Media bias is studied across computer science, journalism, political science, psychology, communication science, and related fields. Yet despite growing attention and technical progress, there is still little agreement on how different forms of media bias should be defined, categorized, annotated, and measured. Similar or overlapping phenomena are often described using different terminology, while the same terms may be operationalized differently across studies. This fragmentation makes findings difficult to compare, limits reproducibility, and complicates the development of shared datasets, benchmarks, and computational methods.

This Dagstuhl Seminar will bring together researchers, journalists, AI policymakers, and industry professionals to work toward a clearer shared conceptual foundation for computational media bias research. The seminar will use an initial taxonomy prepared in advance as a starting point for discussion rather than as a predetermined final result. Participants will examine definitions, conceptual boundaries, levels of analysis, and relationships among different forms of media bias, including biases in content as well as those introduced through production, curation, distribution, interpretation, platforms, algorithms, access conditions, and institutional power.

A second focus will be the implications of these conceptual foundations for computational research. Participants will discuss task design, annotation practices, evaluation, datasets, benchmark design, multimodal and machine-learning-based approaches, and the shared resources needed for more comparable and reproducible work. Particular attention will be paid to how technical advances relate to stakeholder needs in journalism, policy, and industry, and to where current research concepts or evaluation practices remain insufficient.

The seminar is designed around short overview talks, plenary discussion, focused breakout sessions, and substantial time for collaborative exchange. Its intended outcomes are a refined and openly available taxonomy of media bias concepts, a report summarizing the state of the art and key research priorities, and concrete directions for future work on annotation, evaluation, benchmarking, and computational methods. Beyond these outputs, the seminar aims to strengthen coordination across an emerging international and interdisciplinary community and to create foundations for continued collaboration, future events, and shared research resources.

Creative Commons BY 4.0

Timo Spinde, Karsten Donnay, Isao Echizen, Gunda Ehmke, and Gianluca Demartini