(Re)Framing Europe: AI-Driven Analysis of Conceptualisations by Political Elites (1999–2023)

Political elites’ discourses are pivotal tools in the struggle for political hegemony. Among these, empty signifiers —terms like democracy or justice that are widely used but lack fixed definitions following Laclau and Mouffe's terminology— play a key role. This paper argues that Europe itself functions as an empty signifier contested by European political elites. Building on this premise, I analyse all the speeches delivered in the European Parliament from 1999 to 2023 by EU heads of government, heads of state, and European Commissioners, utilising Artificial Intelligence's Large Language Models (LLMs). In total, subdividing speeches in paragraphs, I analyse 45,379. Specifically, I employ ChatGPT —scholarly considered the leading AI tool in text analysis— to conduct two machine-learning-based classifications with human checks. First, I determine whether these speeches conceptualise Europe. I understand conceptualisation as either talking about how Europe should be –prescriptive, or describing what Europe is –descriptive. In the second step, I focus on the speeches that do conceptualise Europe, analysing how they frame it. Using the Comparative Manifesto Project's methodology combined with AI capabilities, I classify conceptualisations into four framings: cosmopolitan, ethno-cultural, economic, and cosmopolitan-communitarian. Each speech is scored on how well they fit to each framing based on word choice, tone (positivity/negativity), and context. Finally, descriptive analyses are conducted using datasets like the Chapel Hill Expert Survey (CHES), accounting for variables such as ideology, year of EU accession, or country of origin. This study serves as a cornerstone for a much broader examination of the determinants and evolution of Europe’s conceptualisation by political elites.