Current Projects

Short description:

The DFG project “Architectures for Vulnerable Road Users” (ACHIEVE) #516946933 is a collaboration between Prof. Klaus David from the Department of Communication Engineering (ComTec) at the University of Kassel and Prof. Delphine Reinhardt from the Department of Computer Security and Privacy (CSP) at the University of Göttingen. The main goal of ACHIEVE is to investigate various architectures for cooperative systems with a focus on collision avoidance and the privacy of pedestrians and cyclists. The performance of these architectures will be measured using quantitative evaluations.

There are various types of system architectures being investigated as part of the project. These include peer-to-peer architectures, client-server architectures, and hybrid architectures. They differ, among other things, in the type of communication (direct, cellular, or hybrid), the location where collision avoidance algorithms are computed (on mobile devices, on-board units, servers, or distributed), and the type of warnings issued.

The first step is to identify the various possible system architectures. Next, custom-developed key performance parameters for collision detection and privacy compliance are defined. These parameters are used to evaluate and rank the architectures accordingly.

The evaluation is therefore based on quantitative measurements designed to accurately assess the performance of each architecture in terms of collision avoidance and privacy. This makes it possible to better understand the advantages and disadvantages of the various architectures and to make informed decisions regarding the further development of cooperative systems in the field of road traffic.

Short description:

The University of Kassel is involved in the joint project "AI for REAL-world NETwork operation (AI4REALNET)", together with 15 other partners from academia and industry. The project is funded under the Horizon Europe Framework Program (HORIZON) and will receive a total of almost 4 million euros for a period of 3.5 years from October 2023.

The scope of AI4REALNET includes the perspective of AI-based solutions for critical systems (power supply, railroad and air traffic management) that are modeled by networks, can be simulated and are traditionally operated by humans, with AI systems complementing and augmenting human capabilities.

The goal is to develop the next generation of decision-making methods driven by supervised and augmented learning, aiming at trustworthiness in AI-assisted human control with augmented cognitive performance, hybrid human co-learning capability and autonomous AI, with resilience and safety of critical infrastructures as core requirements. In addition, AI4REALNET aims to foster the development and validation of novel AI algorithms by the consortium and the AI community, through existing open-source digital environments capable of emulating realistic scenarios of physical systems operation and human decision making.

The core components include AI algorithms consisting mainly of supervised and reinforced learning, combining the benefits of existing heuristics, physical modeling of complex systems and learning methods. These are accompanied by a set of complementary methods that aim to improve transparency, safety, explainability and human acceptance. In addition, there is a focus on decision making with human involvement to facilitate joint learning between AI systems and humans. This includes the integration of considerations such as model uncertainty, human cognitive load and trust. Finally, autonomous AI systems based on human supervision and enriched with human expertise and safety rules are integrated.

The AI4REALNET framework is validated in 6 use cases driven by industry-driven requirements and focusing on 3 network infrastructures with common characteristics. The use cases focus on critical challenges and tasks of network operators, taking into account strategic long-term goals such as decarbonization, digitalization and resilience to disruptions, and are formulated as a unified sequential decision problem where many AI and non-AI algorithms can be applied and benchmarked.

Short description:

Artificial intelligence (AI) is being used in more and more areas. However, only a few still understand which data is used and how. An international research project with the participation of the University of Kassel wants to create transparency.

The project "AI Forensics: Accountability through Interpretability in Visual AI Systems" focuses on the use of image systems. "The field of facial recognition in particular continues to grow. For example, in the U.S., AI with facial recognition is being used to detect potential threats. What serves general security is at the same time an intrusion into personal rights. We want to make AI decisions comprehensible," explains project participant Prof. Dr. Claude Draude from the Scientific Center for Information Technology Design at the University of Kassel.

The goal is an Internet platform that is freely accessible to the public and makes large data sets accessible. Machine learning models should thus be made analyzable and understandable. "Users can upload an image, for example, and activate an automatic examination. The system checks whether the image is already in a dataset (dataset forensics), how it was used in a model (model forensics), and where and for what purpose (application forensics). The scope of these capabilities varies from model to model, illustrating different degrees of transparency and interpretability across the spectrum of models available for research," explains Professor Draude.

Project participants (research): Institute of Artificial Intelligence and Media Philosophy at the Staatliche Hochschule für Gestaltung in Karlsruhe Department of Computer Science at Durham University in the UK Department of Germanic and Slavic Studies at the University of California in Santa Barbara (USA) Technical Partners Steinbuch Centre for Computing at the Karlsruhe Institute of Technology (KIT) NVIDIA CUDA Research Centre at Durham University.

Short description:

Users of digital services often face the challenge of being unable to realistically assess exactly how and for what purposes their personal data is used or shared by providers. Users generally have to acquire the skills for informational self-determination—that is, the responsible selection and control of information—on their own. This often involves asking friends and acquaintances, conducting their own online research on networks, in databases, and through recommender systems, or seeking guidance from professional advisory institutions and government agencies. Users are often confronted with a flood of offers and information, making it difficult for them to filter out reliable and helpful information. This calls for innovative advisory services that enable users to gain the necessary support and skills for informational self-determination so they can navigate the digital world safely.

