The Rashomon Wikipedia: A Data-Perspectivist Analysis of Divergent Historical Narratives
Authors: Claudiu Creanga, Liviu P. Dinu, Anca Dinu
Organizations: Interdisciplinary School of Doctoral Studies, University of Bucharest, Romania · HLT Research Center, University of Bucharest, Romania · Faculty of Mathematics and Computer Science, University of Bucharest, Romania · Faculty of Foreign Languages and Literatures, University of Bucharest, Romania
Wikipedia aims to provide a unified, neutral record of history, yet its independent language editions often function as distinct epistemic communities, creating divergent narratives around contested events. This paper investigates cross-lingual historiographical bias by analyzing Wikipedia articles across five languages (Romanian, Hungarian, Russian, Turkish, and English) focusing on three contentious events in Romanian history: the Battle of Posada (1330), the Soviet occupation of Bessarabia (1940), and the Night Attack at Targoviste (1462). Using human annotators and Large Language Models (LLMs) to classify citation stance and quantify narrative evolution from 2005 to 2024, we identify a phenomenon of "citation isolation". In the case of the Battle of Posada, only 2 out of 119 citations were shared between language editions, with the Romanian edition exhibiting a 91% pro-national bias compared to the balanced Hungarian edition. Longitudinal analysis reveals that these narratives are volatile and responsive to contemporary geopolitics, evidenced by a significant shift in the Russian framing of Bessarabia in 2024. Finally, we propose a "Peace-Maker" pipeline to automate conflict reconciliation. We demonstrate that while standard prompting leads models to hallucinate consensus, "adversarial" prompting, which explicitly instructs the model to preserve and attribute disagreement, achieves near-perfect neutrality scores.
Figures & tables
Figure 1: Raw count of citations by stance. Note the overwhelming volume of Pro-Romanian citations in the Romanian edition compared to the balanced distribution in Hungarian.
Figure 2: Stance distribution of citations in the Battle of Posada articles. Romanian Wikipedia is 91% Pro-Romanian, while Hungarian Wikipedia is balanced.
Figure 3: LLM Confidence Distribution. The high confidence scores across all stances indicate that the "Citation Isolation" is not due to model uncertainty but reflects clear, distinct narrative framing in the sources.
Figure 4: Evolution of the "Pro-National" score over time. Left: Bessarabia (RO vs RU). Right: Night Attack at Târgovi s , te (RO vs TR vs EN).
Figure 5: Detailed metric comparison for Bessarabia (2024). Note the divergence in "Soviet Justification" and "Deportations".
Figure 6: Deportation mentions over time. Romanian Wikipedia consistently highlights Soviet atrocities, while Russian Wikipedia largely omits them.
Figure 7: Radar chart of narrative metrics for the Night Attack (2024). The shapes illustrate distinct narrative priorities: Turkish (Cruelty-focused), Romanian (Heroism-focused), and English (Balanced).
Figure 8: Comparison of "Vlad Cruelty" scores. Turkish Wikipedia consistently emphasizes Vlad’s brutality much more than the Romanian edition.
Figure 9: Neutrality scores (0-1) for different prompting strategies. The Adversarial prompt achieves perfect neutrality.
Figure 10: Detailed evaluation of generated summaries. The Adversarial prompt excels not just in Neutrality, but in Source Attribution and Conflict Preservation.
Prompt Strategy
Neutrality
Conflict Pres.
Standard
0.47
0.60
Mediator
0.95
1.00
Academic
0.97
1.00
Adversarial
1.00
1.00
Table 1: Performance of prompting strategies. Neutrality and Conflict Preservation are scored 0-1.
Who invented the radio, Russia's Alexander Popov or Italy's Guglielmo Marconi? Was the telephone the achievement of Bell in the United States or Meucci in Italy? Does printing belong to China's Bi Sheng or Germany's Gutenberg? The answer depends not only on historical record but also on language and perspective. We analyse eleven widely used large language models across 21 disputed inventions and discoveries, evaluated in twelve languages and 75,896 responses. While models generally acknowledge that credit is contested, query language systematically affects which claimant is surfaced. Lower-status claimants are more likely to appear when questions are asked in their associated language, whereas dominant Anglophone figures remain stable across languages. These patterns persist after controlling for response length, model differences, historical prominence, and levels of national commemoration. Language thus acts as a switch that activates different national versions of the same history, producing systematically different national memories from the same question. We interpret this as evidence that large language models function as distributed systems of cultural memory, where language conditions which histories become visible, contributing to a computational form of banal nationalism.
William Guey, Pierrick Bougault, Wei Zhang +2
Department of Industrial Engineering, Tsinghua University, Beijing, China · School of Social Sciences, Tsinghua University, Beijing, China · Department of Industrial Engineering, Federal University of Rio de Janeiro, Brazil
Existing datasets on cultural knowledge have been typically created by human annotators familiar with each culture. However, the process requires considerable human effort, presenting a significant challenge to scalability. To address this challenge, we introduce WikiOriginQA, a QA dataset automatically constructed by objectively representing the origins of knowledge through the creation dates of different Wikipedia language editions. Experimental results with the created dataset confirm that LLM performance tends to be biased toward Anglophone cultures. Furthermore, the results suggest the importance of knowledge propagation into the Anglosphere. These findings demonstrate the validity of our automated approach, enabling cultural bias evaluations with superior scalability across diverse languages without extensive manual annotation.
Large Language Models(LLMs) have shown strong performance on tasks with verifiable answers, but many real-world queries involve subjectivity, disagreement, and conflicting evidence. Most existing benchmarks for such queries remain English-centric, overlooking the vast diversity of perspectives held by users worldwide. To systematically study how LLM perspectives vary across languages, we curate PerspectivesQA, a manually verified multilingual dataset of globally contested topics covering 18 languages, and propose ROOTS, a framework that profiles linguistic, perspective, and framing diversity. We evaluate a diverse set of eight models covering both open-source and proprietary families at varying scales, establishing that LLMs hedge their answers in English but commit to positions in lower-resource languages. Larger models maintain a more consistent cross-lingual worldview, while smaller models are more easily swayed. All models also exhibit high sensitivity to minor surface-level input variations, producing notable stance even in larger models.