A part of the CORECON project team participated in the third edition of CILIDI (International Conference on Digital Linguistics), held in Valencia on June 26-27, 2025. Aligned with one of the conference’s core themes – the intersection of technology and discourse analysis – Simina-Maria Terian, Monica Borș, Denisa-Maria Bâlc, Iulia-Maria Ticărău and David Morariu delivered presentations within a panel focused on media discourse and social representation.
The panel opened with a presentation titled “Large Language Models and Political Bias: The Case of the Russian-Ukrainian Conflict” authored by Simina-Maria Terian and David Morariu. The authors explored the concept of bias in large language models (LLMs), focusing in particular on ChatGPT, Gemini and DeepSeek. These models were evaluated using five distinct prompts addressing topics such as responsibility for the war, potential solutions and risks posed to other countries. A key methodological component of the study was a qualitative evaluation conducted by both human assessors and AI, which confirmed earlier research findings regarding ideological antinomies between ChatGPT and Gemini on one hand, and DeepSeek on the other.
Next, Denisa-Maria Bâlc presented the paper titled “AI-Assisted News Values Analysis: Trivialization, Sensationalism, and Newsworthiness. But Until When?”, which introduced an AI-based approach to identifying and interpreting news values. Her findings highlighted the promising capabilities of AI in conducting qualitative analyses – both cognitive and linguistic – of news values. However, the presentation also emphasized the inconsistencies observed in large language models when applied to quantitative analysis.
The third presentation, with the title “Unconventional Representations of War Actors, Leading Figures and Affected People in the Russian-Ukrainian Conflict: A Corpus Analysis of Non-Human Animals in Romanian and English-Language Media (2022–2024)” was delivered by Monica Borș and David Morariu. Their study revisited the CORECON corpus through the lens of animal studies and the concept of “citizenship theory” (Donaldson & Kymlicka, 2011). Specifically, they identified linguistic triggers and discursive strategies that portray animals as war victims, thereby reinforcing a discourse of solidarity and legitimizing Ukrainian resilience against Russia.
Closing the panel, Iulia-Maria Ticărău presented “A Brief Analysis of the Role of Multimodality via AI in the News Coverage of the Russian-Ukrainian Conflict in Romanian Mainstream Media” focusing on the pivotal role of images in journalistic discourse. She examined ChatGPT’s interpretations of news related to the Ukraine war under different scenarios: without the news text, with the news text, and by analyzing other images accompanying similar news content. One of her key conclusions highlighted how images reinforce the statements made in the news.
Audience questions sparked lively discussions, enhancing the project’s visibility and inspiring new research directions connected to the themes explored at the conference.