Generative AI and the framing of Spanish electoral news: cognitive perspectives on political fiction and their translation

Abstract

This study explores the cognitive implications of generative artificial intelligence (AI) in the production and translation of electoral news, with a particular focus on framing effects. Drawing on a corpus of Spanish electoral news generated and translated by ChatGPT-3 (68 news items generated in Spanish and their translated English versions), the research examines how AI influence the perception of political narratives across linguistic and cultural contexts. Framing is a central concept in political communication, shaping how audiences process, evaluate, and emotionally respond to information (Lakoff 2014). When electoral news are produced or translated by AI, subtle shifts in macro-structure, lexical choice, syntaxis, and cultural adaptation can amplify or attenuate particular frames, thereby influencing audience cognition. The research applies both discourse analysis in Atlas.TI and translation quality assessment models (in this case, MQM Core error model) to identify patterns and trends in AI-news creation and translation. The framings that AI prioritises have been identified in this study: The electoral victory of the parties is narrated in a context of a positive emotion, almost euphoric and utopian (political-fiction), with large majorities, an overwhelming citizen participation, and positive reading by political analysts regarding the electoral triumph, no matter which party was called as the winner and for which newspaper it was to be published. AI creates news items in which the framing codes linked to political actions (mostly related to economy, social policy, and sustainability) stand out. ChatGPT-3 can create news items that resemble the genre of the usual news, but are more superficial. AI tries to emulate elements such as the use of sources (authoritative voices, figures, or data) in order to give credibility to the pieces of news that it creates. However, it does not always do so accurately, due to its lack of pragmatic knowledge. This weakness becomes even more prominent in translated news which exhibit culture, terminology and style errors. The study contributes to ongoing debates about the cognitive impact of multilingual media environments and the role of AI in shaping journalistic narratives.