This study examines how political parties and their representatives communicate ideological positions across different platforms, leveraging large language models (LLMs) trained on anonymized texts from party manifestos, official party social media accounts, and individual politicians’ posts. By removing party identifiers, the research evaluates AI’s potential to scale party positions along the ideological left-right spectrum and analyze communication strategies.
The analysis compares ideological scores from these sources with expert-coded measures such as the Chapel Hill Expert Survey (CHES) and individual-level voter perception data. This approach identifies which type of content—manifestos, official communications, or politicians’ posts—most influences perceptions of party ideology.
The study hypothesises that mainstream and extreme parties both present more moderate positions in their manifestos and use more polarized narrative on social media. Additionally, the research explores a potential “social media fringe-to-mainstream pipeline,” where radical ideas shift into broader party narratives.
By incorporating temporal dynamics, including Twitter data from January to June 2023 and positions throughout the parliamentary term, the study captures variations in ideological presentation during and outside election campaigns. These findings contribute to understanding how party ideology is strategically adapted across platforms and the differing roles of mainstream and extreme parties in shaping public perceptions through digital communication.