Making Sense of Evolving Threats: A Novel AI Approach to Monitor Violent Organisations and their Defining Characteristics Over Time from Open Sources

Today’s increasingly unstable geo-political environment presents opportunities for violent organisations to thrive – whether terrorist, criminal, insurgent or militant. It is therefore important for NATO and its nations to monitor and understand potentially threatening organisations as they evolve. At the same time, digitalization has led to an explosion in the amount of open source information, bringing with it analytical opportunities but also the challenge of information overload – whilst developments in Artificial Intelligence (AI) offer new technologies to deal with this challenge. In this paper we build on knowledge of organisational theory and AI to present a first step towards an approach for automatically extracting information about threatening organisations from large amounts of open sources over time. Specifically, we use an existing database of violent organisations as a starting point, and combine three different AI methods for extracting several specific organisational characteristics: (1) Supervised Machine Learning (ML) for extracting activity types and target types; (2) Unsupervised ML for extracting ideology; and (3) Natural Language Processing (NLP) for extracting organisation size, number of leaders, and number of deaths and injured in attacks. We evaluate the performance of these methods and reflect on their generalizability. By doing so we take a step towards tools for automatically monitoring threatening organisations and their defining characteristics as they evolve. Future steps, such as implementing models to ‘fill in’ missing or incomplete information and automatically detecting new threatening organisations as they emerge, are discussed