Juan David Gutiérrez. Universidad de los Andes, Colombia
Sarah Muñoz-Cadena. Universidad de Rosario, Colombia
We examine how different artificial intelligence (AI) systems adopted by public bodies in Latin America may contribute to the main stages of the public policy cycle: agenda-setting, formulation, implementation, and evaluation. We selected and analyzed various systems that illustrate diverse technologies and government sectors from different countries for that purpose. Three case selection criteria were considered: a) functions of the AI systems, b) countries that adopted them, and c) availability of information about their implementation. Our analysis covers various systems utilized across multiple government sectors and countries, including Argentina, Brazil, Chile, Colombia, Guatemala, Honduras, Mexico, and Peru. We provide basic information about each system, such as its name, the public body that adopted it, its objectives, its operation (including the type of data used), and its intended beneficiaries. We also examine how these systems can help governments achieve their objectives and the risks associated with their implementation. Our research reveals that Latin American governments embrace and use diverse AI systems. However, there is limited publicly available information on these systems, particularly on critical issues such as who developed the tool, whether results or impacts associated with its use have already been reported, and how the system was financed. This underscores the importance of investigating tools that enhance algorithmic transparency. We also highlight the need to prevent and manage risks associated with implementing different AI systems, especially when human rights may be at stake. Additionally, there is a need for further research on adopting AI in Latin America to understand the situations or contexts that may or may not favor the adoption and implementation of systems by the governments in the region