Developing Artificial Intelligence for Good: Interdisciplinary Research Collaborations and the Making of Ethical AI

Abstract

Much of the technical progress in AI has been championed by computer scientists and engineers with little attention to the ways in which people understand and use technology beyond the immediate tasks at hand. Technology is often framed in terms of “socio-technical imaginaries” [2] to better the human condition, but there is growing awareness about its unintended, sometimes harmful outcomes. Research teams have begun to incorporate ethical questions into their design and development processes, lending to the emerging interdisciplinary field of applied “AI for Good” [4]. My dissertation research focuses on the production of AI systems that seek to address social problems. These research initiatives take place across multiple disciplines and industries to collaborate, create, test and deploy AI. Social science research has documented the effects of AI on society, pointing out the ways in which technologies reshape, enhance and present obstacles for social groups. Much of this scholarship focuses on the outcomes of AI and how it contributes to inequality and reproduces bias [1, 3]. However, scholarship also shows that the social organization of scientific production shapes the knowledge and technologies produced, as well as their subsequent effects [5]. In the field of AI research, the question remains how the organization of scientific and technological collaboration influences how researchers define social problems, evaluate technological solutions and put knowledge into practice. Sociology is primed with tools to address questions about collaboration and evaluation, how the social good is understood, expertise is constructed, and people interpret complex systems. In studying