Twitter is well recognized as a microblogging site, an online social network (OSN), and increas-ingly as a digital news platform. With the changing media usage behavior over the past decade, political actors have now recognized the need to enrich their election campaign efforts by in-cluding social media strategies. However, previous research has shown that users behave het-erogeneously in online political discussions. To better understand how users behave and inter-act in such debates, we conduct an exploratory study to identify emergent user roles from Twit-ter data. We develop a dynamic selection query to collect a representative data set on the Ger-man federal election of 2017. We define features of structure, function, and time for Twitter dis-cussions and conduct a cluster analysis to derive eleven emergent roles from the 30,553 most active users. We then refine those roles by further data-driven analyses to enhance and deepen their understanding. Our results indicate dominance of the online discussion by the populist party Alternative für Deutschland. We also find that media outlets and political parties show somewhat similar behavior, and that the offline popularity of prestigious actors is extended into the online world.