The project “Advising Users: Strengthening Informational Self-Determination Through Collaborative Partnerships in Digital Consumer Protection (BeDeNUTZ)” examines how forms of digital consumer protection counseling must be designed to improve users’ informational self-determination in the digital world. Specifically, this question will be researched by examining the need for counseling on the topics of digitized electricity consumers and digital electricity consumer services. The project’s goal is to identify and analyze consumers’ support needs and, based on this, to design appropriate training programs and network structures. To this end, various methods will be empirically examined and thoroughly analyzed. To enable the implementation of these concepts in consumer counseling, a catalog of measures will be developed that consumer organizations can apply directly in practice. The new concepts are intended to address the needs of both traditional and digital consumer protection.

In this way, the BeDeNUTZ project makes an important contribution to strengthening citizens’ IT security. Through empirical research into various advisory scenarios, the insights gained can be applied not only to issues related to digital electricity meters but also to other topics in which informational self-determination plays a central role. With the help of these collaborative advisory formats, citizens can learn to use digital media more competently while safeguarding their informational self-determination. This enables them to make more informed decisions in complex information environments, which significantly contributes to upholding European values and fundamental rights in the context of digital technologies.

Short Description:

The BioDroneAI project uses low-noise LTA-UAV drones and AI-powered technology to make the monitoring of animal populations more comprehensive, efficient, and less disruptive to wildlife. The use of multimodal sensors—including thermal imaging and high-resolution cameras, as well as acoustic detection—enables precise and comprehensive species surveys in large and hard-to-access areas. BioDroneAI aims to protect biodiversity in sensitive ecosystems such as wetlands and strengthen their role as CO₂ sinks. Automated data collection and analysis are intended to alleviate the shortage of skilled workers in biodiversity monitoring and accelerate planning processes for sustainable energy infrastructure projects. In the long term, BioDroneAI supports species conservation and the achievement of the UN Sustainable Development Goals through an innovative combination of environmental technology, robotics, and artificial intelligence.

Coordination: Prof. Dr. Gerhard Heyer, Institute for Applied Computer Science e.V.
Partners: University of Kassel, Anhalt University of Applied Sciences, and the German National Library

Motivation: The use of language models is often subject to restrictive requirements, known as constraints. For example, if data subject to licensing restrictions is used for training, it should generally not be reproduced in artificially generated texts. Furthermore, statements in generated texts should be transparently traceable to their sources. For many institutions and companies, these and similar requirements are essential for the productive and secure use of language models. The question, then, is whether and how constraints can be taken into account in the design of language models. This is precisely where the “CORAL” research project comes in, aiming to make artificial intelligence more flexible, resilient, and efficient.

Objectives and Approach: The project investigates whether practically usable language models can be trained using texts that may only be made available in various restricted forms. In addition, methods are being developed to generate texts that incorporate domain expertise and include source citations. In particular, the goal is to avoid simply reproducing text from the training data while accurately reflecting specified sources. These methods will be evaluated in extensive experiments and tested with partners from the financial sector, GLAM institutions (cultural and memory institutions), and the private sector.

Innovations and Prospects: Innovative results and insights are expected in three key areas related to the development and use of language models in society, academia, and industry: (1) Incorporation of previously limited training data; (2) model architectures that account for constraints—including those that prevent the reproduction of training data; and (3) references to relevant and reliable sources on which the generated text is based. The exemplary transfer of these approaches will reliably demonstrate both flexibility and effectiveness.

Brief description:

The “HUB: Shaping Sustainable Data-Driven Value Creation” (DaWeNa-HUB) project aims to establish a comprehensive innovation ecosystem that guides SMEs and the public sector through the challenges of digital transformation. As a central hub, it connects various research projects to develop advanced digital solutions and business models based on sustainable data utilization. By leveraging state-of-the-art technologies such as generative AI, large language models, and low-code/no-code platforms, the project promotes the widespread adoption of innovative methods and tools that increase efficiency and support organizational resilience.

Short Description: 
With the increasing availability of data in recent years, a repertoire has emerged that not only offers benefits but also holds significant potential for abuse. A highly topical and sensitive issue in this context is the creation of so-called “deepfakes,” which were first mentioned in 2017 on the “Reddit” platform. The prevalence of this image, audio, or video content—generated or manipulated using artificial intelligence—is demonstrably on the rise and poses a real threat to citizens’ personal rights and trust in digital media.

Beyond their use for harmless entertainment and artistic purposes—and for generally beneficial purposes—various areas of law are affected, so particular attention is being paid to the implications of deepfake technologies in the fields of tort law, copyright law, platform law, data protection law, constitutional law, criminal law, and criminal procedure law.

The overarching goal of the collaborative project “Protecting Citizens’ Privacy and Other Rights from Abuse by Deepfakes (DEEP-PRISMA)” is therefore to improve the conditions for exercising the right to informational self-determination and related rights in the context of deepfakes. To this end, it is first necessary to examine the effects of abusive deepfakes and to systematically assess the effectiveness of both existing rights and regulations designed to protect against such abuse through multi-layered empirical studies. Building on this, recommendations for action regarding the exercise and enforcement of rights will be derived through interdisciplinary dialogue, and proposals for reform will be developed.

