Artificial intelligence governance: Understanding how public organizations implement it

Abstract

While observing the race for Artificial Intelligence (AI) regulation and global governance, public organizations are faced with the need to structure themselves so that their AI systems consider ethical principles. This research aimed to investigate how public organizations have incorporated the guidelines presented by academia, legislation, and international standards into their governance, management, and AI system development processes, focusing on ethical principles. Propositions were elaborated on the processes and practices recommended by literature specialized in AI governance. This entailed a comprehensive search that reached out to 28 public organizations across five continents that have AI systems in operation. Through an exploratory and descriptive aim, based on a qualitative and quantitative approach, the empirical analysis was carried out by means of proposition analysis using the Qualitative Comparative Analysis (QCA) method in crisp-set and fuzzy modes, based on questionnaire responses, combined with an interview and document content analysis. The analyses identified how processes and practices, across multiple layers and directed at the application of ethical principles in AI system production, have been combined and internalized in those public institutions. Organizations that trained decision-makers, AI system developers, and users showed a more advanced stage of AI governance; on the other hand, low scores were found on actions towards AI governance when those professionals did not receive any training. The results also revealed how governments can boost AI governance in public organizations by designing AI strategy, AI policy, AI ethical principles and publishing standards for that purpose to government agencies. The results also ground the design of the AIGov4Gov framework for public organizations to implement their own AI governance.