after a few days. The analyses determined what type of actions respondents were motivated to take in the wake of one-sided violence, conditional on their demographics, their attitudes toward perpetrators and state authorities, and their emotional responses to the experience of a terrorist attack. Theoretically, we found that terrorist attacks and their critical aftermaths, through social media, offered empirical opportunity windows of “big data.” These windows are capable of revealing adaptive responses, leading to a better understanding of the connection between motivations and actions, such as grievance and conflict. Methodologically, we found that triangulation of data sources provide higher resolution information landscapes about the relationship between attitudes and violent events, how one affects the other and vice versa. This research provides a theoretical foundation for the next phase of the project, on grievances (“motivations”), opportunity (“mobilization”), and political instability (“contentious actions”).

In the second phase of the project, we aim to predict how motivations, conditional on opportunity, drive contentious behavior. To this end, the project’s methodological work on political instability developed the MELTT method, which aggregates conflict event datasets together to provide the most encompassing look at conflict data, disaggregated from the country-year to the most local unit of analysis. This research relies on proven methodologies on assignment problems and robust technical solutions from computer science to integrate conflict event datasets by location and type. This dynamic conflict event dataset merges the many prominent, open-source geo-coded datasets available and allows researchers two major advantages. The first is integration to allow for a greater range of types of events. The second is disambiguation, where, through iteration through multiple observations, researchers can determine the uniqueness of events, and compile information surrounding them. Researchers input criteria into the software, including where (within x kilometers), when (same or within x days), and what type (from protest to conflict), and with what actors (violent, non-violent, religious, state, etc.). With this computational solution, we aim to give researchers a greater degree of confidence and clarity about patterns of dynamic contentious events and conflict.

Methodology:

Preliminary research on the independent variable, or how attitudes and sentiment produce motivations, triangulates quantitative analyses of cross-sectional, panel survey data, and data from a survey embedded within an experiment, to assess changing attitudes and emotional reactions of publics in response to terror attacks, as well as the resulting actions they take in information-sharing, information-seeking, protective action-taking, and willingness to help police. Sentiment analysis of Twitter data before and after one-sided violence demonstrates how socially-mediated environments alter the way publics respond to crises.

The second phase of the research, using lessons learned from triangulating methods and data on attitudes and sentiment, will computationally model the manifestation of grievances in Sub-Saharan Africa. We will disaggregate the components of grievance to include objective factors such as structural inequalities and political exclusion, subjective perceptions such as unfairness and discrimination, affective reactions such as anger or fear, and a target of blame. Further, we add that the impact of an event itself on attitudes may contribute to – or temporarily reveal – what the literature refers to as “grievance.” This new conceptualization of grievance and contentious actions will be computationally modeled with survey data from the Afrobarometer, data from electronic and social media (Twitter) in Africa, and finally the dynamic dataset of conflict events in Africa ranging from non-violent protests and strikes to civil conflict and one-sided violence.

This research will first engage in a pilot study across cases of different events of grievance. Ultimately, it will engage in a cross-country, time-series analysis using geo-coded structural indicators, survey responses, and social media data (Twitter) to identify a signal of grievance. Finally, the grievance indicator will be modeled with the dynamic conflict event-data to forecast instability.

Timeframe

Project Period:
to