The project thus contributes to strengthening fundamental rights and safeguarding privacy and personality rights in connection with artificial intelligence technologies. In addition, the project promotes the sovereign shaping of technology according to democratic standards while also fostering a broader understanding of how individuals handle their own data.

Short Description:

This priority program explores fundamental theoretical questions that lie at the intersection of several disciplines. To date, for example, research on fake news and conspiracy theories has been conducted in philosophy, psychology, communication and media studies, and media law, without researchers being aware of the current state of research and discussion in the respective neighboring disciplines. Re:DIS aims to create a shared research platform for disinformation research across various disciplines.

Among Re:DIS’s goals is to develop a common conceptual and theoretical framework and, in particular, to better understand borderline cases of disinformation in which the recipients’ belief formation is subtly manipulated. Among other things, this is essential for the legal regulation of disinformation: Society and lawmakers must make well-founded decisions regarding which types of disinformation are acceptable in a liberal democracy and which fall outside the scope of protection afforded by freedom of speech, freedom of expression, and freedom of the press.

Subproject by Prof. Dr. Gerrit Hornung (ITeG, University of Kassel) in collaboration with Prof. Dr. Nicole Krämer (Psychology, University of Duisburg): “Resilience to Disinformation – An Investigation of Legal and Psychological Constructs of Disinformation on Online Platforms”

Short Description:

Assessments of the crisis and challenges facing digital journalism are often linked to the role of large platform companies. While social media platforms open up new opportunities for participation on the one hand, on the other hand they make it easier for unverified information to be published and for important curation tasks to be delegated to algorithmic recommendation systems. Talk of a “post-truth” era, as well as aspects of the coarsening and polarization of public discourse, are symptomatic of these developments. As a result, established publishers find themselves forced to respond to these developments by, for example, adapting their business models. Furthermore, newer players in digital journalism are experimenting with platform models to uncover innovative ways to disseminate journalistic content. However, there is often a lack of concepts for platforms and their technical implementation that take into account the multifaceted aspects of establishing structures that promote democracy while simultaneously ensuring economic profitability.

Against this backdrop, the Democracy-X project—funded by the BMFTR—is developing relevant design parameters for a nonprofit news platform intended to ensure economic viability within a model oriented toward the common good. This lays the groundwork for initiative groups and operator consortia seeking to establish similar platform models. The project’s findings are intended to contribute to a deeper understanding of how privacy-preserving, fair, and public-interest-oriented news platforms can be designed.

Short description:

The COVID-19 pandemic and the resulting shift to distance learning in schools brings the question of how to use school information systems (e.g. learning apps, infrastructure apps, learning platforms) into focus. The wide range of offerings and the requirements that the General Data Protection Regulation (GDPR) in particular places on the operation of these information systems lead to a variety of legal uncertainties in everyday school life. Teachers as well as parents and students are often insufficiently informed when evaluating whether an application they use meets the requirements of data protection law. Developers are also uncertain about the legal requirements that a product they have developed must meet. Overall, there has been a lack of clear identifiability of applications that comply with the complex data protection requirements.

To address this problem, the BMBF joint project "Data Protection Certification for Educational Information Systems" ("DIRECTIONS") is developing a national data protection certification for school information systems. Two stages of development are planned: First, a seal of approval will be designed that can be awarded for compliance with certain data protection criteria. In a second step, this seal of approval will be further developed into a fully-fledged data protection certification in accordance with Art. 42 of the GDPR.

The University of Kassel, Department of Public Law, IT Law and Environmental Law (Prof. Dr. Gerrit Hornung, LL.M.), is working on the DIRECTIONS sub-project "Legal Requirements for Certification Criteria and Assessment Programs". Within the framework of the sub-project, the University of Kassel is involved in the more detailed definition of the object of certification as well as the development of the criteria catalog. Among other things, the regulations of the DSGVO, the TTDSG, the BDSG, the state data protection laws and the state school laws as well as the planned e-Privacy Regulation will be taken into account. Based on this, a seal of approval will first be designed, tested and awarded, which will then be further developed into a certification in the sense of Art. 42 DSGVO. In this context, the University of Kassel is involved in the development of a conformity assessment program as well as a practicable modularization and protection class concept that takes into account the different risks of various processing operations as well as the needs of small and medium-sized enterprises (SMEs).

DIRECTIONS is coordinated by the Karlsruhe Institute of Technology (Prof. Dr. Ali Sunyaev). Other partners are the University of Kassel (Prof. Dr. Gerrit Hornung, LL.M.) and datenschutz cert GmbH. In addition, the project is accompanied by a heterogeneous expert advisory board, various associated partners, and a panel of data protection supervisory authorities.

For more information, see the DIRECTIONS website.

Short description:

Many users of digital services often agree to the collection and processing of their data without being aware of the consequences this may have. This problem is particularly acute for those groups of people who, for structural or individual reasons, have less background knowledge about the advantages or even necessity of digital privacy protection. They also often lack the feeling of being able to make self-determined decisions in technical areas. As it makes little sense or is feasible to strengthen the privacy literacy of these groups using traditional means, the project aims to research and test alternative approaches to guaranteeing the fundamental right to informational self-determination.

