Violence, whether it is focused on individuals such as crime or at larger scales such as war, seems to be almost inevitable in our world. However, researchers are asking if…
Violence, whether it is focused on individuals such as crime or at larger scales such as war, seems to be almost inevitable in our world. However, researchers are asking if it can be prevented and if geospatial big data techniques, including machine learning methods, could potentially be used to prevent violence from going out of control.
A recent World Economic Form blog has highlighted varied efforts that attempt to mitigate violence at different scales, with geospatial data often a core feature of different methods discussed in these tools.
Two tools have been recently developed that focus on small-scale acts of violence and harassment. Using crowdsourcing and hot spot mapping, Safecity and HarassMap have been created, which depict recent trends of assaults, sexual harassment, and local crime to help individuals determine areas to avoid.
These tools focus on individual urban blocks, helping users gain a more fine-scale understanding of safety in different areas. The tools use police and individual reports and online data, which helps users better navigate neighborhoods in real-time and determine safety in different areas.
Mapping Violence at a Larger Scale
At larger scales, an initiative that started in 2017 has been developed to help forecast political violence in Africa. The initiative is called Violence Early-Warning System (ViEWS). The effort collects data related to armed conflict involving states and groups, including non-state actors, and violence focused on civilians.
Data are state-level but also within states and at small-scales, with resolution being 0.5 x 0.5 decimal degrees for data captured, helping to forecast violence within regions in states. The current effort uses statistical modeling and machine learning techniques to forecast where violence will occur, with the long-term intent being the interruption of violence.
Using Natural Language Processing to Map Violence
Some early success of the effort demonstrates its potential as a tool for users to better understand where violence could occur. In another effort that focuses on political violence, an effort that uses natural language processing and can be tuned to monitor different media data has been used in conflict and politically volatile zones to assess possibilities of conflict.
This effort, called Ushahidi, initially also focused on Africa but has now spread to more than 160 countries to assess their potential for violence. Types of violence include armed conflict but also less organized acts of violence. Data are mainly crowdsourced and machine learning methods are used to assess sentiment and attitudes that could lead to conflict. Varied tools are created by this effort that are also expanding to assess non-fatal violence.
Increasingly, it has become clear that a variety of data, online and offline, are needed to monitor the potential for violence. The GDELT project is a major effort that is gathering online and offline data at a global scale, helping efforts that then create forecasting tools to assess an determine the likelihood of conflict in different countries and even sub-regions.
The data are also varied, with over100 languages assessed that help identify specific groups, locations, organizations, emotions, and even images that can be used to help demonstrate potential for violence. With image data being captured, this opens up the potential for computer vision techniques to be used in assessing sentiment and attitudes that may lead to violence.
Other efforts have focused on environmental issues that can lead to conflict. For instance, theWorld Resources Institute’s Water, Peace and Security initiativehas created an early warning tool that uses economic, socio-political and environmental data to help forecast potential conflict using random forest machine learning models and techniques around the resource of water.
The initiative is open source and allows contributors and users of the tool to integrate into their own datasets and workflow. The effort attempts to forecast conflicts up to a year in advance, giving ample time to efforts that could potentially be used to mitigate conflict.
What we have seen in recent years are new efforts that are attempting to find new ways to resolve conflict by attempting to forecast where violence could occur, whether at large- or small-scales. These efforts are using geospatial data, often online but also offline data, which increasingly have geospatial information as part of the dataset. Such efforts give policymakers and authorities new tools in efforts to mitigate violence.