The paper titled “Ethically Aligned Stakeholder Elicitation (EASE): Case Study in Music-AI” has been accepted to the New Interfaces for Musical Expression Conference (NIME) 2023. It describes a new tool for engineers to explore the stakeholders and ethical implications of machine learning projects, and we explore it with a number of development teams in Music-Ai.

Our abstract:

“Engineering communities whose development work feeds the current proliferation of artificial intelligence (AI) have historically been slow to recognise the spectrum of societal impacts this work has on a diverse range of stakeholders. Frequent controversies around AI applications in creative domains demonstrate insufficient consideration of their ethical predicaments, but the abstract principles of current AI and data ethics documents provide little practical guidance for development work. Hence, there is an urgent need for pragmatic methods that support developers in ethical reflection of their work on creative-AI tools.

In the wider context of value sensitive, people-oriented design, we present an analytical method that implements an ethically informed and power-sensitive stakeholder identification and mapping: Ethically Aligned Stakeholder Elicitation (EASE). As a case study, we test our method in workshops with six research groups that develop AI in musical contexts. Our results demonstrate that EASE supports critical self-reflection of the research and outreach practices among developers, discloses power relations and value tensions in the development processes, and foregrounds opportunities for stakeholder engagement. This can guide developers and the wider NIME community towards ethically aligned research and development of creative-AI.”