The basis for the development of new protection concepts is an ethically sound systematization of various vulnerable groups. The aim is to increase awareness of the disclosure of private data in order to avert potentially negative consequences. The results of the research will be used to develop appropriate attention-grabbing signals (e.g. visual overlays) and other suitable mechanisms to prevent people from agreeing to the collection of data too quickly. In particular, it is being investigated whether the perception of intrusions into privacy can and must be made instinctively tangible. The aim is to strengthen intuitive behaviors for the protection of privacy.

Based on ethical and psychological approaches, the project will develop mechanisms that lead to greater data protection competence among users and strengthen the self-protection of their own privacy in consent processes. Building on this, the construct of informed consent will be reconceptualized in relation to vulnerable groups. Concrete recommendations for the development and design of appropriate user interfaces will be derived from the results.

The project "Diversity-oriented privacy protection in digital environments (Divers Privat)" is the subject of the funding guideline "Platform Privacy - Supporting citizens in exercising the fundamental right to informational self-determination" as part of the Federal Government's research framework program on IT security "Digital. Secure. Sovereign". The BMFTR is funding the project for three years with 1.23 million euros.

Further information can be found on the BMFTR website.

Project partners:

- University of Duisburg-Essen, Social Psychology: Media and Communication: Prof. Dr. Nicole Krämer (Coordinator)
- University of Passau, Chair of Applied Ethics: Prof. Dr. Karoline Reinhardt
- University of Tübingen, International Center for Ethics in the Sciences and Humanities (IZEW): PD Dr. Jessica Heesen

Short description:

The GenKITS project aims to integrate generative AI into the business processes of Starke + Reichert GmbH & Co. KG (SR) in order to optimize internal processes and open up new opportunities for SMEs. This technology could complement or replace existing products, which is why a GPT orchestrator is being developed to enable SME employees to create their own GPT agents even without programming knowledge. At the same time, comprehensive training programs are being conducted to actively involve the workforce in the development and use of AI technologies. Following a successful pilot phase, the solutions will also be made available to SR’s partner companies, with the specific goal of helping SMEs get started with AI and increasing their competitiveness.

Short Description:

Current Situation
In 2024, a total of nearly €49 billion was spent on integration assistance and social assistance in Germany. Of this amount, €28.7 billion went toward integration assistance benefits for over one million people. Nevertheless, matching those seeking assistance with suitable services is often a lengthy and inefficient process. Social workers primarily search for available services manually; existing platforms require significant maintenance and are not widely adopted in the social sector. There is a lack of data-driven approaches to recognize needs early on, identify gaps in services, and develop new services in a targeted manner. Those seeking help usually have access to information only through service providers, which severely limits their independent participation. Digital support services, such as health or inclusion apps, are rarely taken into account by existing platforms. All of this means that available services often go unused despite high demand, and those seeking help must wait a long time for appropriate support.

Project Objectives
KIDaS is developing an AI-powered data platform that facilitates and analyzes social services and systematically promotes innovation in this field. The platform brings together service seekers, service providers, and service administrators, enabling the targeted exchange of data on service offerings, needs, and demand to better balance supply and demand. AI agents based on generative AI automate the collection and maintenance of service data, thereby significantly reducing manual effort. An innovative “Service Sonar” analyzes publicly available data sources to identify trends and unmet needs, and uses this information to generate proposals for new social services. A matching service with a service configurator connects people seeking assistance with suitable services in a targeted manner. From a scientific perspective, the project develops design principles for the use of agent-based AI on platforms, a method for data-driven innovation in social services, and a transferable process model for establishing digital service platforms. In the long term, the platform is intended to serve as a public-interest-oriented hub that promotes knowledge transfer among stakeholders and establishes a network of expertise among regional social welfare agencies.

Approach
Four consortium partners are collaborating in an iterative and practice-oriented manner: FAU Erlangen-Nürnberg and the University of Kassel as academic partners, and ANLEI-Service GmbH and Dataciders PRODATO as implementation partners. The consortium is complemented by associated practice partners: the LWV Hessen, Bathildisheim e.V., the Rhineland Regional Council, and the Westphalia-Lippe Regional Council. The research approach combines Action Design Research with Value-Sensitive Design and ensures that technical innovation and ethical principles are given equal consideration. During so-called “e-Van tours,” the project team visits social service organizations on-site to identify requirements, gather feedback, and build a network of expertise. Development proceeds in stages: from requirements analysis and technology assessment, through the development of the platform and AI-based data crawlers, to the core services and the rollout strategy.

Results and Application Potential

The project delivers an expandable data platform with integrated AI components that significantly reduce the manual maintenance workload, precisely match those seeking assistance with appropriate services, and support decision-makers with data-driven recommendations for new services. In addition, the project develops transferable scientific methods for the use of agent-based AI, for data-driven service innovation, and for the sustainable establishment of digital service platforms. Through a targeted openness strategy and an incentive system, the platform is positioned as an incubator for social tech innovations and, not least, strengthens digitalization in the social sector.
 

