Autori
Sampugnaro, Rossana
Marino, Simone

Titolo
L’intelligenza artificiale e il suo contenuto politico. Il caso del conflitto russo-ucraino
Periodico
Comunicazione politica - ComPol
Anno: 2025 - Fascicolo: 2 - Pagina iniziale: 177 - Pagina finale: 210

The present study examines the use of generative artificial intelligence, particularly Large Language Models (LLMs) such as ChatGPT, Copilot and Bard, in a political context, focusing on the analysis of the Russia-Ukraine conflict. The study underscores the potential of these models to influence public perception by generating narratives that may contain distorted or misleading information. The experimental section of the study involved the analysis of responses provided by four AI systems (ChatGPT-3.5, ChatGPT-4, Bard, and Copilot) to twenty questions related to the Russia-Ukraine conflict. The responses were then categorized as either «support» or «non-support» in order to identify any potential biases. Subsequently, a clustering analysis was conducted using T-Lab to examine the specific terms that emerged, grouped into five distinct clusters. The analysis included the evaluation of term frequency within clusters, deviation from expected frequency, and the associated p-value to determine the statistical significance of the observations. The results demonstrate that some AI systems exhibit a marked tendency towards pro-Ukraine positions, while others provide more neutral responses. This study underscores the potential and limitations of LLMs in handling complex and controversial content, emphasizing the importance of understanding inherent biases and the risks associated with the manipulation of information in political contexts.



SICI: 1594-6061(2025)2<177:LAEISC>2.0.ZU;2-1
Testo completo: https://www.rivisteweb.it/download/article/10.3270/116583
Testo completo alternativo: https://www.rivisteweb.it/doi/10.3270/116583

Esportazione dati in Refworks (solo per utenti abilitati)

Biblioteche ACNP che possiedono il periodico