Leveraging AI for Predictive Analytics With Survey-Based Rubrics in the Public Service Sector in Canada and the USA

In developed countries, a significant gap persists in the absence of empirical, survey-based rubrics to measure AI's technical characteristics and predictive analytics in the public service sector. This study fills this gap by developing and validating survey-based rubrics through a comparison between Canada and the United States (US). Following MacKenzie, Shiau, and Huang's scale development procedures, this research utilized data from Canadian and US government AI websites twice, employing structural equation modeling (SEM) and PLS methods to examine the scale properties and test the relationships between AI's technical characteristics and predictive analytics. Findings show that supervised and unsupervised machine learning, along with deep learning, are positively associated with predictive analytics across public service sectors in both countries. Conversely, artificial neural networks are not positively associated with predictive analytics in Canada, whereas they are in the US. The relationship between artificial neural networks and predictive analytics varies across countries.