A User Interface Solution for Enabling Media Professionals to Create Trusted Datasets

The FAIRmedia research project aims to develop tools and workflows that enable media professionals to create trustworthy, secure datasets for training and validating AI models, with a focus on tasks such as classification and question answering. This paper details the user interface design process for the annotation workflow, prioritizing usability for the target group of non-expert users such as journalists and archivists. Employing a user-centered design methodology, participatory workshops and a design workshop were conducted in order to gather requirements and inform conceptual development, which led to a prototype user interface that simplifies AI training tasks through intuitive design, clear language, and guided interactions via an avatar named Taylor. The paper demonstrates a user-centered design approach for a possible reduction of cognitive barriers and enhancement of media workflows, which is still a work in progress. Future work involves further development and infield user testing of the functional prototype.