Short description:

The research project Artificial Intelligence for Foreign Object Detection in Filled Beverage Bottles (KI4FKD) is a joint project between the Department of Intelligent Embedded Systems (IES) at the University of Kassel and miho Inspektionssysteme GmbH from Ahnatal near Kassel. This project is funded by WIBank under the Distr@l funding program of the Hessian State Chancellery in the area of the Minister for Digital Strategy and Development.

The aim of the project is to develop an inline foreign body detection system based on artificial intelligence (AI). The focus is also on the interaction between humans and AI for a high level of user-friendliness and continuous improvement.

Project partners:

• University of Kassel - Intelligent Embedded Systems IES
• miho Inspektionssysteme GmbH

Short Description:

The goal of this three-year project is to combine methods of continuous, adaptive learning and “machine learning in operation” to meet the growing demands of a highly data-driven energy system in terms of automation, scalability, and operational resilience. The intelligent utilization of flexibility by new consumers, generators, and storage systems is placing increasing demands on operational planning and, consequently, on the predictability of an ever-growing number of components and actors in the energy system. At the same time, the high dynamics of the system require constant real-time adaptation and expansion of planning and forecasting models. The associated processes, such as model training runs and evaluations, must therefore inevitably be handled reliably and securely by highly autonomous, computer-aided procedures. The subproject at the University of Kassel focuses on the development of algorithms and models using methods of machine learning and artificial intelligence. 

The focus is on methods of transfer learning and representation learning to facilitate the reuse of learned knowledge, methods of self-monitoring and self-awareness in technical systems, enabling the learning system itself to determine the need for model adaptation (e.g., for forecasting models), as well as methods of continuous learning aimed at self-reflection, so that the learning system can adapt to new conditions (e.g., new facilities) and guarantee the required performance. The project results include implementations of corresponding algorithms and models, which are being investigated in collaboration with project partners using use cases for wind power forecasts, photovoltaic forecasts, and forecasts of vertical power flows at transformers.

The LIFT project aims to improve self-directed learning phases in the flipped classroom through the use of a Gen-AI-based tutoring system. The system employs a multimodal approach and creates a learning-friendly environment that is used in conjunction with existing course materials such as videos or presentations. Instead of merely passively consuming content, students can receive targeted feedback on their work, prepare for exams in a structured manner, and gain an overview of the course at any time. In this way, they are actively involved in shaping their own learning path, which strengthens both their motivation to learn and their long-term learning success.

Short Description:

The goal of DyNaMo is to use a holistic socio-technical approach, incorporating AI, to explore a wide range of methods and measures aimed at significantly increasing traffic safety—and thereby fostering sustainable mobility for the city of tomorrow—using bicycle traffic as a case study.

The research objectives of the LOEWE DyNaMo research initiative are (1) to investigate AI-based detection of cyclists’ riding behavior. This behavior—such as raising an arm to signal a turn—is captured using vehicle- and wearable-based sensors and analyzed using various AI methods (e.g., transformers, hyperparameter optimization, transfer learning, sensor fusion). The data can then be used to (2) investigate effective infrastructure and training measures that will substantially contribute to traffic safety and safe, sustainable mobility in the cities of tomorrow. To address these goals, expertise in computer science, traffic psychology, transportation planning, and law must be combined.

Short Description:

By 2050, about two-thirds of the world’s population will live in cities. In 1950, that figure was 30 percent; by 2010, it had already reached about 50 percent. The continued growth of cities is only possible through more efficient design of (critical) infrastructure such as energy, transportation, logistics, and water. Information and communication technologies (ICT) offer a wide range of opportunities to achieve these efficiency gains and serve as a catalyst for digital cities. However, this development also poses a threat to the functioning of cities in crisis situations. Increasing interconnectedness and dependence on digital services make societies more vulnerable to disruptive events that affect regular ICT functions. ICT-based infrastructures are threatened by human-caused disruptions, natural disasters, violence, and terrorism. At the same time, the functioning of digital cities in the event of extreme events, crises, and disasters has not yet been sufficiently researched. The LOEWE Center “emergenCITY – The Resilient Digital City,” funded by the Hessian State Government, aims to investigate the fundamentals, methods, and solutions needed to establish a novel approach to crisis management through ICT-based infrastructures. This approach is referred to as “Resilient Digital Cities.” In this context, resilience refers to a system’s ability to absorb crises, recover from them promptly and sustainably, or achieve comparable or new levels of functionality through transformation.

Short Description:
Data and metrics have always played a significant role in the music industry, for example, in measuring the market success of releases through chart rankings. However, just as in other areas of society, the digitalization process—and in particular the proliferation of streaming services—has led to a massive expansion in the collection and use of (behavioral) data within the music industry as well. While a range of euphoric and dystopian assessments of this development are circulating, relatively little is known about how data is actually constructed, used, and evaluated by various stakeholders in the music industry. This project addresses this research gap. Qualitative interviews with various stakeholders in the music industry, as well as comparative case studies of different music analytics firms, serve as the empirical starting point.

