Symposium Day 1: Panel Description

 

Panel I: Text as Data

Chair: TBD

In recent years, academic research has witnessed a remarkable surge in efforts to diversify and enhance approaches to forecasting armed conflict. Among these innovative methodologies, the widespread adoption of Natural Language Processing (NLP) methods and Large Language Models (LLM) to extract valuable insights from textual data excels. As diverse as textual data is, so is the field of application for NLP and LLM methods in conflict research. This panel investigates three text-based research projects that deal with the following topics: forecasting autocratization processes, generating event data on situations of climate-related conflict and cooperation, and using ChatGPT-4 to support the development of future scenarios.

 

Panel II: Forecasting Instability

Chair: Nadine O‘Shea, Technical University of Munich (TUM)

Conflict's causes and impact appear in diverse guises of direct and indirect forms and cover a wide range of social, economic, political, and security phenomena. The complexity of conflict necessitates an integrated approach to grasp a conflict’s core fully. Hence, different factors, phenomena, and developments serve as a potential data source and indicators to forecast instability and potential conflict risk. This panel investigates topics ranging from data on riots and unemployment to forecasting climate-change-related mobility and the escalation of violence. Moreover, this panel will address the stakes of combining satellite and ground data as data sources for early warning of crises.

 

Panel III: Foresight and Communication

Chair: Boukje Kistemaker, King’s College London

In recent years, foresight methods for risk and resilience analysis in conflict research have increasingly gained more attention. Different foresight techniques encourage researchers, practitioners and decision-makers to profit from iterative, participatory and multidisciplinary discourse and the exploration of innovative approaches and methods. This panel investigates the use of foresight methods to analyse the future impacts of quantum technology on security issues and discusses the role of strategic communication of prediction data to decision-makers. Additionally, as a comprehensive crisis prevention process that profits from data-driven early warning and substantial foresight methods, this panel further investigates the combination of human judgement and model-based forecasts.

 

Panel IV: Predicting Violence

Chair: Paola Vesco, Uppsala University

In recent years, academic research has expanded its approaches to diversify and enhance approaches to forecasting armed conflict. In this surge in efforts, various Machine Learning (ML) techniques have been increasingly employed to capture complex relationships and dynamics inherent in conflict prediction models. In addition, geospatial techniques for discerning spatial patterns indicative of potential conflict zones receive a great deal of attention. Therefore, this panel investigates the usage of a set of different Markov-type latent state models and negative binomial distribution to predict fatalities from state-based conflicts on the country-month level. Moreover, the panel addresses a Gaussian process approach to estimate temporal and spatial trends and transform spatiotemporal data into a 3D shape to analyse complex diffusion of conflict patterns. Additionally, the panel investigates new data sources, modelling approaches and ways of improving transparency in detecting specific events.