Towards a Sociomaterial Theory of Interfacing AI Data Infrastructures

Abstract

The implementation of advanced data analytics and AI in organisations to support decision-making has led to the development of rich data infrastructures. In this context, real-time operational data is increasingly perceived as a valuable source of data. However, leveraging this data demands two distinct technologies and respective groups to interact in new ways – Information Technology (IT) and Operational Technology (OT). We study how interfacing IT-OT helps further our understanding of the process of building data infrastructures. Drawing on Pickering’s (1995) Mangle of Practice, and his recent work on cybernetics, we conduct a qualitative study of interfacing IT-OT within a property management company and develop a sociomaterial conceptualization of interfacing which describes the ‘dance of agency’ (human, non-human) in both IT and OT domains to transition from material agency defined by knowledge from human heuristics to material agency from AI.