The exercise of discretionary competence by artificial intelligence: A legal analysis considering the democratic principle

Abstract

The article seeks to demonstrate that delegating the exercise of discretionary authority to an artificial intelligence (AI) system, in its current stage of development, is prohibited due to a violation of the democratic principle. The idea starts with the fact that AI-based systems, by being autonomous like human intelligence, can present unprecedented solutions due to the introduction of new decision-making criteria throughout their execution. Thus, the final structure of the program’s algorithms cannot be known in advance, which is not the case with simple automation, where the entire algorithmic chain is written into the program from the outset. The uncontested core of discretion lies in the choice of one solution over another that is also permitted by law. This choice, although delimited by law, inevitably involves the ideological framework of the agent. If the agent was elected to serve as a public administrator, it means that the people who elected them have a public subjective right to see them—along with their subordinates—implement the legal order, and also to see their ideas being realized in discretionary decisions, rather than those of their defeated opponent or a machine whose criteria for decision-making cannot be known.