Human-centered AI

The team aims to combine methods of AI, data visualization and human-machine interaction to support the development of data-driven clinical decision models. Visual analytics techniques are used to process and explore complex data with the integration of expert knowledge. The aim is to identify correlations, patterns and artifacts in order to verify existing hypotheses and generate new ones. Visual analytics methods also support the design and evaluation of AI models by giving developers a deeper understanding of the impact of design changes on performance.

The combination with automated machine learning (AutoML) enables efficient suggestions for new AI designs and provides insights that are important for data processing steps and model decisions. AutoML can also draw on existing expert knowledge to provide targeted support in AI development. Finally, visual analytics methods should help to make the AI-based prediction of decision models understandable and comprehensible for various target groups, from developers to clinicians and patients.

The junior research group's work will initially focus on oncology issues, particularly in the use case of AI-based image analysis for individual therapy decisions. In collaboration with other junior research groups, further applications will be successively tackled.