Example-based Explanations in Machine Learning

In this talk I will give an overview of my work and the open challenges in example-based explanations such as contrastive counterfactual explanations and class prototypes. I will argue for a shift towards more human-centered and goal orientated explanations from machine learning systems. Specifically, this talk will look at how explanations can be misleading if not appropriately grounded and I will discuss the open challenges in evaluating explanations.