Opening the Black Box of AI: A Sociological Study of AI as a Network

Abstract

This study provides a sociological understanding of the production of AI, which is underexplored in the sociology of AI. To achieve this, the study focuses on the AI development process. Utilizing Actor-Network Theory (ANT), this study demonstrates how the development of AI creates a network consisting of both human and nonhuman actors. The sociological literature focuses on how AI is adopted in various social contexts, identifying the social effects of its introduction and use. We investigate AI itself, showing the values and politics that constitute AI sociotechnical systems in the United States. Based on interviews with software engineers residing in Northeastern USA who work on AI and music platforms, the study highlights how humans and nonhuman actors and social forces such as capitalism and imperialism coproduce AI systems. Engineers' technicality-bound worldview plays a crucial role in their interpretation of AI and the drive for efficiency and profit are foundational values that justify including nonhuman actors such as generative AI platforms and datasets as participants in AI networks. This ultimately results in the production of AI sociotechnical systems that recreate values central to capitalism and imperialism.