Answers are grounded in this paper and cite their sources.
This article investigates how standalone Large Language Models (LLMs) analyse and simulate highly polarised, screen-mediated populism in digital memory wars. Using a thread of Facebook interactions (N = 320) surrounding the 2021 removal of the Petras Cvirka monument in Vilnius as a diagnostic testbed, we conduct a human-in-the-loop qualitative audit of ChatGPT through a cultural-semiotic lens. Our findings expose a profound semiotic translation gap, and algorithmic normalisation that sanitises the affective, non-normative, and vernacular friction of online political discourse. We identify a cross-lingual semiospheric inversion in how the model identifies populism: when queried in Lithuanian, the model’s definition classifies only pro-removal supporters as populist; conversely, in English, despite producing a comprehensive definition, its empirical application pathologises legitimate civic preservationists, assigning populism exclusively to monument opponents while remaining blind to pro-removal exclusionary nationalist populism. This double asymmetry demonstrates how commercial LLMs, when running unchecked, can act as biased epistemic engines that flatten cultural complexities of phenomena such as East European memory wars. The study highlights the dual role of AI as both a reflective and potentially reductive tool. It contributes to emerging debates on the role of AI in the study of memory practices, populism, and agonistic interaction in online environments.
No takes yet. Share an insight, caveat, or question.
Dallas et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: