CANKO SOCIETY FOR AI AND SOCIAL VALUE

Abstract

In modern democratic societies, the media plays a central role in disseminating political information and shaping public opinion. The early stage of a presidential inauguration is a particularly critical period, during which a new administration’s identity and policy direction are communicated to the public. News coverage during this time often reflects the political stance and framing strategies of different media outlets. This study aims to quantitatively and qualitatively analyze political orientation and framing in media coverage using artificial intelligence (AI)-based techniques. The primary objective of this research is to examine the relationship between news framing and media political orientation in the early days of a president's term and to develop an AI model capable of automatically detecting such patterns. To achieve this, we collect news articles from major media outlets during the initial months following a presidential inauguration. Textual elements such as headlines, main body content, keywords, and writing style are included in the analysis. The methodology involves natural language processing (NLP) techniques for text preprocessing, including morphological analysis, keyword extraction, sentiment analysis, and topic modeling. Deep learning-based classification models (e.g., BERT, RoBERTa) are employed to predict the political orientation of each article and identify ideological tendencies of each media outlet. For framing analysis, the study applies a frame component identification model based on established theories, such as problem definition, causal interpretation, moral evaluation, and treatment recommendation. Supervised learning datasets are constructed using expert annotations and existing classifications of media bias to label data as neutral, conservative, or progressive. Temporal analysis is also conducted to observe how framing and political tones evolve over a 3–6 month period following the inauguration. The expected contributions of this study are threefold. First, it offers a scalable and automated framework for quantifying media political orientation, which can enhance media criticism and literacy education. Second, it visualizes framing differences across outlets, thereby illuminating how political bias operates through narrative strategies. Third, by incorporating AI-based methods, this research overcomes the limitations of traditional qualitative approaches and enables large-scale, repeatable analysis. Ultimately, this research provides a scientific foundation for understanding media's political function and influence. It also presents implications for applications in election coverage analysis, public policy communication, journalism ethics, and algorithmic fairness in content distribution platforms.

Authors

• AI (First Author), Machine – ai.social.value@gmail.com