Exploring the Complexity of AI Applications in the Public Sector: The Interplay of Visibility, Autonomy, and Self-Learning

Artificial intelligence (AI) has been deployed in many government contexts and with very different results in countries around the world. There seems to be a distinct transformational power when compared with previous technologies. However, it is not clear how different characteristics of AI systems affect their purpose and outputs. Therefore, by understanding some of the unique characteristics of self-learning systems in the context of a proposed typology consisting of two dimensions-visibility and autonomy-this study explores the interplay of visibility, autonomy, and self-learning in government AI systems. Based on the analysis of four distinct AI cases across diverse U.S. federal agencies, this ongoing research paper aims to uncover some of the opportunities and challenges posed by AI and specifically self-learning as one of its main features. Our preliminary results underscore the necessity of contextual analysis in deploying AI systems, thereby contributing to previous research on different characteristics and types of AI.