Scientists are currently trying to predict wars and other political conflicts using computer-based text analysis. To this end, they analyse masses of archives and online data.
Dr Karsten Donnay from ETH Zurich is using a computer-based text analysis to analyse media reports and social media posts in an effort to determine the mood of the population based on the “tone of voice”, the frequency of certain terms and changes in the reports.
Thomas Chadefaux is conducting similar research at Trinity College in Dublin. He is using a big data tool to search posts from the archives of Google News and social networks for terms that indicate tensions (“conflict”, “crisis”, “dispute”, “unrest”, “quarrel”, “discord”, etc.). If the occurrence of such words increases over the course of a given week, the software deduces rising tensions; unusually large deviations reveal an increased likelihood of armed conflict. According to Chadefaux, he managed to “predict” the outbreak of 200 conflicts between 1900 and today with an accuracy of up to 85% in this way by analysing 60 million newspaper pages.
In contrast, Dr Kalev Leetaru from Georgetown University in Washington is searching print, radio and online media coverage for events such as an increased frequency of meetings or increased registrations with Facebook and Twitter. His software generates mood barometers, and in the analysis of the 2011 Arab Revolution he successfully identified a violent mood slump from mid-January. The uprising in Egypt began on January 25.
Conflict research using big data analysis is nevertheless still in its infancy. Scientists are currently working on the indicators that signalled conflicts in the past. They have not thus far succeeded in making a prediction that later came true. Critics also question whether the evaluation of media coverage and social networks is indeed a promising approach, since there is no free press in many countries and tweets and Facebook posts are always made only after an event. However, researchers are confident that their work is already improving their understanding of how conflicts arise. Indications from data analysis may actually help make reasonably accurate predictions on political developments possible someday. (Source: Technology Review/rf)