Project Partner:  Professor Dr. Martin Potthast; Professor Dr. Henning Wachsmuth 

Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing 
General and Comparative Linguistics, Experimental Linguistics, Typology, Non-European Languages 

Piloting a cooperative open web search infrastructure to support Europe's digital sovereignty.
Partners: M. Granitzer (Passau University), B. Stein (Bauhaus-Universität Weimar) and 11 other European partners

Short Description:

The PharmaXR research project, funded by the Federal Ministry of Research, Technology, and Space, is developing a digital training environment designed to provide targeted training for employees. This enables companies to begin training their staff using eXtended Realities as early as the planning and setup phases of production processes—such as those for vaccines—without having to wait for the physical infrastructure to be in place. Through the use of AI-based activity recognition combined with XR, employees can be better trained to perform safety-critical work steps correctly. This results in significant time and quality benefits.

ComtTec is conducting research into AI-based algorithms for activity recognition using wearables. By capturing data with devices such as smartphones and smartwatches, activities—i.e., work steps—can be recognized. This makes it possible to ensure the quality of production processes as early as the training phase using XR and AI technologies. It also provides the foundation for a quality assistance and documentation system.

Project partners:
PTS Training Service, FLAVIA IT-Management GmbH, Sigmund Lindner GmbH, Universität Kassel (FG Arbeits- und Organisationspsychologie), Wirtschaftsgenossenschaft deutscher Tierärzte eG, als antizipierter Partner die CUP Laboratorien Dr. Freitag GmbH

Short Description:

The PIONIERIN project (a program designed to inspire and guide female high school students toward computer science studies) aims to pave the way for talented and interested female high school students aged 17 and older to pursue computer science studies, with the long-term goal of increasing the proportion of academically qualified female computer science professionals and thereby contributing to the equal participation of women in research and innovation processes. In digital fabrication workshops, the students actively engage with computer science content that bridges the gap between previous (extracurricular) educational opportunities and future university courses during their degree programs. Furthermore, the students learn to view a computer science degree—through role model presentations and examples of computer science career paths, among other things—as a viable future path for themselves; they become familiar with the university as a place of learning; and they build their own network with like-minded peers.

The workshops focus on a core topic that is both a fundamental part of a computer science degree program and closely linked to the respective research group. The Cognitive Systems research group in Bamberg is developing workshops on the topic of artificial intelligence; the Didactics of Computer Science research group in Hildesheim is focusing its workshops on environmental informatics; and the Computer Science and Its Didactics research group in Koblenz will develop workshops on programming and modeling. From a research perspective, the workshops will be qualitatively evaluated by the Participatory IT Design research group in Kassel—with a particular focus on intersectional perspectives.

Brief description:

Identifying the effects of digitization at an early stage, assessing them, and designing them in such a way that privacy, self-determination, and democracy gain from digitization is of high research priority. So far, such efforts have only been undertaken by the Forum Privacy and Self-Determined Living in the Digital World, in which eight research institutions collaborated from 2014 to 2021.

In 2021, BMFTR funding was further intensified and bundled under the umbrella of a "Privacy Platform." The previous tasks of the "Forum Privatheit" in terms of intra-scientific exchange, scientific communication and as a group of experts commenting on current issues of data protection and privacy have been taken over by a scientific coordination (acronym: PlattformPrivat) in the sense of project or program support since April 2021.

The Privacy Forum bundles the research projects funded by the BMFTR in the thematic area of privacy and digitization, especially those with an interdisciplinary approach. For these projects, the basic framework conditions (such as cooperation with scientific coordination) are defined in a framework announcement. At least once a year, a funding announcement is published, from which about four projects should emerge each time.

Website of the Forum Privatheit

Project partnersFraunhofer Institute for Systems and Innovation Research ISI, Karlsruhe, Germany

Press releaseResearch for a functioning democracy

MOTIVATION

Until a few years ago, eye-tracking technologies were primarily used in very specialized applications. However, advances in miniaturization and algorithms have led to a rapid expansion of potential areas of application. The analytical capabilities of eye tracking in everyday life carry the risk of significantly infringing on people’s privacy. For example, eye tracking can be used to draw conclusions about a person and their intentions. Additionally, there is a risk that people’s behavior could be subtly influenced based on the information gathered—for instance, during elections or when making purchasing decisions.

GOALS AND APPROACH

The goal of “PRETINA” is to provide concrete practical recommendations for the privacy-preserving and legally compliant use of eye tracking by developers for users. To this end, the implications of the technology, potential privacy risks, and how these risks are perceived are examined holistically. At the same time, the researchers use these findings to derive criteria for the use of eye tracking in a manner that complies with data protection regulations and is socially acceptable. These criteria are incorporated into concrete, technical solutions.

INNOVATIONS AND PROSPECTS

This research helps protect citizens when they use eye-tracking technology. It also supports developers of eye-tracking applications in fulfilling their ethical, legal, and social responsibilities and in implementing data protection measures through technical means. Ultimately, this is intended to enable companies to offer eye-tracking technologies and products that combine data-driven value creation and advanced business models with effective privacy protection.

Network Coordinator
Fraunhofer Institut für System- und Innovationsforschung (ISI), Karlsruhe

Project Partners
Universität Tübingen – Internationales Zentrum für Ethik in den Wissenschaften
Blickshift GmbH, Stuttgart
Universität Stuttgart – Institut für Visualisierung und interaktive Systeme – Abteilung Mensch-Computer-Interaktion und kognitive Systeme
Universität Kassel – Fachgebiet Öffentliches Recht, Projektgruppe verfassungsverträgliche Technikgestaltung

Brief description:

Funding guideline "Platform Privacy - IT Security Protects Privacy and Supports Democracy" as part of the Federal Government's research framework program on IT security "Digital. Secure. Sovereign"

Network partner:

Stefan Voigt, Board of Directors, Open Search Foundation e.V.

