The aim of the project is to establish an open, transparent and user-friendly platform for analysing media bias in news, social media and other text sources.
Using large language models, human feedback and a jointly developed taxonomy, various forms of bias – such as linguistic or ideological bias, hate speech or selective reporting – can be systematically identified. The platform provides specialised classifiers for this purpose, which can be accessed via an open API and a user-friendly web interface.
The project pursues three key objectives:
- New standards for bias detection: A uniform taxonomy enables the consistent definition of different types of bias and their interdisciplinary, comparable analysis.
- Scalable and open tools: Researchers and journalists can use BARI without prior technical knowledge to analyse large volumes of text efficiently.
- Participatory development: via a feedback portal, users can actively contribute to improving the classifiers and definitions.
The BARI infrastructure thus opens up a wide range of potential applications in communication and media studies, linguistics, psychology, as well as in political education and library work.