Three different aspects of current deep learning algorithms are inherently problematic: the role of bias in training data, the black-box nature of deep layered networks and the fact that the networks are vulnerable to adversarial attacks. These aspects are analyzed from a technical point of view in this project. Alongside, the project also investigates peoples’ preferences regarding the explainability of AI decision making. That is, the project at the same time investigates deep learning algorithms from a technical as well as an ethical view point in order to provide new insights across disciplines.