This dissertation investigates the socio-cultural and institutional factors - or determinants - shaping the design and implementation of automated decision-making (ADM) systems in public services, through a case study in Italy and the Netherlands. It engages with the contemporary discourse surrounding artificial intelligence (AI), real cases of failure of automated discriminatory practices and high-risk uses, highlighting the narratives of inevitability and existential threat often associated with technological advancements. The research draws on Science and Technology Studies (STS), social constructivism and feminist technology scholarship to challenge the view of algorithms as autonomous systems. Instead, it proposes to follow the actors behind algorithms - designers, bureaucrats, institutions - revealing the human and political work embedded in their construction and deployment. Through a detailed case study approach, employing qualitative methods such as ethnographic methods, interviews, and document analysis, the thesis explores how specific cultural values and institutional frameworks influence the design, implementation, and public perception of ADM systems. The first empirical chapter examines the Dutch welfare system and some recent scandals, such as the childcare benefits case - exposing network determinism as a concept and a practice of institutional discrimination carried out using machine learning, correlation and statistical pattern that resemble eugenics and classification systems. The thesis turns to the Italian context, analyzing a failed teacher allocation algorithm to illustrate how administrative opacity and entrenched bureaucratic habits contribute to institutional blackboxing - a process by which responsibilities are obscured within complex technical systems, even prior to algorithmic opacity, to shield human responsibility. Finally, the research provides a cross-cultural analysis of these dynamics, revealing how public officials in both countries increasingly defer responsibility to automated systems, further eroding accountability and complicating classical notions of legitimacy. The thesis contributes to the field by demonstrating how ADM systems are not just technological artifacts but socio-political projects shaped by values, institutions, and histories. It encourages the social sciences to go beyond the metaphor of the “black box” by re-centering empirical inquiry on constructivist focuses on the actors, contexts and practices that shape each technological application. Ultimately, this work reaffirms the importance of contextual, actor-oriented approaches to understanding automation, particularly in domains where public legitimacy and social justice are at stake.

CONSTRUCTING AUTOMATED SOCIETIES. SOCIO-CULTURAL DETERMINANTS AND IMPACTS OF AUTOMATED DECISION-MAKING IN PUBLIC SERVICES

HUYSKES, DILETTA
2025

Abstract

This dissertation investigates the socio-cultural and institutional factors - or determinants - shaping the design and implementation of automated decision-making (ADM) systems in public services, through a case study in Italy and the Netherlands. It engages with the contemporary discourse surrounding artificial intelligence (AI), real cases of failure of automated discriminatory practices and high-risk uses, highlighting the narratives of inevitability and existential threat often associated with technological advancements. The research draws on Science and Technology Studies (STS), social constructivism and feminist technology scholarship to challenge the view of algorithms as autonomous systems. Instead, it proposes to follow the actors behind algorithms - designers, bureaucrats, institutions - revealing the human and political work embedded in their construction and deployment. Through a detailed case study approach, employing qualitative methods such as ethnographic methods, interviews, and document analysis, the thesis explores how specific cultural values and institutional frameworks influence the design, implementation, and public perception of ADM systems. The first empirical chapter examines the Dutch welfare system and some recent scandals, such as the childcare benefits case - exposing network determinism as a concept and a practice of institutional discrimination carried out using machine learning, correlation and statistical pattern that resemble eugenics and classification systems. The thesis turns to the Italian context, analyzing a failed teacher allocation algorithm to illustrate how administrative opacity and entrenched bureaucratic habits contribute to institutional blackboxing - a process by which responsibilities are obscured within complex technical systems, even prior to algorithmic opacity, to shield human responsibility. Finally, the research provides a cross-cultural analysis of these dynamics, revealing how public officials in both countries increasingly defer responsibility to automated systems, further eroding accountability and complicating classical notions of legitimacy. The thesis contributes to the field by demonstrating how ADM systems are not just technological artifacts but socio-political projects shaped by values, institutions, and histories. It encourages the social sciences to go beyond the metaphor of the “black box” by re-centering empirical inquiry on constructivist focuses on the actors, contexts and practices that shape each technological application. Ultimately, this work reaffirms the importance of contextual, actor-oriented approaches to understanding automation, particularly in domains where public legitimacy and social justice are at stake.
5-mag-2025
Inglese
GANDINI, ALESSANDRO
REBUGHINI, PAOLA ALESSANDRA
Università degli Studi di Milano
206
File in questo prodotto:
File Dimensione Formato  
phd_unimi_R13283.pdf

accesso aperto

Dimensione 4.16 MB
Formato Adobe PDF
4.16 MB Adobe PDF Visualizza/Apri

I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.14242/209369
Il codice NBN di questa tesi è URN:NBN:IT:UNIMI-209369