Since the ‘narrative turn’ in the social sciences, questions about how narratives are formed, how they impact, and how they shape political and public discourse have become central across various disciplines, including political science, sociology, psychology, cognitive sciences, communication studies, history, and more recently, computational social sciences. The workshop on ‘Political Narratives in Public Discourse’ held at the Max Planck Institute for Mathematics in the Sciences in Leipzig brought together researchers from different fields that contributed their perspective on the topic that sparked lively discussions at the intersection of sociology, cognitive science, political science and natural language processing.

The first part of the workshop was dedicated to the various concepts and definitions of narrative. First, following work from narratology, narratives can be defined as representations of events. These representations can take different forms, ranging from causal sequences of events that form a ‘story’ or ‘plot’, to cognitive representations by which humans to make sense of their world, to structuralist definitions like Greimas’ actantial schema that conceptualizes narratives according to predefined relationships between actors. Each of those approaches highlights one aspect of narratives and their functions, and relating them to political processes was one of the overarching goals of the workshop.

The second part of the workshop focused on various approaches to extracting (political) narratives, or signals thereof, from textual corpora. Several speakers showcased the use of LLMs for the extraction of various aspects of narratives, for instance for the extraction of causal event chains from text, the measure of ‘narrativity’ of a text, or the extraction of Greimasian schemata from text. Furthermore, an actor-centered approach using Abstract Meaning Representation was presented that could uncover narrative signals from text (https://arxiv.org/abs/2411.00702). Finally, deductive approaches based on seeded topic modelling were shown to prove useful for the extraction of pre-defined narratives, such as political identity narratives on Twitter, or immigration-based narratives in large newspaper corpora.

The final day of the workshop was dedicated to the relationship between narratives and democracy, in particular with deliberation. Following W. Fisher’s narrative paradigm (1984), narratives were discussed as an alternative to rational argumentation. However, this dichotomy is not necessarily valid, as narratives, or stories, and arguments may very well complement each other in a deliberative democracy.