Recently, several initiatives have made attempts to regulate the use of artificial intelligence (AI) research results in applications. Respective approaches are either legislative (European AI Act), of a governing nature (United Nations AI Advisory Board Initiative), or formulated as guidelines (e.g., for the use of AI technologies in schools and universities). For the two application domains covered in this special issue, namely healthcare and public sector projects, regulations definitely have immediate impact.

The European AI ActFootnote 1 is the world’s first comprehensive legal framework for AI systems, and it aims to promote the trustworthy use of AI technology in Europe and sets out clear requirements for AI developers and those responsible for the deployment of AI systems. Citizens have the right to submit complaints about AI systems and receive explanations, especially for AI systems used in so-called high-risk settings. High-risk systems according to the classification of the European AI Act are AI systems that pose a significant risk to health, safety, or fundamental rights of humans. There are greater transparency and public access requirements to high-risk systems. These systems are divided into different categories and are subject to strict compliance requirements in accordance with the provisions of the AI Act. Providers of such high-risk systems must carry out conformity assessments and adhere to specific requirements to ensure compliance with the regulations. Whoever classifies a system as being of high risks for humans must consider the social mechanism in which humans interact with systems and also with one another. What might be of high risk in one mechanism might be tolerable in another.Footnote 2 In the statement on “Governing AI for Humanity”Footnote 3 from the UN AI Advisory Body, for example, direct reference is made to the OECD definition of AI with the classic agent-based view on AI systems.Footnote 4 It is suggested that the design of the implementation of an agent architecture should be regulated in an appropriate manner.

Thus, any application of AI technology must intimately consider the notions of agents and mechanisms in the future, and therefore, the special issue will shed light on the impact of all dedicated research articles on a respective general AI perspective, such that new results of AI research can ultimately be brought to bear in applications under the control of the AI Act and, e.g., UN guidelines. It will become apparent that regulation actually turns out to be very fruitful for applications on the one hand, and that regulation also stimulates new research on the other. With a general AI perspective, different research contributions stemming from application areas as diverse as healthcare and private sector can indeed be seamlessly integrated into a coherent whole, namely into the AI weltanschauung behind agents and mechanisms.

1 Content

This special issue covers a variety of topics within the area of AI in healthcare and the public sector, with four technical contributions, five project reports, a system description, and two dissertation abstracts, next to a survey that also presents short summaries and interpretations of the above mentioned contributions and an interview with Katharina Morik on lessons learned from resource-aware machine learning for healthcare applications. Together, the articles highlight the ongoing activity within the field.

1.1 Survey

  • AI in Healthcare and the Public Sector: How to Face the Challenges of High-risk Applications and What AI Research Can Get Out of It [1] Tanya Braun, Ralf Möller.

1.2 Technical Contributions

  • Lifting in Support of Privacy-preserving Probabilistic Inference [6] Marcel Gehrke, Johannes Liebenow, Esfandiar Mohammadi, Tanya Braun.

  • Partial Image Active Annotation (PIAA): Efficient Active Learning using Edge Information in Limited Data Scenarios [7] Md Abdul Kadir, Hasan Md Tusfiqur Alam, Devansh Srivastav, Hans-Jürgen Profitlich, Daniel Sonntag.

  • Automated Computation of Therapies Using Failure Mode and Effects Analysis in the Medical Domain [9] Malte Luttermann, Edgar Baake, Juljan Bouchagiar, Benjamin Gebel, Philipp Grüning, Dilini Manikwadura, Franziska Schollemann, Elisa Teifke, Philipp Rostalski, Ralf Möller.

  • AutoRAG: Grounding Text and Symbols [11] Tim Schulz, Malte Luttermann, Ralf Möller.

1.3 Project Reports

  • Auditive emotion recognition for emphatic AI Assistants [3] Roswitha Duwenbeck and Elsa Andrea Kirchner.

  • Multimodality in Explanations: Lessons Learned from Image Classification for Medical and Clinical Decision Making [4] Bettina Finzel.

  • Building an AI Support Tool for Realtime Ulcerative Colitis Diagnosis [10] Bjørn Leth Møller, Bobby Zhao Sheng Lo, Johan Burisch, Flemming Bendtsen, Ida Vind, Bulat Ibragimov, Christian Igel.

  • Requirements for a social robot as an information provider in the public sector [13] Thomas Sievers, Nele Russwinkel.

  • EpiPredict: Agent-Based Modeling of Infectious Diseases [14] Janik Suer, Johannes Ponge, Bernd Hellingrath.

1.4 System Descriptions

  • A Toolchain for Privacy-Preserving Distributed Aggregation on Edge-Devices [8] Johannes Liebenow, Timothy Imort, Yannick Fuchs, Marcel Heisel, Nadja Käding, Jan Rupp, Esfandiar Mohammadi.

1.5 Dissertation Abstracts

  • Taming Exact Inference in Temporal Probabilistic Relational Models [5] Marcel Gehrke.

  • Analysing Semantically Enriched Trajectories [12] Jana Seep.

1.6 Interview

  • Lessons from Resource-aware Machine Learning for Healthcare – An Interview with Katharina Morik [2] Tanya Braun, Ralf Möller.

2 Service

2.1 Conferences

AI in healthcare and the public sector is an interdisciplinary topic that is regularly covered by articles submitted to all of the major AI conferences such as AAAI, ECAI, IJCAI, or NeurIPS, to name a few, as well as major machine learning conferences such as ICML or ECML. Additionally, there are also a host of application-specific conferences, focusing on different aspects such as image processing, robotics, human-computer interaction, or ethical aspects.

2.2 Workshops

There is a wide variety of workshops that regularly co-locate with major AI conferences highlighting different important aspects of AI in healthcare and the public sector such as “AI for Critical Infrastructure”, “Trustworthy AI for Healthcare”, and “AI Governance: Alignment, Morality and Law” at IJCAI-24 or “AI for Credible Elections: A Call to Action with Trusted AI”, “Health Intelligence”, “Machine Learning for Cognitive and Mental Health”, and “Public Sector Large Language Models: Algorithmic and Sociotechnical Design” at AAAI-24 to name a few. (The list is necessarily not complete. We apologise to all colleagues whose workshops have not made it onto this list.)

2.3 Journals

Several journals have presented special issues on the topics, including special application-oriented tracks at AI journals such as “AI and COVID-19” at JAIR (Vol. 76, 2023) and AI-oriented tracks at non-AI journals (e.g., special issue “AI and Public Policy: Many Dimensions and Key Challenges” at the “Global Public Policy and Governance” journal).