AI-Mediated Systems and the Co-Construction of Decision-Making Practice in a Public Sector Organization: An Ethnographic Analysis of Practice Reconfiguration in the Transport Division's UTURN System

Abstract

This thesis examines how Artificial Intelligence (AI)–mediated systems are enacted, interpreted, and negotiated in a public-sector context. Through an organizational ethnography of the UTurn traffic management system in Trinidad and Tobago’s Transport Division, the study investigates how AI becomes woven into communicative and material practices of decision-making. Rooted in social constructionism and drawing on Practice-Based Studies, sociomateriality, Actor-Network Theory, and ventriloquism, it treats UTurn not as a neutral tool but as a mediating actor. Data were collected via participant observation, semi-structured interviews, ethnographic interviews and document analysis across agencies including the Ministry of Works and Transport, the Transport Division, the Traffic Enforcement Centre Unit (TECU), and the Judiciary. Findings show that the UTurn embeds legal and procedural mandates into daily operations, enabling coordination and traceability while creating tensions around discretion, control, and interpretation of the law. Actors negotiate the system’s constraints, demonstrating that AI-mediated governance is an ongoing, communicative, and material reconfiguration of work practices. The study contributes theoretically by reframing AI in public administration as a sociomaterial mediator, and methodologically by showing how ethnography can surface the lived dynamics of digital transformation. Practically, it offers insights for policymakers and administrators seeking to harmonize technological efficiency with human judgment, institutional legitimacy and legislative frameworks.