Use of Domain Engineering in Hyperautomation Applied to Decision Making in Government

Abstract

This article presents the domain engineering process carried out to obtain the requirements for the implementation of an Artificial Intelligence (AI) compliance framework aimed at the public sector. Owing to the current competitive and fast economy, which generates huge demand for increasingly efficient, reliable, and transparent intelligent systems, decision-support architectures should also be developed under strong restrictions of cost and time. Such a context requires adequate structures, processes, and technologies for coping with the complexity of building such intelligent systems. Currently, many public organizations have adopted applications for process automation, with the aim of refraining from repetitive work and producing more efficient results. However, what is not so often observed is the development of intelligent engines to support complex public decision-making. Possible explanations are the plethora of available data sources and the number of legal norms to be abided by. Moreover, it is important to highlight the need to incorporate transparency, auditability, reusability, and flexibility into such systems. Thus, they can be safely utilized in various analogous situations, reducing the need to develop new applications from scratch. An architecture suitable for supporting public decision-making with so many features and increasingly unstructured data, as well as abundant regulation, needs well-crafted formal specifications. This article aims to analyze three existing frameworks and carry out domain engineering studies in three cases to produce some guidance for future public applications and services based on AI. Next, we provide a conceptual preliminary architectural definition for the public sector. The proposed architecture targets were identified in the three cases studied, namely, frequent tasks of process mining requirements, detection of anomalies, and extraction of rules and public policies for helping public servants. All these aim at expedient AI development for public decision-making.