Human-Centered Explanations of Machine Learning Results

Funding Agency

Federal State of Lower Saxony

"Hybrid Intelligence through Interpretable AI in Machine Perception and Interaction" (HybrInt) is a joint project of the Leibniz University Hannover and the University of Osnabrück. The main goal of the project is to combine the strengths of human and machine intelligence to support farmers in agricultural water management to optimize irrigation.

Within HybrInt, we lead the subproject "Human-Centered Explanations of Machine Learning Results". The focus of our subproject is to explain the machine learning results to groups of people with different expertise in order to enable the idea of hybrid intelligence. To make explanations understandable effectively, human-centered explanations are generated by adapting them to the individual skill level of the explainee. In addition, to provide trustworthy explanations, different knowledge representations such as knowledge graphs, model interpretations, and language models are combined and integrated.
Three central questions are investigated: (1) How can neural language models be trained to adapt explanations individually to a given text style? (2) How can different knowledge representations be integrated on-the-fly into the explanation generation process? (3) To what extent can explanations be guaranteed to be factually correct and at the same time adapted to the text style?

Lead at LUHAI: Prof. Wachsmuth

Funding Program: NDS

Project Period:  2023–2026