Authors: Vasiliki Voukelatou, Ioanna Miliou, Fosca Giannotti, Luca Pappalardo. Dashboard design: Daniele Fadda

The world news tells us about peace

Measuring and predicting the Global Peace Index though the world news



There is not well-being without peace. Therefore, the United Nations Development Program (UNDP) has included Peace, Justice, and Strong Institutions in the Sustainable Development Goals (SDGs) list.

Armed violence is on the rise and it is challenging to prevent it [1]. Governments and peacekeeping organisations often have little warning of abrupt changes in peace and safety, while the war expenses for the war-torn countries weaken their economies.

The Expert Panel on Technology and Innovation in UN Peacekeeping recognizes the importance of harnessing the data revolution for the benefit of the international community and peace [3].

To demonstrate the critical role of AI in accomplishing SDG 16, we measure the Global Peace Index (GPI), an official yearly index, at a monthly frequency and we forecast the index 6 months ahead.

The map visualises the predicted monthly Global Peace Index (GPI) as measured with the use of digital news data. We use data from the GDELT database to measure the yearly Global Peace Index at a monthly frequency.

Objective of the study and the map

This map visualises the Global Peace Index monthly values as measured with the use of digital news data. In particular, we use data from the GDELT database to measure the Global Peace Index at a higher frequency. Selecting a country, either from the map or from the countries' buttons below, a user can find both the calculated from the news and the official Global Peace Index monthly values. In addition, the user can compare the predicted and the official Global Peace Index from the plots provided. Last, the user can explore the most important features related to news events that contribute to the measurement of the predicted Global Peace Index.
Frequent estimation updates of peace could flag conflict or war spots months in advance by revealing considerable month-to-month peace fluctuations and significant events that would be otherwise neglected. As a consequence, we believe that this study and this dashboard is valuable to policy-makers, peacekeepers, and the scientific community. Early warnings of peace fluctuations could be valuable for adequate policy making and for lasting peace.

Official Code

Paper - Github Repository

Dashboard - Github Repository

References
  1. World Bank. (2018). Pathways for peace: Inclusive approaches to preventing violent conflict.
  2. Hillier, D. (2007). Africa's Missing Billions: International arms flows and the cost of conflict.
  3. Perera, S. (2017). To boldly know: Knowledge, peacekeeping and remote data gathering in conflict-affected states. International Peacekeeping, 24(5), 803-822.
  4. The Institute for Economics and Peace: Global Peace Index 2020 (2020)
  5. Leetaru, K.: The GDELT Project. LINK (2013)
  6. Lundberg, S. M., & Lee, S. I. (2017, December). A unified approach to interpreting model predictions. In Proceedings of the 31st international conference on neural information processing systems (pp. 4768-4777)
  7. Schrodt, P. A. (2012). Cameo: Conflict and mediation event observations event and actor codebook. Pennsylvania State University.

Frequently Asked Questions

The scientific papers can be found below:
  • Voukelatou, V., Pappalardo, L., Miliou, I., Gabrielli, L., & Giannotti, F. (2020, October). Estimating countries’ peace index through the lens of the world news as monitored by GDELT. In 2020 IEEE 7th International Conference on Data Science and Advanced Analytics (DSAA) (pp. 216-225). IEEE. LINK
  • Voukelatou, V., Miliou, I., Giannotti, F., & Pappalardo, L. (2021). Understanding peacefulness through the world news. arXiv preprint arXiv:2106.00306. LINK

The Global Peace Index is the world's leading measurement of national peacefulness, produced by the Institute for Economics and Peace. The Global Peace Index is captured yearly by institutional surveys and governmental data, such as economic data, police data, etc. [4]. The score for each country is continuous, normalized on a scale of 1 to 5, where the higher the score, the less peaceful a country is. For example, in 2019, Iceland has been the most peaceful country with GPI = 1.072, whereas Somalia has been the least peaceful country with GPI = 3.574. The index is constructed from 23 indicators related to Ongoing Domestic and International Conflict, Societal Safety and Security, and Militarisation domains [4], such as Number of Internal Security Officers and Police per 100,000 People, Exhibit military or police power, etc.

The Global Data on Events, Location, and Tone database (GDELT) is a major news data source that describes the worldwide socio-economic and political situation through the eyes of the news media, making it ideal for measuring well-being and peace [5]. For the prediction of GPI, we derive several variables from GDELT. These variables correspond to the total number of events (No. events) of each GDELT category at country and monthly level. For example, the GDELT variable No. events related to “Exhibit military or police power'' could cover the GPI indicator “Number of Internal Security Officers and Police per 100,000 People''. In addition, the GDELT variables No. events related to “Fight with small arms and light weapons'' and No. events related to “Use conventional military force'' could cover the GPI indicators “Ease of Access to Small Arms and Light Weapons'' and “Volume of Transfers of Major Conventional Weapons, as recipient (imports) per 100,000 people''.
We extract the GDELT variables for the models using the Google Big Query.

The most important variables visualized are calculated through the Average Gain, which is used by the XGBoost models and it is the average improvement in model fit each time the feature is used in the trees. In the paper, we present the feature importance as calculated with the Shapley Additive Explanations (SHAP) [6]. This is the reason you might notice small differences between the feature importance plots visualized in the dashboard and the paper.

The Cameo manual [7] presents all the variables used from GDELT and gives an extended explanation with the relevant examples. In this study we use the 3 digits codes. For example, the news text “...East German Foreign Minister Oskar Fischer will visit Albania in June, the first Warsaw Pact foreign minister to do so since Tirana split with Moscow in 1961, the Albanian embassy said…” is labelled as an “Express intent to meet or negotiate” event. We extract the GDELT variables for the models using the Google Big Query.

We apply predictive machine learning models to demonstrate that news media attention from GDELT can be used as a proxy for measuring GPI at a monthly level. XGBoost shows the highest model performance and thus the values visualised on the map are produced from the XGBoost models. The training's set period for each prediction is 6 years, i.e., 72 months.