AI governance systems: a multi-scale analysis framework, empirical findings, and future directions

The governance of artificial intelligence (AI) has recently emerged as a field of research and practice, but the structural and functional components of AI governance (AIG) systems are not well understood by researchers and practitioners. To address that gap, we apply service system analysis methods and thematic analysis methods to develop a novel framework for conceptualizing and analyzing AI governance systems across multiple scales of activity, including international, national, subnational, sectoral, and organizational systems of governance. We apply our analysis framework to an empirical study of Canada’s national AIG system. Drawing upon qualitative data collected from 20 leaders of Canadian AIG initiatives and subject matter experts, we identify and discuss the actors, impacts, resources, networks, activities, logics, norms, and rules involved in structuring and operating a national AIG system, using Canada as a case study. Based on our findings, we propose three directions for future research: (1) conduct additional analysis of the 610 topics in our dataset, (2) further investigate institutional and ecosystem-level structures and dynamics in Canada’s national AIG system, (3) apply our framework, data, and findings to study AIG systems in other contexts. We also outline four strategic objectives for strengthening Canada’s AIG system: (1) implement new collaboration and coordination mechanisms, (2) create guidance for designing and implementing participatory AIG initiatives, (3) expand access to key resources needed for effective AIG practices, (4) advance diversity, equity, and inclusion in AIG activities.