How Do AI Ethics Principles Work? From Process to Product Point of View

Abstract

Discussing the potential negative impacts of AI systems and how to address them has been the core idea of AI ethics more recently. Based on this discussion, various principles summarizing and categorizing ethical issues have been proposed. To bring these principles into practice, it has been common to repackage them into guidelines for AI ethics. The impact of these guidelines seems to remain small, however, and is considered to be a result of a lack of interest in them. To remedy this issue, other ways of implementing these principles have also been proposed. In this paper, we wish to motivate more discussion on the role of the product in AI ethics. While the lack of adoption of these guidelines and their principles is an issue, we argue that there are also issues with the principles themselves. The principles overlap and conflict and commonly include discussion on issues that seem distant from practice. Given the lack of empirical studies in AI ethics, we wish to motivate further empirical studies by highlighting current gaps in the research area.