Explainable AI

The aim of this working group is to develop a comprehensive assessment basis for finding the most suitable and, if necessary, newly developed methods of explainability based on machine learning. This is to ensure that the black box modelling to be developed for the design of active surfaces of forming systems takes into account human-readable explanations and interpretations.

• Analysis of the application potential of existing ML models and methods for forming processes / feature recognition for explainability or congruence with known physical models. models
• If necessary, further and/or new development of ML models and modelling techniques
• Consideration of explainability aspects (xAI) to increase the interpretability of black-box models
• Approaches developed and used in sub-projects for the explainability of model predictions, e.g. based on certain characteristics of time series, are likely to be applicable across contexts. Corresponding synergies are to be identified and utilised in a targeted manner.
• Causal Inference
• Tutorial / workshop on different models
• Create a repository so that data is available during FÖP1
• Development of new data formats if necessary