PhD Defence: Tamara Thuis

Abstract

Artificial intelligence (AI) is increasingly woven into work and social life, offering new opportunities while raising complex ethical and organizational challenges. As AI systems become more opaque, unpredictable, and dynamic, organizations face growing difficulty in explaining their behavior, determining responsibility, and governing their development and use. Although guidelines, frameworks, and regulatory requirements for AI ethics have been introduced, how organizations translate these aspirations into practice remains poorly understood. This dissertation therefore shifts attention from what AI ethics should look like to how organizations organize for it, examining how values, technologies, and rules are enacted in everyday AI work. 
Drawing on over three years of qualitative field research within a large organization, this dissertation shows how aspirations for AI explainability, responsibility, and governance unfold in practice. It demonstrates how data scientists construct explanations tailored to different algorithmic contexts; how responsibility for AI shifts as people, algorithms, tools, and organizational processes evolve; and how organizations seek to govern AI not only through technological controls but also through discursive practices that socialize values and align actions around emerging standards. Rather than approaching AI ethics as a compliance-oriented checklist or abstract ideal, the findings reveal AI ethics as situated, relational, and continuously negotiated within organizational settings.
This dissertation highlights that organizing for AI ethics requires sustained, reflexive effort rather than one-off, static solutions. By tracing how ethical aspirations are enacted in practice, it offers insight into how organizations can meaningfully explain, take responsibility for, and govern their AI activities.