Article Preview
Top1. Introduction
Artificial intelligence (AI) has a deep potential to change various aspects of citizens’ daily lives and of society as a whole. A systematic review of academic literature has shown growth in the uptake of artificial intelligence in the public sector (Gomes de Sousa et al., 2019; Berryhill et al., 2019; van Noordt & Misuraca, 2020). In Europe alone, the use of AI in public services is increasing, with over 230 empirical use cases identified (van Noordt & Misuraca, 2020). Researchers have noted that AI applications bring significant benefits to institutions that deploy them, from improving public services to reducing the costs and administrative burden (Mehr, 2017; Misuraca et al., 2020). However, these benefits are countered with sobering risks. Concerns for citizens’ privacy and security, loss of decision-making autonomy, and unintentional harm that arise from AI systems may reinforce existing discriminatory practices (Sun & Medaglia, 2019).
As a response to the risks, international organizations and institutions have increasingly advocated for the ethical design and development of AI. The results of their endeavors are realized through the introduction of ethical guidelines, standards, and governance frameworks, or soft law (Bartneck et al., 2021). More recently concrete actions toward operationalizing ethics have emerged in the form of legislative proposals for AI (EU Proposal AI Regulation, 2021). As technical developments in AI flourish, the ethics of AI persists as a contentious yet important discussion for communities, putting into question the human values that are deemed important by society.
Against the background of the multidisciplinary field of AI, empirical research on AI in the public sector has been inadequate (Sun & Medaglia, 2019; Zuiderwijk et al., 2021). Even less has been published about the practical implementation of the ethics of AI in this sector. Only a handful of empirical studies address the state of AI ethics in practice, and they have either focused on companies in the private sector (Vakkuri et al., 2020) or on a broad mix of both (Desouza et al., 2020; Ryan et al., 2021). Researchers note that in practice, most governments have a limited understanding of the implications of the use of AI. They hypothesize that insufficient research on empirical, context-based AI usage in governments can induce systemic failures that may negatively impact not only governments but also societies as a whole (Zuiderwijk et al., 2021). Therefore, this research aims to address this knowledge gap in the rapidly-evolving field of AI by addressing the following questions: