Get Connected 2019

HCSS Conflict Risk Assessment Project | Mr. Hannes Rõõs, HCSS



Mr. Hannes Rõõs has a master’s degree in sociology from the University of Tartu, and bachelor’s degrees in both political science and sociology from the same university. In addition, he studied at the University of Oslo and the University of Mannheim for a total of three semesters. Prior to HCSS, Mr. Rõõs worked as a research and teaching assistant at the University of Tartu and at the associated Centre of Excellence for Strategic Sustainability.


At HCSS, Mr. Rõõs is primarily involved in quantitative data analysis, contributing to all related activities – data gathering, tidying, storing, transformation, modeling, visualization, and communicating the results of the analyses. Mr. Rõõs also assists in the building of interactive web solutions. His substantive research questions include assessing the risks of future conflict and instability using a data-driven approach.



The Hague Centre for Strategic Studies (HCSS), a small independent think tank, has developed quantitative models that assess the risk of armed conflict on the subnational level for the next one, six and twelve months. Mr. Rõõs presented the Conflict Risk Assessment Project in detail and gave examples of its use. The Conflict Risk Assessment Project applies various machine learning algorithms to a variety of publicly available data sets on conflict history, recent events and structural indicators. Recent events are coded using natural language processing and, in combination with structural indicators (notably political, business, environmental, social, cultural and conflict history), form the basis for calculating the likelihood of conflict in a particular region. The project uses a quantitative approach for risk assessment and therefore applies a singular approach on a global scale. It will have to be supplemented with qualitative data on particular regions.


Mr. Rõõs explained in his presentation why HCSS is running this project and how it has been set up. He elaborated on the risks, the different types of conflict, and the way they are investigated. Following the methodology, he went into the reliability of the predictions and explained how the results can be used.


The enormous amount of data processed allows conclusions to be drawn which would be difficult to conceive, or only with a considerable delay, through purely human data collection and evaluation.


Overall, the pool of predictors exceeds one thousand, the number of observations per region exceeds one million per month, and the number of recorded event mentions is close to a billion. The model portfolio is updated daily to take the latest events on the ground into account. The model could be generalized and applied to other use cases, such as assessing the risk of interstate disputes, terrorism and coup d’états.


However, Mr. Rõõs also mentioned that machine-supported data collection has its challenges. The multitude of international sources places the highest demands on the language software used for translation. Although available language software is constantly becoming more powerful, there is still miscoding. A human sanity check is still needed to verify the results. In particular, the correct classification of historical data while taking developments in the recent past into account, requires common sense which is not yet fully available from machines or systems.


The online Conflict Risk Assessment tool, which can only be accessed through membership, is currently being used by Dutch national airline KLM, among others, to assess the risk of air traffic to and from troubled regions and the risk to crews at and around destination airports.


The HSS Conflict Risk Assessment project has now been running for a year and is constantly being improved.


The tool itself was demonstrated at the seminar to great interest. A preview of the way ahead, accompanied by interested questions from the plenum, rounded off the lecture.


Key take-away

Automated data analysis combined with automatic assessment goes a long way in predicting risk of conflict in given geographical areas. Human validation however is at times required to evaluate machine-generated results.

Comments are closed.

Close Search Window