Over the past two decades, scholars have generated a large and sophisticated literature on genocide. Nevertheless, there are still several research areas that require further work. This article outlines a research agenda that analyzes the conditions under which genocide is likely to occur, the multilevel processes of violent escalation and de-escalation, and the ways in which these processes are shaped by, connect to, reinforce, accelerate and impede one another. I argue that scholars should 1) model elite and follower radicalization processes by disaggregating genocidal “intent” over time and space, and exploring how intent emerges rather than taking it as pre-given. Doing so will permit researchers to 2) situate genocide research within a broader context of political violence in order to understand how they are related temporally and spatially, and to decenter analytical domains beyond the standard country level and single victim group in order to gain insight into the dynamics of genocide, including how perpetrator policies vary by group; 3) draw on recent advances in microanalyses of civil war to theorize about subnational patterns of violence diffusion; 4) move beyond problematic contrasts between ideology and rationality to analyze how ideologies frame the strategic choices “available” to genocidal elites.