- Prof. Dr. Marc-Thorsten Hütt, Jacobs University
- Prof. Dr. Wolfgang Marwan, Otto-von-Guericke-Universität Magdeburg
- Prof. Dr. Thorsten Fehr, Universität Bremen
With its domain Transdisciplinary Physics, the NFDI4Phys consortium enables and strengthens bridges between Physics and other disciplines near and far. It gathers all communities and collaborators with whom members of our consortium interact in reaching out across disciplines and spanning a large part of the DFG panel classification scheme. It is built around the unifying power of Physics and its search for common principles and universal laws in systems across a wide range of application domains. We do not claim to cover all interactions of Physics with other disciplines but certainly represent some of the conceptually most important. Specific aims and objectives in RDM may be found under each disciplines page link.
Systems Biology & Systems Medicine
- Prof. Dr. Marc-Thorsten Hütt, Jacobs University
- Prof. Dr. Wolfgang Marwan, Otto-von-Guericke-Universität Magdeburg
Biophysical analysis of the structure and dynamics of gene regulatory networks controlling cell fate decisions and the differentiation of living cells requires the integration of complex, high throughput datasets of markedly different formats and data types. Through mutual interactions with partners of the NFDI4Phys consortium, we will develop innovative concepts leading to a unifying framework which can cope with this unavoidable heterogeneity in data formats. We are also happy to contribute and elaborate approaches for the automatic interpretation of experimental time series data sets with the help of algorithms that compute the structure and dynamics of regulatory networks in terms of executable models directly from the data sets. In general, such executable models represent the structure and dynamics of causal intersections within a system and are relevant for systems of different complexity, from molecular networks up to socio-economic systems and beyond. Indeed, category theory provides a powerful algebraic method for a general classification scheme. It will serve as an universal instrument to establish structural hierarchical order.
- Prof. Dr. Thorsten Fehr, Universität Bremen
Biological Psychology, as one of the main sub domains of Psychology, provides an inherent transdisciplanary bridge between the humanities, social sciences and the natural sciences as it adumbrates and links a variety of aspects from psychological, biological, physical and other disciplines. One sub-domain of the Biological Psychology, the comparative psychology, compares the physiological principles underlying behaviour and experience between species. Different methodological approaches come into play that cover a variety of levels between micro-(i.e., subnuclear) and macroscopic (i.e. complex system interaction) perspectives, such as magnetic resonance imaging, positron emission tomography, electroencephalography, electrocardiography, electromyography, and many more.
Computational Social Sciences
- Prof. Dr. Uwe Engel, Universität Bremen
Computational social science (CSS) is a dynamically developing discipline in the intersection of data science and social science. A lot of social interaction and interpersonal communication takes place exclusively on the internet. This produces digital trace data in the form of texts and/or behavioral marks, which people leave when surfing the web. These digital survey data accumulate on a grand scale. It is the original four Vs of Big Data (volume, velocity, variety, veracity) that make such great demands on a contemporary management of the digital research data in CSS. A transdisciplinary element is provided by “variety”: the sensing of physical data while collecting survey data [Bosse & Engel 2019]. In addition, CSS consists of more elements than big data and data analytics. CSS is also deeply rooted in mathematical modeling and simulation in sociology. Social simulation is a third supporting pillar of CSS with a transdisciplinary core: its multilevel approach towards social complexity; a concept that designates the genuinely emergent aggregate features and inherent regularities of social systems and their major constituents such as social groups, networks, and individual agents. A challenge which is currently tackled in this context is the data-driven performance of social simulation to achieve more than images of artificial societies. [Engel et al. 2021]
Philosophy of Law
- Prof. Dr. Lorenz Kähler, Universität Bremen
- Prof. Dr Hans-Günther Döbereiner, Universität Bremen
Empirical Legal Studies are concerned with general legal phenomena in society. These are expected to follow patterns that correspond to similar phenomena as studied, e.g. in the Domain Socio-economic Systems. In order to conduct our studies, we will benefit from discussions with colleagues from this domain and with members of sections Biological Psychology, Computational Social Sciences, and Digital Humanities within the Domain Transdisciplinary Physics. We will employ the survey methodology available via task area Surveys to conduct our studies remotely. We are looking forward to a FAIR handling of our remote data. Apart from similar disciplinary topics, we share the concern and challenges of data privacy and security. We will potentially utilize data available via interfacing with consortia which provide data from business studies, economics and related fields, integrating expertise from both research and infrastructure. We will utilize integrated management of unstructured (Big) data and algorithms of artificial intelligence provided by interfaced consortia. In addition to pioneering FAIR legal data management already available from field experiments or acquired a new by our surveys, we will uncover hidden correlations in qualitative legal texts. In collaboration with the task areas FAIR Laboratory and Metadata and Ontologies , we will attempt automated semantic analyses of legal documents and built up relevant ontologies. Semantic analysis, which we conduct under close guidance by the SEM working group of the FDO forum, lowers the divide between qualitative and quantitative data. Applying a semantic metric to a qualitative text quantifies it to a certain degree and enables the digital transformation to a FAIR Digital Object which is available on a federated repository. We will apply for additional DFG funds for a research project enabled by integrative RDM within NFDI.
