Simon Polichinel von der Maase defends his PhD thesis at the Department of Political Science

PHD defence

Candidate

Simon Polichinel von der Maase

Title

"A Lens to Learn Through – Conflict Studies and the Age of Computational Methods".

The thesis

The thesis can be loaned from the Royal Danish Library.

Time and venue

Thursday 23 March, 2023 from 14:00-17:00 at Centre for Health and Society, Øster Farimagsgade 5, DK-1353 Copenhagen K., room 1.1.02. Kindly note that the defence will start precisely at 14:00.

Assessment committee

  • Associate Professor Frederik Georg Hjorth, University of Copenhagen (chair) 
  • Professor Lisa Hultman, University of Uppsala 
  • Professor Nils B. Weidmann, University of Konstanz

Abstract

The aim of this Ph.D. thesis is to explore the potential of using advanced computational methods in the field of conflict studies. It consists of a frame and five articles: One introducing the Bodies As Battleground Dataset (BABD) and four research articles.

The BABD, created as part of the thesis, contains detailed information on approximately 150,000 photographs taken by a photojournalist during the Iraq war from 2003 to 2009. It was compiled using computer vision to infer the visual content of the images. Afterward, data from the Uppsala Conflict Data Program and PRIO grid on the level of conflict and structural features
at the time the images were captured were added.

The second article, From Front-line to Front-page, uses the BABD to investigate which visual features are associated with photographs being chosen for submission to news outlets. The third article, Where Have All the Women Gone, also uses the BABD, here to examine the relationship between the presence of women in images and observed conflict levels.

The final two articles focus on conflict forecasting. The Currents of Conflict shows how Gaussian processes can be used to estimate and extrapolate tempo-spatial conflict exposure patterns and distinguish between long-term and short-term conflict trends. The final article, ConflictNet 1.0, presents a new deep learning architecture for conflict forecasting.