The development process of Responsible AI: The case of ASSISTANT

The development of artificial intelligence (AI) for manufacturing comes with the question of why and how to make sure that the resulting system acts responsibly. In this paper, we discuss why responsibility matters in the context of AI in manufacturing and introduce an approach to put these considerations into practice of an AI development project. We will start with the observation that many abstract guidelines and approaches exist to address responsibility issues in AI contexts, but the question remains of how these guidelines can be translated into practice in a development project. Therefore we will review different approaches towards issues of responsibility, including'ethics by design', responsible research and innovation, as well as methodological approaches such as design for values and human-centric design. These approaches informed the theoretical framework we created and adapted within the AI development project ASSISTANT. Our contribution encompasses an ex-ante and an ex-post approach. The ex-ante approach aims at making sure that issues of responsibility can be formulated and discussed in all phases of the development process and therefore draws on a design for values methodology. We aim at making sure that questions of responsibility are considered in the course of the development process.