Media and State Stability Lessons Learned Prof. Kathleen M. Carley kathleen.carley@cs.cmu.edu Center for Computational Analysis of Social and Organizational Systems http://www.casos.cs.cmu.edu/
2 America is now threatened less by conquering states than by failing ones. President George W. Bush, 2002 National Security Strategy Report. Of all the states on the map of the world in 1816, nearly half no longer exist today. Tanisha M. Fazal State Death: The Politics and Geography of Conquest, Occupation and Annexation, Princeton University Press, 2007, p.1
Arab Spring August 2013 Copyright 2013 CASOS, ISR, CMU Kathleen M. Carley - Director 3
4 Focus Cyber Mediated State Stability Grounded theory development and testing geo-temporally aware multi-methodological, multicultural, multi-country, multi-level
5 Social Media Revolution Evolution
6 Crisis Data is Big Data Multi-dimensional, and relational data Physical sensor + social Dramatically time variant Time sensitive and tends to atrophy Most input is largely irrelevant data of interest is the needle in the haystack Big, but the issues are about more than scalability
7 Crisis Data Presents Unique Challenges Multiple forms Text, Video, Images, Audio Limits to data access Critical information is hidden by vast irrelevant data Data is multi-lingual and multi-cultural Information grows stale rapidly and response needed immediately Information that is valuable to the responders and analysts and can support resiliency; but, the dark side, this data provides guidance to others about system vulnerabilities The data and data source can themselves be security threats from a cyber-perspective
8 Understanding State Instability Requires Having a Baseline Tweets per 4 hours Mentioning Libya News Mentioning Libya twice daily every 8 hours
Overall Tweet Network Note there are a few sources that are picked up August 2013 Copyright 2013 CASOS, ISR, CMU Kathleen M. Carley - Director 9
Most Retweeted Actors August 2013 Copyright 2013 CASOS, ISR, CMU Kathleen M. Carley - Director 10
Hashtag network Note no direct linkage between Arabic topic-group and English topic-group August 2013 Copyright 2013 CASOS, ISR, CMU Kathleen M. Carley - Director 11
Multi-country Comparison note Egyptian attack is against a background of violent events, Libya and Yemen are anomalous August 2013 Copyright 2013 CASOS, ISR, CMU Kathleen M. Carley - Director 12
13 News Hot Topics Muslims and Islam Embassies and Consulates Protests Terrorism War and Conflict
14 the Movie 1.6% mention movie Large fraction of news mention movie Mentions are after the event
15 Accuracy of Methods Overall, violence belief much more predictive than revolution belief maybe because keywords for revolution harder to define, less often used? Source for timeline of events: http://en.wikipedia.org/wiki/timeline_of_the_arab_sprin g#overview_diagram and similar
16 Violence Belief Change Protests in Egypt begin Jan 2011 Feb 2011 ->large spike in violence, start of civil war in Libya August 2011, Gaddafi s gov practically overthrown
17 Revolution Belief Change Interpretations In general similar conclusions to Violence, but Violence appears to be more predictive Egypt: revolution beliefs increase as protests begin in Jan 2011 Libya: revolution beliefs increase as civil war begins in Libya, and spike again during battle of Tripoli, decreases as revolution comes to end in Oct 2011 Syria: revolution beliefs show large increase from September 2011, Yemen: remained neutral in terms of revolution beliefs
Score Sheet Expectation Tunisia first event Libya and Egypt High violence, revolution, overthrow Egypt Continued violence, revolution Tunisia, Yemen Less violence, revolution, overthrow Syria Increasing violence Qatar, UAE Low violence, low revolution, no state change Kuwait, Oman, Morocco Low violence, moderate revolution, minor state change Iraq, Iran High violence, low revolution Find Yes and spike August 2013 Copyright 2013 CASOS, ISR, CMU Kathleen M. Carley - Director 18 Yes Yes Yes Yemen weak on revolution Yes Yes Yes Oman strongest, Morocco slow Yes
19 Summary Rapid assessment of situation Every 6-12 hours is plenty Having a baseline supported analysis Multi-source is critical But cross-media ontology needed Enables rumor debunking Difficult to disambiguate public sentiment from newsreporting bias Data collection is slower than data analysis News complexity predicts state instability