Erklärbare Künstliche Intelligenz - Steigerung der Nachvollziehbarkeit überwachter maschineller Lernverfahren

Abstract

Artificial Intelligence is increasingly used in everyday applications. However, its high complexity often leads to a lack of transparency. Explainable Artificial Intelligence addresses this issue by making models and their decisions more understandable. This work contributes to explainable supervised machine learning by introducing a procedure model with five types of explanations and proposing global and local surrogate models to explain both overall model behavior and individual decisions.