Technical Briefing: Hands-On Session on the Development of Trustworthy AI Software

Abstract

Following various real-world incidents involving both purely digital and cyber-physical Artificial Intelligence (AI) systems, AI Ethics has become a prominent topic of discussion in both research and practice, accompanied by various calls for trustworthy AI systems. Failures are often costly, and many of them stem from issues that could have been avoided during development. For example, AI ethics issues, such as data privacy are currently highly topical. However, implementing AI ethics in practice remains a challenge for organizations. Various guidelines have been published to aid companies in doing so, but these have not seen widespread adoption and may feel impractical. In this technical briefing, we discuss how to implement AI ethics. We showcase a method developed for this purpose, ECCOLA, which is based on academic research. ECCOLA is intended to make AI ethics more practical for developers in order to make it easier to incorporate into AI development to create trustworthy AI systems. It is a sprint-based and adaptive tool designed for agile development that facilitates reflection within the development team and helps developers make ethics into tangible product backlog items.