91原创

Abstract: The disruptive potential of machine learning (ML) technologies depends on their capacity to perform the tasks of domain experts. Thus, most studies focus on the interactions between domain experts and ML, examining how domain experts respond to, understand, and use these technologies to improve performance on a single task. However, expert or professional work rarely involves one task. Thus, examining  how domain experts collaborate with AI on a predefined task or single decision is a particular and limiting unit of analysis. Drawing on an eighteen-month ethnographic study of how gastroenterologists use ML technologies, we shift the unit of analysis from a task to the full work process. By doing so we observe that ML plays six possible roles – signal, noise, catalyzer, inhibitor, jammed signal, and irrelevant signal – in the work process. We show how the same expert can experience ML differently and examine the consequences on work performance. This multiplicity of a ML technology emerges because we examine temporalities, conditions under which expertise is generated, and the various forms of knowledge used to generate expertise. The study shows that the overestimation of the collaborative potential between humans and ML technologies is likely the result of an underestimation of how tasks are linked, performed, and embedded within a work process and a relational system of expertise.

(This is work co-authored with Marjolaine Rostain from Warwick Business School.)

 

 

About Ruthanne Huising

My research is motivated by the documented difficulty of achieving organizational compliance with regulations, standards, and other ethical and social expectations. To examine this problem, I study the work and professions that are central in shaping how compliance is understood and practiced within and across organizations. Within organizations, I study how prosocial demands are harmonized, translated, and implemented, analyzing issues of expertise and power. Across organizations, I examine governance processes, including the organized spaces in which stakeholders and members of designated professions interact to negotiate the meaning of organizational compliance and develop resources to facilitate organizational compliance. My work has been published in Administrative Science Quarterly, Organization Science, and Regulation & Governance. I am an Associate Editor at Administrative Science Quarterly, a former deputy editor and senior editor at Organization Science, and founder of the Ethnography Atelier (), a collaborative space dedicated to teaching qualitative methods and sharing qualitative research. I received my Ph.D. from the Sloan School of Management at M.I.T. Prior to joining E91原创C I was Associate Professor and William Dawson Scholar at McGill University and Professor of Management at Emlyon Business School.