To realize the potential benefits of artificial intelligence (AI) in industrial applications Professionals and end-users need to gain insights into the internal processing of machine learning (ML) models. These insights can be obtained through research on eXplainable Artificial Intelligence (XAI), which has received much attention recently. The needs and characteristics of users and their context of their work differ significantly between various industrial applications. Therefore, it is essential to consider various industrial users, use cases, and data, to better understand the context and requirements of appropriate AI solutions. These aspects need to be considered when developing industrial AI systems.
The successful implementation of this idea requires close collaboration between AI use case providers, AI vendors, and research stakeholders from machine learning, XAI, software development, user experience, and human factors. The EXPLAIN project aims to realize an end-to-end ML lifecycle that is interactive and explainable to industry experts. This goal requires in particular the intensive involvement of domain experts. The development also requires advanced MLOps approaches that directly support and incorporate XAI.
The EXPLAIN project will realize:
Conventional machine learning techniques, explanation components, and a feedback system will be developed in the project. The respective components will be evaluated in practice. In addition, various requirements for explainability will be identified.
The Software Systems Engineering (SSE) group led by Prof. Dr. Klaus Schmid is already working on methods and techniques to support the efficient development of qualitative software with artificial intelligence, including the IIP-Ecosphere and HAISEM-Lab projects.
Within the project, the working group will contribute its expertise in software engineering for AI systems and focus on MLOps for XAI in the context of Industry 4.0.
Duration: 3 years
Contact: Prof. Dr. Klaus Schmid
The EXPLAIN project is funded by Grant 01IS22030E from the BMBF.