Artificial intelligence in public administration: dangers, challenges and responsibility

07. 05. 2024.

The application of artificial intelligence in public administration presents numerous challenges, including reliability, transparency, and changes in the labour market. It is a complex and sensitive area, yet it is a fact that its use can create new opportunities. – reads an article in the Public Administration Science professional scientific journal.

The growing importance of artificial intelligence (AI) and its social impact is increasingly felt across various fields. We have previously presented the advantages one can expect when using AI in the financial sector, but how can artificial intelligence be utilized in substantial public administration?

The proliferation of artificial intelligence in public administration

AI-driven computer systems, based on existing data, are capable of drawing conclusions and learning human patterns. AI is present in almost every aspect of life, whether it be online shopping, machine translation, smart homes, digital personal assistants, cybersecurity, production, services, or indeed, public administration. Initially, the use of AI in public administration was unimaginable as it was known for its paper-based operations, but it has now begun to exert an innovative impact in certain areas of the state and local government sectors as well.

When examining the applicability of AI in public administration, it is essential to mention the introduction of automated decision-making. This development in public administration enables decisions to be made without human intervention under certain conditions – such as when the necessary data is automatically available and in cases that do not require deliberation – according to the Administrative Procedures Act* and the Health and Social Services Administration Act**. Although the current use of AI in public administration is still limited, it is already used in certain data management and decision-support tasks, such as facial recognition and chatbots. The Ministry of Innovation and Technology’s initiative, the Artificial Intelligence Coalition, aims to promote the use of AI in other areas of public administration as well, to enhance efficiency and convenience.

The National Tax and Customs Administration of Hungary is clearly a pioneer in the use of AI, not only in Hungary but also among tax authorities worldwide. The Government established the AI Working Group of the NTCA at the end of February 2022 as a consultative, advisory, and opinion-forming body. Within this framework, AI contributes to the analysis of information sets and the identification of tax evaders.

Dangers and challenges: potential side effects of AI application

A global debate is ongoing about the regulation of AI systems, focusing on the protection of fundamental human rights and balancing the potential competitive disadvantages of over-regulation. Despite the benefits of automated surveillance for law enforcement, the development of AI raises significant concerns, particularly regarding the development of autonomous weapon systems, the so-called “killer robots,” for which many international lawyers support pre-emptive restrictions. The opacity of AI systems, the “black box” phenomenon, the damages they may cause, and discriminatory practices, along with the loss of jobs and the widespread dissemination of false information, also cause significant concern. The European Commission’s White Paper proposes a risk-based approach, analysing the main legal and security risks associated with the use of AI. Treating public administration as a particularly risky area is justified since it can significantly affect citizens’ privacy and the public interest alike.

Who is responsible?

The issue of responsibility in AI demands distinguishing important differences among AI algorithms. “Strong” AI, or Artificial General Intelligence (AGI), refers to algorithms capable of independent thinking and self-awareness, and are not confined to predefined tasks. In contrast, pre-programmed algorithms follow predetermined steps and are not capable of creative solutions. In terms of AGI, responsibility issues foresee that damages can be attributed to either the programmer or the user, as they foresee the possible outcomes. New types of liability challenges arise from deep learning machines, which can make decisions based on data learning, as their operation is not entirely transparent. The AI systems used in public administration currently fall below the level of AGI, but developments are ongoing.

Future prospects

Digitalization must include digital governance, which does not merely mean the introduction of individual IT solutions but also the provision of end-to-end digital services. In the area of chatbot applications, we have not yet reached the stage where the liability of AI needs to be considered but given the development of the European Union and its member states, as well as their digitalization policies, the issue of liability might soon arise in public administration. It is conceivable that soon everyone will have their own personal assistant through artificial intelligence, which can help people in various life situations, including administrative procedures. It is crucial that regulatory systems in public administration are prepared to manage these changes effectively, to handle the associated risks and challenges, and to be compatible with all aspects of privacy.

* Act CL of 2016 on General Administrative Procedures

** Act CCXXII of 2015 on the General Rules of Electronic Administration and Trust Services