The Bias Analysis Research Infrastructure (BARI)

• Applicants: Professor Dr.-Ing. Bela Gipp; Professor Dr. Michael Granitzer; Professor Dr. Philipp Wieder
• Subject Area: Data Management, Data-Intensive Systems, Computer Science Methods in Business Informatics ; Methods in Artificial Intelligence and Machine Learning
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 565115197

The BARI project aims to create a scalable platform for transparent analysis of media bias in textual content, designed to be accessible to non-experts. Media bias, or one-sided reporting, can shape public perception by presenting particular viewpoints or omitting key information. Such bias can take many forms, including linguistic bias, hate speech, selection bias, phrasing bias, and coverage bias, often linked to disinformation. While many academic fields perform various bias analysis approaches, identifying and categorizing media bias is complex, typically requiring manual annotation or machine learning models based on unclear definitions of bias. Large language models (LLMs) add further complexity due to ambiguous definitions of bias and limited high-quality annotated data. BARI seeks to overcome these challenges by developing a standardized, transparent, community-driven media bias taxonomy and leveraging human-AI collaboration to improve the accuracy and adaptability of bias detection. Using LLMs as annotation agents, BARI will use the created taxonomy and its definitions to supply specialized classifiers for different types of bias, helping researchers across all domains. BARI will first offer a freely available API for bias classification in news content and later be extended to other media types such as social media, websites, and outputs from LLMs like GPT. The platform will support interdisciplinary research across different fields, e.g., journalism, social sciences, computer science, and information science, allowing researchers to analyze bias with minimal training. By providing a standardized, transparent methodology for media bias detection, BARI aims to improve the reproducibility and credibility of research integrating any information about media bias. It addresses the gap in research infrastructure for media bias analysis, thereby contributing to developing robust, accessible systems for studying bias in today’s rapidly changing media landscape.

• DFG Programme: Research data and software (Scientific Library Services and Information Systems)
• Co-Investigators: Dr. Daniel Kurzawe; Dr. Timo Spinde