Short description:

Due to the important role that communication networks, and the Internet in particular, play in our society, such failures are seen as significant threats for our daily life’s. The main reason behind these failures? Networks are becoming more and more complex. While network operators are already struggling to control and to configure well-established protocols to patch failures or to prepare the Internet for the future, new technologies such as reconfigurable, softwarized, and programmable networks increase the amount of possible configuration knobs.

Yet, network operators get left behind with their traditional network operation tools and solutions. Machine learning based network management solutions propose a way to handle complexity in network management. Yet, ML can itself harm network resilience by making wrong decisions and being understood as black-box solutions that cannot be fully understood.

The goal of this project is to improve network resilience by identifying situations in which machine learning helps in making networks more resilient. To realize this goal, this project aims at deriving a methodology for performing risk assessment of (ML-driven) network management approaches. The influence of a given network management approach (e.g., ML-based) on network resilience will be captured in a resilience function that is derived by our methodology.

In a second step, we capture the non-trivial interaction effects that can occur when deploying ML-based network management functions independently in different networks. To exemplify our methods, we target a wide area networking use case with focus on two problems: i) Internet routing and ii) attack detection and mitigation. We argue that ML can be suitable for managing complex networks, when it can be sufficiently controlled. The application of ML should actually increase but not harm network resilience. Our framework helps to realize the vision of having resilience-by-design when new ML-based solutions are proposed: it should become an integral part of analyzing ML solutions before putting them into effect.

Brief description:

Funding guideline "Plattform Privatheit - IT-Sicherheit schützt Privatheit und stützt Demokratie" within the framework of the Federal Government's research framework program on IT security "Digital. Secure. Sovereign"; total volume: ~ € 1.24 million, own contribution: ~ € 312 thousand (incl. project lump sum)

Network partners:

* Jessica Szczuka, INTITEC (Intimacy with and through technology) Junior Research Group, University of Duisburg-Essen (Coordinator),

* Joel Baumann, New Media, School of Art and Design,

* Christof Paar/Theodor Schnitzler, Research Center Trustworthy Data Science and Security, Ruhr University Bochum

Short Description:

In the recently launched collaborative project “SOURCE,” partners from the research and professional communities are working together to develop a Europe-wide research and analysis infrastructure designed to help combat targeted disinformation campaigns online. This infrastructure is intended to automatically collect large volumes of web and social media data and analyze it using innovative AI methods. Both content and dissemination channels will be analyzed to expose inauthentic information campaigns and strengthen Europe’s digital sovereignty.

Disinformation—that is, the targeted dissemination of information with the intent to deceive and mislead—poses a challenge to social cohesion, democratic processes, and trust in (digital) media. The enormous volume of online content makes it difficult to detect and analyze coordinated information campaigns on the web at an early stage. This is a problem that is further exacerbated by the proliferation of generative artificial intelligence.

“The goal is to provide researchers, fact-checking organizations, the media, civil society, and government agencies with a tool to detect potential disinformation on the web at an early stage, trace its origin, and develop effective countermeasures,” explains Prof. Dr. Martin Potthast of the University of Kassel.

The University of Kassel is participating in the project through its “Deep Semantic Learning” research group, which is funded by hessian.AI and is a member of the Scientific Center for Information Technology Design (ITeG). Prof. Potthast’s research group focuses on content-based disinformation detection. This includes the detection of AI-generated texts, stylometric analyses to determine authorship, and text watermarking techniques. The research group is also contributing its expertise in the field of argumentation mining to the project.

Short Description:

Video-based disinformation and deepfakes can pose a particular threat to young people. Today, video platforms such as TikTok—which are especially popular among young people—are the primary drivers of rapid, mass dissemination. The fast, algorithm-driven spread of short videos makes it possible to quickly make false or distorted information accessible to a large audience. Particularly problematic are emotionally charged content and AI-generated deepfakes, which appear deceptively real and can be used for political influence, defamation, or opinion manipulation. Furthermore, influencers can deliberately spread this content, often without disclosing their motives or sources of funding, which—given their reach among adolescents—poses a significant risk of manipulation.

The Interdisciplinary Collaborative Project

As part of the collaborative project “Video Use in Adolescence: Manipulation, Propaganda, Information, and Resilience (VAMPIR),” the project aims to examine technical, societal, and legal measures to counter video-based disinformation and deepfakes. The core objectives are to identify and analyze disinformation strategies and to develop AI-supported methods for detecting deepfakes. In addition, the project will investigate how visually conveyed disinformation differs psychologically from text-based disinformation and what specific effects it has on adolescents. Finally, the legal framework for detecting and combating video-based disinformation and deepfakes will be examined. Building on these findings, joint recommendations for action will be developed for lawmakers and platform operators, including strategies to strengthen young people’s media literacy.
 

