The next speaker in this session at the DigiWorld network ECREA preconference on election campaigning in Brno is my colleague Tariq Choucair, whose focus is on assessing polarisation amongst political leaders during election campaigns. The emphasis here is on affective, identity-based, and relational polarisation: the presentation of ‘us vs. them’ and sectarian in- and outgroup perspectives.
Computational methods might be valuable for this, but current methods do not fully capture this; they do not usually capture whom polarised language is directed against, and manual methods are commonly used instead. This paper develops this further, and applies this to a study of electoral campaigning in Peru 2021, Australia 2022, Brazil 2022, and Denmark 2022, focussing especially on the two leading candidates in each election.
Key innovations in this work are to improve the identification of rhetorical constructions of the in-group; assessing different levels of affiliation and opposition towards in- and out-groups rather than merely identifying support or opposition; and identifying the out-group target of polarising rhetoric. This is operationalised through an analysis of mentions of other entities within the political actors’ texts; an assessment of affiliation (from strong support to strong opposition) between the actor and that entity; and an assessment of the overall patterns for each actor. This is done using LLMs, with human oversight.
This produces a diverse set of patterns across the four actors; Australia’s Albanese shows less hard affiliation or oppositions than Peru’s Castillo, for instance; and the types of opposition groups being targetted through hard opposition (or hard affiliation) also differ substantially – for some, they are merely other political actors, but for others they are the judiciary, journalists, scientists, other countries, and societal minorities.
