Currently, I am a PhD Researcher at PaCE Patterns of Conflict Emergence: Developing an Automated Pattern Recognition System for Conflict, which is an ERC funded project at Trinity College Dublin. Together with my supervisor Thomas Chadefaux, we work on extracting recurring patterns from times series data on different forms of political contention, most importantly, protest and armed conflict. We argue that the interactions between the government and the challenger, protesters or rebel group, produce recurring temporal patterns in event data. These patterns are indicative of the future and help us to anticipate developments in protest, armed conflict and battle tactics.

In my PhD dissertation, I am studying the emergence of civil war, by differentiating between collective action and the onset of civil war. Variables put forward by the Correlates of War literature explain collective action more broadly, such as protest, riots, terrorism and low intensity armed conflict, rather than the onset of civil war exclusively. Since collective action is a prerequisite for the onset of civil war, these predictors serve to filter out high-risk cases to anticipate conflict onsets. To validate the theoretical arguments, I use tools from conflict prediction and interpretable machine learning, which are particularly suitable to validate civil conflict theory, due to civil war being a high-complexity phenomenon.

Together with my supervisor Thomas Chadefaux and my colleague Thomas Schincariol, we are developing a new conflict prediction model—the Shape finder. Due to the underlying data structure, existing conflict prediction models tend to produce overly conservative predictions, ranging around the mean value which is often close to zero. Indeed, most country-months have zero fatalities from armed conflict, making zero the best guess. Our approach seeks to complement risk-averse conflict prediction models, excelling at anticipating a baseline risk of armed conflict, with a risk-taking approach, focussing on capturing variability in time series of fatalities—or rather sudden escalations and de-escalations in battle activity.

You can find my CV here.