The Legal Subproject

The legal subproject is responsible for developing proposals for legal measures that are consistent with the empirical findings of the partner disciplines.

The University of Kassel will first examine the legal options for accessing data from matching services for researchers and apply for the project consortium’s access to relevant services (see Art. 40 DSA). In addition, data protection requirements for the project infrastructure will be defined, and the specific protection needs of minors will be identified. This will involve close collaboration with Fraunhofer SIT to assess existing solutions for detecting deepfakes from a data protection perspective and to evaluate the “privacy-by-design” approach.

The focus of the legal subproject is the analysis and evaluation of the legal framework for detecting and combating video-based disinformation and deepfakes. The analysis begins by examining the requirements of German constitutional law and EU primary law. It then examines specific provisions of current EU secondary law (e.g., the Digital Services Act, the General Data Protection Regulation, the AI Regulation, and the Audiovisual Media Services Directive). At the national level, particular attention will be given to the DDG, the State Media Treaty, and the State Treaty on the Protection of Minors in the Media. In particular, the study will assess whether these legal acts contain sufficiently concrete provisions for protecting minors from video-based disinformation, with an in-depth analysis of regulations on age restrictions and risk management obligations. In addition, the requirements of the AI Regulation will be examined, such as the obligations to label synthetic content and deepfakes. Soft-law instruments (e.g., guidelines, codes of conduct, and crisis protocols) as well as national self-regulatory bodies must also be included in the examination of the legal framework, as must the interplay between state and private regulation (hybrid governance). Furthermore, the implementation and oversight of the regulations will be evaluated, and current EU proceedings against very large online platforms (e.g., TikTok, Meta, X) and comparative analyses of third countries will be conducted.#

Finally, proposals for legal policy reforms will be evaluated (e.g., reform of the AI Regulation, the draft Digital Fairness Act, and legislative initiatives regarding the criminalization of deepfakes). In particular, the study will determine what scope remains for national regulations to combat video-based disinformation and deepfakes.

Building on these studies, proposals for legal measures will be developed based on the empirical findings of the partner disciplines. Both preventive and repressive measures tailored to the specific protection needs of young people will be identified.
 

Motivation

The increasing digitalization and interconnectedness of our world are enabling groundbreaking innovations across industries, from medicine to mobility. As applications like virtual reality in medical technology and remote control of autonomous vehicles become more prominent, ensuring minimal latency is critical. Even delays of just a few milliseconds can have significant consequences. Current internet protocols often fail to adequately address this challenge, especially in networks with numerous nodes. To support the demands of future applications, new approaches for optimizing latency in optical networks are essential.

About XLerate

Our project, "IXP-Based Optimization of Latency and Routing" (XLerate), focuses on optimizing BGP route selection processes at Internet Exchange Points (IXPs). One objective can be to drastically reducing delays in fiber-optic networks by optimizing BGP routes for latency. We are developing innovative measurement techniques to determine the minimum achievable latency through IXPs. These techniques are designed to scale across an IXP's full connection range while being resource-efficient to minimize infrastructure load. Using data from extensive measurement campaigns, we will analyze latency stability and influencing factors to create advanced routing algorithms. These will be integrated into a test infrastructure to showcase their potential.

Kooperation with DE-CIX

The XLerate research project is conducted with DE-CIX, one of the world's largest IXPs. Together, we work on making BGP route selection more flexible to meet the demands of emerging Internet applications.

Short Description:

The ZuKIPro future center will create a practice-oriented format for consulting, qualification, testing and diffusion of digital technologies. To this end, participatory work and technology design is intended to enable small and medium-sized enterprises to make better use of the potential of digital technologies in work and business processes. ZuKIPro is intended to be a contact and transfer point for SMEs, funding agencies, technology providers and social partners. Via a platform module, SMEs can bring a mobile AI lab directly onto the company premises and experience prototypical AI applications, try them out and consider their applicability in their own operations. Planned qualification offerings, such as a learning and knowledge platform or Massive Open Online Courses (MOOCs), bring the topics of digitization and artificial intelligence to life for employees. Through these formats, the Future Center enables companies and their employees to recognize and analyze the challenges in their own operations, to independently initiate, design and implement learning and solution processes, and to evaluate the results.

The overarching goal of the project is to create low-threshold access to the topic of AI for SMEs. To this end, SMEs are supported in the participatory, co-creative introduction and human-centered design of AI systems. Across modules, the Future Center informs and advises SMEs according to their needs and enables them to implement concrete solutions. To this end, a total of three offices will be established: Northern Hesse (Kassel), Central Hesse (Frankfurt a. M.) and Southern Hesse (Darmstadt).

Consortium: - Machine Tool Laboratory WZL of RWTH Aachen University - Chair of Production Metrology and Quality Management and Chair of Manufacturing Process Technology - German Research Institute for Artificial Intelligence (DFKI) - Smart Service Engineering Research Group - University of Kassel - Departments of Information Systems, Human-Machine Systems Engineering, Communication Technology (ComTec) - Darmstadt University of Technology - Institute of Industrial Engineering - Institute for Technology and Work e.V. - IHK Kassel-Marburg and IHK Hessen-Innovativ - Regionalmanagement Nordhessen GmbH