Prediksjon i Internasjonal politikk

The author challenges the consensus in international relations that the field can explain but not predict political phenomena. He argues that there is little theoretical basis for distinguishing between explanation and prediction, and that modern data and machine learning make prediction empirically possible. The paper outlines three generations of conflict prediction, from early behavioralist ambitions to current sophisticated early warning systems used by entities like the UN and EU. The author contends that focusing on prediction improves theory testing by moving beyond statistical significance to evaluate whether models can actually reproduce data-generating mechanisms. While acknowledging the challenges of 'black swan' events and non-linear complex systems, he asserts that the alternative to scientific prediction is often reliance on unreliable expert intuition.

• The author argues against the prevailing view in international relations that the discipline is capable of explanation but not prediction, stating that in most cases there is little theoretical reason to distinguish between the two.
• Conflict prediction has evolved through three generations: the first (1960s-70s) was driven by behavioralism and the Correlates of War project; the second introduced game theory, machine learning, and automated news data analysis (e.g., KEDS); the third is characterized by massive data access and active collaboration with decision-makers.
• Current predictive tools are utilized by major international bodies, including the UN's Department of Political Affairs for genocide prediction, the UK Cabinet Office's 'Strategic Early Warning' unit, and the EU's Global Conflict Risk Index (GCRI) to inform resource allocation.
• The author contends that relying solely on statistical significance (NHST) is problematic and that 'out of sample' prediction is a more robust method for testing theories and avoiding 'overfitting' to idiosyncratic data.
• The paper highlights the risk of 'black swan' events and the difficulty of predicting non-linear complex systems, where outcomes are not proportional to input factors, unlike linear domains such as physics.
• The author asserts a social responsibility for researchers to provide scientific predictions because the alternative is often reliance on 'experts' whose predictive accuracy is frequently lower than a coin flip.

APA

Nygård, H. M. (2015). Prediksjon i Internasjonal politikk. Peace Research Institute Oslo. https://cdn.cloud.prio.org/files/462e80f8-6fba-4556-b52d-0c6cbdfa2213/Prediksjon i internasjonal politikk.pdf?inline=true

Chicago

Nygård, Håvard Mokleiv. Prediksjon i Internasjonal politikk. Peace Research Institute Oslo, 2015. https://cdn.cloud.prio.org/files/462e80f8-6fba-4556-b52d-0c6cbdfa2213/Prediksjon i internasjonal politikk.pdf?inline=true.

Wikipedia

{{cite report |last1=Nygård |first1=Håvard Mokleiv |title=Prediksjon i Internasjonal politikk |publisher=Peace Research Institute Oslo |date=2015 |url=https://cdn.cloud.prio.org/files/462e80f8-6fba-4556-b52d-0c6cbdfa2213/Prediksjon i internasjonal politikk.pdf?inline=true |access-date=23 October 2026 |via=Climate Insights Directory}}

BibTeX

@techreport{nygrd2015prediksjon, author = {Nygård, Håvard Mokleiv}, title = {{Prediksjon i Internasjonal politikk}}, institution = {Peace Research Institute Oslo}, year = {2015}, url = {https://cdn.cloud.prio.org/files/462e80f8-6fba-4556-b52d-0c6cbdfa2213/Prediksjon i internasjonal politikk.pdf?inline=true}, urldate = {2026-10-23}, note = {Indexed by Climate Insights Directory} }