CANKO SOCIETY FOR AI AND SOCIAL VALUE

Abstract

This study aims to develop an artificial intelligence (AI)-based model to detect political bias and ideological tendencies in news articles covering Yoon Suk-yeol, the 20th President of South Korea, and Lee Jae-myung, the 21st presidential candidate. In recent years, the issue of media bias in political reporting has gained critical attention, as it influences public opinion, exacerbates political polarization, and potentially undermines democratic values. Against this backdrop, this research seeks to collect and analyze political news data over time—spanning from the pre-election period to the mid- and post-term stages—to identify latent political inclinations in journalistic narratives. The dataset consists of news articles retrieved through web crawling from major Korean media outlets, including both conservative and progressive sources (e.g., Chosun Ilbo, Hankyoreh, JoongAng Ilbo, Kyunghyang Shinmun, Yonhap News, News1), using “Yoon Suk-yeol” and “Lee Jae-myung” as target keywords. Articles published between 2021 and 2025 were collected and preprocessed for textual analysis. Natural language processing (NLP) techniques were employed, including sentiment analysis, topic modeling, and contextual embedding using pre-trained language models such as BERT, KoBERT, and RoBERTa. In addition, metadata such as journalist names, publication dates, and media companies were integrated into the analysis to build a comprehensive bias detection framework. Temporal visualization techniques were used to track shifts in political bias over time. The results reveal that certain news outlets maintain a consistent ideological stance—either conservative or progressive—while some journalists exhibit clear signs of bias through emotionally charged language and framing techniques. The level of bias becomes particularly pronounced during critical political events such as presidential elections, legislative audits, and major judicial investigations. Variations were also found in the intensity and emotional polarity of coverage depending on the political context. This study offers an empirical approach to assessing media neutrality and objectivity in political reporting. The AI-based model developed herein can be utilized to support media monitoring systems, fact-checking platforms, and digital media literacy programs. Moreover, the proposed methodology can serve as a valuable tool for scholars, policymakers, and the general public to evaluate the reliability of political information in the digital age, ultimately contributing to the enhancement of democratic discourse and informed citizenship.

Authors

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