Abstract
This study introduces a computational framework to quantify the "conversational styles" of political parties on Twitter by mapping sociological theories to mathematical measures. By analyzing German politicians during the 2013 federal election, the authors demonstrate how different practices—like hashtagging, retweeting, and mentioning—reveal distinct patterns of party identity, inter-party debate, and reactions to major media events.
Introduction
While social media research often focuses on network structures (who follows whom) or sentiment analysis (what is being said), there is a lack of integrated methods to measure conversational practices as socio-cultural processes. Political parties use Twitter to define themselves through unique topical focuses and styles, but these are inherently difficult to assess quantitatively. This research bridges the gap between qualitative sociological theory—concepts like homophily, reproduction, and punctuation—and quantitative data science to understand how political identities are constructed and disrupted online.
Quantifying Conversational Style
The researchers follow a deductive approach, deriving five quantitative measures from sociological constructs:
- Cultural Focus ($F$): Uses Normalized Shannon Entropy to measure how concentrated a party’s attention is on specific hashtags or users. High focus indicates a party is strictly "on message."
- Cultural Similarity ($S$): Uses Cosine Similarity to determine the overlap between two parties' referenced hashtags or users.
- Cultural Reproduction ($R$): Employs Rank Biased Overlap (RBO) to measure the stability of a party's focus from week to week, quantifying "style" as a regularity maintained over time.
- Institutionness ($I$): Adapts the h-index to identify "stable facts"—hashtags or users that remain highly referenced over many weeks.
- Burstiness ($B$): Uses Kleinberg’s Burst Weight to detect sudden disruptions in established digital styles, signaling external "shocks" or punctuations.
The Duality of Online Practices
A key finding of the study is the distinction between different types of Twitter interactions. Retweeting and following are highly homophilic; politicians primarily engage with their own party members to demonstrate solidarity. In contrast, mentioning shows significantly lower homophily, serving as the primary tool for inter-party debate and confrontation.
The Pirate Party presented a notable anomaly: despite their "new media" branding, they exhibited the most homophilic behavior and the highest focus, effectively isolating themselves from the broader inter-party "mentioning space."
Temporal Dynamics and the Impact of Events
The study distinguishes between the long-term "style" of a campaign and short-term "punctuations." Hashtagging (tagging) emerged as the most volatile and culturally significant layer, showing a steady increase in focus and similarity as election day approached. Tagging focus increased from 0.13 in week 1 to 0.27 in the election week, before crashing immediately after.
External events, specifically the televised election debate in week 7, acted as a major "punctuation." This event forced parties to temporarily abandon their unique niches to discuss a shared topic (#tvduell), causing a massive but brief spike in cultural similarity across the board. While this disrupted the stability (reproduction) of major parties like the CDU/CSU, it had little impact on parties like the Left or Pirates who were excluded from the broadcast.
Conclusion
This framework provides a robust method for assessing how political actors navigate digital spaces. While limited by the static nature of follower network data and the specific context of the 2013 German election, the methodology offers a scalable way to study the evolution of political discourse and the impact of media events on party identity. Future work could apply these measures to non-political groups or different cultural contexts to further refine the understanding of online conversational practices.
Summary updated