Automated decision-making

This contribution summarises the main amendments adopted by the European Parliament during its first reading of the EU’s Proposal for a Regulation on Artificial Intelligence (AI Act). It outlines the impact of this Act, if adopted with such amendments, on automated administrative decision-making (“adm-ADM”), and examines the margin Member States will have to supplement such provisions in their respective national administrative procedure acts. It concludes that the AI Act is a necessary piece of legislation and that, if adopted with some of the Parliament’s amendments, it will adequately regulate the development and use of AI systems by European public authorities, setting a high regulatory standard that can be reinforced by national legislators.

This paper examines the implementation of artificial intelligence in decision-making processes within public administration, with a focus on addressing the challenges of transparency, accountability, and the intelligibility of AI-generated decisions. The paper discusses the importance of imputability in decisions made with deep learning algorithms. It emphasises that by granting public administrations full control over the training dataset, source code, and knowledge base, the imputability of the decision can be ensured. This control enables administrations to validate the relevancy and accuracy of the algorithm's training data, address potential biases, and comply with legal and ethical requirements. The paper then proposes the use of Large Language Models (LLM) as a solution to enhance the transparency and motivation behind AI-assisted decisions. It highlights that LLMs can generate articulate and comprehensible textual outputs that closely resemble human-generated decisions, allowing for a deeper understanding of the decision-making process. Furthermore, the paper emphasises the significance of providing access to the training dataset, source code, and individual administrative precedents to enhance transparency and accountability. It argues that by offering these components, stakeholders can evaluate the validity and reliability of AI-assisted decisions, fostering trust in the decision-making process.

Spain's legislation on automated administrative action is among the most advanced in the world. Enacted in 2007, it has remained the same in the current Administrative Procedure Act of 2015. Currently, more than one third of the Spanish public sector uses these systems. Some problems of compliance with general administrative law remain, such as the right to correct applications, the right to present arguments in the face of a draft before a rejection decision is issued, or the motivation of automated decisions or the transparency of programming. But despite the fact that thousands of automated decisions are taken every day, judicial litigation in this area has so far been merely anecdotal. This paper describes the requirements established by Spanish law for the implementation of automated decision-making systems used by Public Administration: their express prior approval, the auditing of the system, the electronic signature of the decisions, etc.; and the critical aspects that still remain.

Digitalization of the Swedish administration has been the center of attention for the Government and legislator for decades, with the explicit goal of being the «best in the world» at using the possibilities of digitalization. The Swedish public authorities were early adapters and have utilized automated decision-making procedures since the 1970’s. The effects on two areas central to Swedish constitutional and administrative law are addressed here: general administrative law, with a focus on administrative decision-making, and transparency rules and access to documents. The regulatory responses in the two areas differ. While concerns regarding the negative effects of digitalization on access to documents prompted the Swedish legislator to include digital recordings among documents encompassed by the right to access already in the 1970s, general administrative law has been adapted to a very limited extent. The Government has over the years commissioned several governmental enquiries on technical, societal, and legal challenges of digitalization of the administration, but the main conclusions in regard to the suitability of automated decision-making have been left to the administration and the courts. Two reasons can be identified. First, it has been a longstanding aim that administrative procedural rules are to be technology-neutral, in order not to become obsolete through future developments. Second, administrative authorities hold a strong and partially independent position in the Swedish constitutional setting, combined with a tradition of pragmatism. Authorities can thus be trusted to perform their decision-making in accordance with the law, whether the procedures are carried out manually, digitally, or by automated means. Swedish courts have thus far refrained from intervening. Still, it is clear that the Swedish administrative law landscape in automated decision-making contains several central lacunas, which ought to be filled in order to safeguard the principle of good administration and the rule of law.

In spite of the image of a developed e-governance, advanced automated decision-making (ADM) systems have not been widely used in Estonian public administration and there is still no general legal framework for them. The draft bill to amend the Administrative Procedure Act, which was presented to Parliament in 2022, takes a rather cautious approach to the issue too, significantly limiting the automation of discretionary decisions and in particular the use of self-learning algorithms. Automated administrative decisions would not be discouraged by the application of procedural principles inherent to the rule of law, such as hearing and reasoning. However, for the automation of discretionary decisions in appropriate cases, a solution has been proposed whereby typical cases would be solved in a fully automated way by means of predefined algorithms based on internal administrative guidelines. This solution is not an universal magic bullet for every situation, but may allow for a certain degree of innovation, provided appropriate procedural and organisational safeguards are respected. Fundamental preconditions for that are the categorical separation of the guidance and algorithm, as well as the publication of the guide. An optimal model of public accountability has to encourage authorities to take appropriate precautions when implementing algorithms.

This contribution analyses the German legal framework concerning automated decision-making systems. Public administration in Germany uses automated decision-making systems primarily to adopt partially or fully automated administrative acts. Accordingly, this paper focuses on automated administrative acts, however, it also discusses the first comprehensive regulation of artificial intelligence enacted by a German “Bundesland”. The paper approaches the topic from three different perspectives: The German legislative framework, case-law and soft-law concerning automated decision-making systems.

Automated decision-making has been discussed in Austrian administrative law for more than 40 years. The focus has always been on the administrative act (in the sense of a formal individual decision) and the pertaining procedure. In this area, there are established principles, although new technologies raise new questions. Beyond the administrative act, we are still very much in the dark.

This article aims at analysing the decision-automation-systems currently used by public administrations in Italy. After an analysis of the legal framework, the different systems are classified and illustrated: in particular, the case of the so-called “good school” algorithm is discussed. The conclusions dwell on the reason for the scarce use of these tools in the Italian landscape, also due to the slow and uneven digitisation of the public sector.

Jurisdictions within the EU and countries around the world are beginning to regulate the use of public Automated Decision Making (ADM). The legal framework thereof differs considerably, and its development is at an early stage. This contribution sets out a possible comparative research framework, with other words elements to compare the different solutions developed by the legal systems in the face of challenges of ADM.