This Open Access book explores the automated identification of media bias, particularly focusing on bias by word choice in digital media. The increasing prevalence of digital information presents opportunities and challenges for analyzing language, with cultural, geographic, and contextual factors shaping how content is portrayed. Despite the interdisciplinary nature of media bias research across fields like linguistics, psychology, and computer science, existing work often tackles the problem from limited perspectives, lacking comprehensive frameworks and reliable datasets. The book aims to advance the field by addressing these gaps and proposing a systematic approach to media bias detection. It develops feature-based and deep-learning approaches for automated bias detection, including a BERT-based model and MAGPIE, a multi-task learning model. These methods demonstrate improved performance on established benchmarks, showcasing the potential of deep learning in detecting media bias. Finally, the author addresses the practical applications of automated bias detection, such as enhancing news reading with forewarning messages, text annotations, and political classifiers, and examines the impact of bias on social media engagement.

About the author

Timo Spinde is a postdoctoral researcher specializing in media bias. He is the founder and coordinator of the Media Bias Group research network. He is affiliated with the University of Göttingen and the National Institute of Informatics (NII) in Tokyo.

This Open Access book explores the automated identification of media bias, particularly focusing on bias by word choice in digital media. The increasing prevalence of digital information presents opportunities and challenges for analyzing language, with cultural, geographic, and contextual factors shaping how content is portrayed. Despite the interdisciplinary nature of media bias research across fields like linguistics, psychology, and computer science, existing work often tackles the problem from limited perspectives, lacking comprehensive frameworks and reliable datasets. The book aims to advance the field by addressing these gaps and proposing a systematic approach to media bias detection. It develops feature-based and deep-learning approaches for automated bias detection, including a BERT-based model and MAGPIE, a multi-task learning model. These methods demonstrate improved performance on established benchmarks, showcasing the potential of deep learning in detecting media bias. Finally, the author addresses the practical applications of automated bias detection, such as enhancing news reading with forewarning messages, text annotations, and political classifiers, and examines the impact of bias on social media engagement.