Ms Schwaar, where do AI-based systems already deliver added value in interactions between the state and citizens today – and where would you consciously choose to do without them?
Stefanie Schwaar: When considering the use of AI-based systems in interactions between the state and citizens, it is important to distinguish between different levels. Currently, such systems are mainly used in mobile applications. Many apps, for example, enable or improve real-time communication between people with different language skills. In this area, standardised solutions could be useful in public authorities to facilitate initial, low-threshold communication. It is also conceivable to support citizens in filling out applications with AI-supported help functions. Initial approaches already exist, such as simple chatbots like the advice bots Bobbi, Mia and Theo.
In my experience, however, these systems – similar to traditional FAQ sections – are primarily capable of handling standardised enquiries. When it comes to individual or more complex issues, direct communication between people remains indispensable. Against this backdrop, processes that follow a clear and repeatable pattern or involve large amounts of data offer considerable potential for increasing efficiency through AI-based systems. At the same time, however, it must be ensured that no automated decisions are made that significantly affect people's lives or living conditions. Special protective measures are required, particularly when dealing with sensitive health data. The well-known problem of ‘hallucinations’ in AI systems makes it clear that automated evaluation of sensitive data is only permissible with strictly tested, specialised algorithms.
Furthermore, it should be noted that there are certainly processes in government procedures that follow easily verifiable rules. Simple algorithmic solutions should also be used for such processes in order to increase efficiency, reduce errors and shorten processing times.
What minimum technical requirements for transparency and objection should apply to AI-supported administrative processes?
Stefanie Schwaar: Public administration already has established mechanisms in place to ensure transparency and legal protection, including clear rights of appeal and transparent decision-making processes, as applied, for example, to digitally generated notices. These basic principles form the foundation of trustworthy administrative action and should continue to apply unchanged when AI-supported procedures are introduced. Citizens must continue to receive all information relevant to decisions and the underlying legal norms in an understandable form.
It is also important to clearly indicate when an administrative act has been created as part of an automated process. Labelling this information not only ensures transparency, but also strengthens trust in new digital processes. Knowing that an algorithm was used and that no direct human intervention took place allows those affected to properly assess the outcome and take action if necessary.
A comparison from the healthcare sector illustrates the importance of such transparency: citizens also receive information on the areas of application and possible risks of medicines so that they can take timely action if necessary. Similarly, knowledge of automated decision-making processes supports the proper evaluation of a decision. If it is apparent that a letter has been generated by a machine, citizens can check it particularly carefully and initiate corrections if necessary. Conversely, knowledge of human processing often enables direct clarification in personal exchanges.
In this way, the use of AI in public administration can both increase efficiency and maintain public trust – provided that transparency and comprehensible information remain central components of administrative action. Modern technologies offer great opportunities, but they must always be in harmony with the rights, expectations and security needs of citizens.
How can we ensure that AI makes administration more inclusive?
Stefanie Schwaar: Inclusivity begins with the basic accessibility of administrative services. This includes accessible websites, understandable digital user guidance, and clear orientation and signage in buildings. Many of these areas still present barriers today that make equal participation difficult.
AI-based systems can provide targeted support here by providing citizens with needs-based digital tools – for example, translation and read-aloud functions, simplified language or assistive apps for people with physical or cognitive impairments. In this way, AI helps to break down individual barriers and enables people to use administrative services independently.
At the same time, inclusion must be anchored in the planning of digital offerings and spaces. AI can be used as a testing and analysis tool, for example through digital twins that simulate how inclusive an offering or environment is. In this way, weaknesses can be identified at an early stage and improved in a targeted manner.
AI can thus make an important contribution to making administration more accessible, understandable and fair – provided that inclusion remains a binding guiding principle.
The whitepaper „The State in Digital Dialogue – AI in the Interaction between Government, Citizens, and Businesses“ (in German) is available for download here.
The interview is released for editorial use (provided the source is cited © Plattform Lernende Systeme).