"How do external actors in international relations respond to impending or ongoing violence against civilians? Does the magnitude or type of victimization in atrocity violence affect the political will of external actors to respond? If so, what is the mechanism that links violence to external action? Does the 20th Century Liberal human rights narrative of "never again" translate into empirical evidence of political action in individual cases, or do Realist calculations of strategic interest framed through risk and reward drive decision-making of external actors to respond? This thesis explores a persistent theoretical blind spot in the literature on mass atrocities in international affairs: the centrality of political will in preventive and responsive action. While multiple domain scholars have explored the internal causes and consequence of mass killing and genocide since at least the end of the Cold War - including both relatively recent insights into micro-level patterns of violence and resistance, and developments in sophisticated machine forecasting of potential events - to date there has been limited empirical exploration of the concept of external political will, what accounts for variation in response among observers in the face of violence against civilians, and its uneven role in the development of mass atrocities. This thesis targets the gap and builds upon theories of decision-making and the conduct of mass atrocities by attempting to explore one part of the prevention equation - political will to respond. I operationalize political will as an ordinal outcome variable, expanding upon its narrow, dichotomized conceptualization, capturing a more nuanced construct. Leveraging advancements in the collection of disaggregated, event-level data on mass atrocities since 1996, I build an atrocity event database matched to discrete external response to test hypotheses about what variables catalyze specific levels of political will in individual atrocity events. I test hypotheses of response using risk and reward as a treatment in a sub-national matching design to determine its effect on subsequent patterns of external response (or lack thereof) across matched pairs of similar regions."