Cultural memory – shared reservoir of historical events, rituals, but also collective traumas – shapes communities by providing a sense of identity and continuity across generations. However, debates about history and commemoration can become deeply divisive, fueling political polarization and social antagonism, particularly, if cultural memory is contested or misused for different political or ideological purposes. Traditionally, the construction of cultural memory has been mediated by institutions (e.g., museums, media outlets). Today, social media allows individuals and grassroots groups to reframe or contest historical and memory narratives on an unprecedented scale. This phenomenon bears directly on democratic stability and deliberative quality. 

To study social media's impact on cultural memory, IRRIS has developed a unique methodology that comprises three analytical components, each theoretically supported by deliberative democracy theory or narrative structure analysis. This methodology utilizes large language models (LLMs) to parse corpora of (social) media posts (Twitter/X and news articles). 

This pipeline was first validated through human evaluation, then tested on an extensive dataset of Slovenian social media posts and news articles covering the Slovenian ideologically charged The Day of Uprising Against Occupation (April 27, national holiday). The main results are presented in Tables 1 and 2.

Based on this methodology, IRRIS was able to clearly identify key narratives in social and traditional media to compare both them in terms of embedded narrative structures and classify them as antagonistic, agonistic, or deliberative, based on the underlying motivation of authors. 

For instance, as can be noted in Table 1, narrative structure 11 in the news articles is marked by a distinct actantial structure. Overall, it has high deliberative function ratings: epistemic 76.9, ideological 72.8, ethical 71.8, and democratic 56.7. Despite exhibiting strong agonism, high epistemic, ethical, and democratic scores indicate an intense yet constructive debate, with well-defined roles and a recognized need to preserve unity. Most similar narrative structures clearly identify an opponent - either “historical distortion” or “societal divisions” - which is challenged in the name of the “Slovenian social future.”

Conversely, narrative cluster 6 in the news article database centers on the framing of Slovenia’s World War II history and its socialist/communist legacy as a distorted narrative imposed by the left, with articles denouncing alleged “communist myths,” highlighting perceived “rewriting” of history, and emphasizing threats to national and cultural identity. It is by far the most antagonistic cluster in our dataset, marked by strong, accusatory language and polarizing rhetoric. The majority of these texts come from the populist right-wing outlet Demokracija, which portrays leftists as undermining true national heritage and glorifying a “socialist past” at the expense of historical facts.

Narrative cluster 9 in the X dataset (Table 2) similarly revolves around an accusation that the official narrative of WWII commemoration is a falsification—portraying the “27 April” holiday not as a day of resistance against Nazi occupiers but as evidence of communist alignment with Hitler under the Molotov–Ribbentrop Pact. Posts repeatedly claim that Slovenia’s leftist tradition has intentionally “recast” Protiimperialistična fronta into a narrative of anti-fascist resistance, thus committing “historical distortions.” Like the parallel news-article cluster, this set of comments is also the most antagonistic in our X dataset, largely echoing right-wing populist discourse that disputes mainstream historical accounts and frames leftist actors as willfully deceptive.

On the other hand, cluster 12, comprising 30 user posts, is particularly notable for its overwhelming support of the established idea of resistance, independence, and sovereignty. This cluster exhibits relatively high deliberative function ratings for dataset X (i.e. epistemic 30.53, ideological 53.6, ethical 39.7, democratic 23.8), suggesting that the discussion is partially structured around reflection and historical interpretation.

Overall, the findings underscore both the transformative power of social media - fueling fragmentation and intense conflicts - and the latent possibilities for public deliberation in an online ecosystem. By mapping out where and how antagonism emerges, the results highlight strategies to mitigate polarizing tendencies, such as promoting fact-based content, clarifying narrative roles, and encouraging respectful exchanges. More broadly, this methodology paves the way for large-scale semantic analyses of other culturally or politically divisive topics, offering a systematic lens to assess how digital discourse can either sustain or undermine the shared historical ground on which democratic societies depend.