To improve our understanding of the behaviors of humans and social animals, a more thorough study of conflict is necessary. We argue that the dynamics that occur within groups in the presence of conflict can significantly affect the conventions and norms that are adopted, and that greater insight into these interactions might ultimately be useful in enhancing group stability and cohesion. Historically, the methodologies required to effectively study conflict in group settings have been underdeveloped to deal with the complexities involved. To address this issue, we propose an interdisciplinary approach that employs robust computational and network science methods. First, using agent based modeling, we examine the adoption of conventions in games of conflict when dynamic network learning is present. We find that when agents are allowed to choose their neighbors, the adoption of host-guest norms is strongly favored over the adoption of ownership norms, and the dynamic network topologies that facilitate this difference are heavy-tailed in similar ways to many real world networks. Next, we run an experiment on teams tasked with solving a difficult problem, in which we collect time-stamped head-pose and audio interaction data. Using relational event modeling, we look to explore the dynamics surrounding contentious and conflicting interactions in this dataset. Finally, in another experiment, we test our agent-based model in an online laboratory setting.