- Wolfgang Schäffner, Humboldt Universität zu Berlin, Director of Matters of Activity
In the Cluster of Excellence »Matters of Activity« researchers from more than 40 disciplines are investigating the activity of matter. Our central vision is to rediscover analog processes in the age of the digital within the activity of images, spaces and materials driven by the integrative nature of the Humanities and with the Natural Sciences and the different Design disciplines. Biology and technology, mind and material, nature and culture intertwine in a new way. In this context, the interdisciplinary research and development of sustainable practices and structures is a central concern in areas such as architecture and soft robotics, textiles, materials and digital filters, and surgical cutting techniques. Objects and materials are not thought of as passive and unadaptable, but rather as active, changeable and recyclable materials. The three closely linked research units »Practices«, »Structures« and »Code« focus on three different approaches to material activity: from the basic level of cultural material practices and the material’s inherent active structures to the challenging idea of a novel kind of material code. Thus, Material Humanities designates a new field which adds to Material Sciences a cultural aspect in contrast to the often purely technological view of the Natural Sciences. It adds a totally new perspective to FDM as qualitative concepts from the Humanities guide quantitative workflows of the Natural Sciences which constitutes a major challenge for the construction of metadata ontologies defining the boundary condition of the creational process inherent in material production.
- Prof. Dr. Christoph Eberl
Materials science is at the crossroad between equilibrium material properties as well as short- and long-term process-driven non-equilibrium materials behavior. Novel Materials, structured on the nano-, micro- and mesoscopic scale, can be manufactured process-driven as non-equilibrium yet multi-stable compounds. Long term material stability and durability depends typically on features like granularity and defect structure. The physics of these phenomenon requires techniques beyond solid state physics or linear elasticity theory. This is the core expertise of material sciences. Furthermore, we can learn and derive effective material models from biological materials and systems (e.g. ExIni Cluster livMatS). Moreover, an integrated socio-technological approach requires transdisciplinary thinking with the human in the loop as future products will be planed and manufactured in a process where architects, engineers, material scientist, as well as psychologists and sociologists work together to adapt material properties and product specifications to human needs and processing boundary conditions. With its transdisciplinary agenda NFDI4Phys provides an ideal platform for such an endeavor.
- Prof. Dr. Lars Hornuf, Universität Bremen
- Prof. Dr. Martin Kocher, Wien
- Prof. Liss C. Werner, Institute of Architecture, Technische Universität Berlin
The advent of Digital Architecture in the 1960s has influenced the discipline of architecture insofar that the creation of buildings, cities, and regions relocated from an analog physical design process to one that was partially supported by digital computation processes (Conway’s Game of Life. MIT Architecture Machine). The late 1990s offered the first steps into CADCAM, known as Computer-Aided Drawing and Manufacturing. In comparison, Digital Architecture in the 2020s engages with the evolution of the very design process as an in-silico computational generation of architecture that integrates social, biological and material intelligence alike. The rising understanding of ecology as computational complexity paired with climatic and population growth issues challenges the discipline. It requires comprehensive sets of disparate data. Digital Architecture aims at designing the transformation of regions, cities, buildings, and construction components as a partial data-driven cybernetic design process. The goal is to learn about the impact and causal relationships of intelligent materials, sensors, biology, AI, biodiversity, population growth, social structures, and climate phenomena on the design process and ultimately the built form – terrestrial and extraterrestrial.
Climate Impact Research
- Prof. Dr. Dr. h.c. mult. Jürgen Kurths, Potsdam Institute for Climate Impact Research (PIK)
It is PIK’s twofold mission to advance the scientific frontier on interdisciplinary climate impact research for global sustainability and to contribute knowledge and solutions for a safe and just climate future. The Research Department 4 “Complexity Science” investigates the properties of the natural and societal complex systems in the realm of climate change and its impacts. In particular, research in RD4 intends to advance the understanding of the complex nature of these systems in order to find new principles that can improve and complement existing modelling and empirical approaches. RD4 explores new territory to find principles, analysis methods, and modelling techniques. With guidelines for Data Management Plans (DMP), PIK provides for its scientists an elaborated framework in terms of data usage, maintenance and dissemination settings to ensure data transparency.
- Prof. Dr. Rolf Drechsler, DFKI & Universität Bremen
- Dr. Lena Steinmann, Data Science Center, Universität Bremen
The cross-cutting discipline data science opens up new possibilities to extract knowledge from heterogeneous, complex datasets using modern analysis methods such as machine learning. Hence, data science is regarded as key discipline of the digital era. Data science requires high-quality data and, thus, a sustainable research-data management as envisioned by the NFDI. The Data Science Center (DSC) is an interdisciplinary institute that acts as focal point for data-driven research and data science at the University of Bremen. Our goal is to strengthen data science in research, education, and application across all disciplines as well as to advance scientific discoveries through cross-disciplinary collaborations. Within the context of NFDI4Phys, we aim to promote data literacy across all status groups and support the use of data science methods, especially from the field of artificial intelligence, in the physics community.
- Prof. Dr. Udo Ernst, Theoretical Neurophysics, Universität Bremen
- Prof. Dr. Andreas Kreiter, Cognitive Neurophysiology, Universität Bremen
- Prof. Dr. Klaus Pawelzik, Theoretical Neurophysics, Universität Bremen
The brain as an information processing machine is one of the ultimate challenges of science. Particularly promising are interdisciplinary approaches where physicists and neurobiologists develop and test quantitative theories to explain neurobiological phenomena and functions of the central nervous system. Also the recent progresses in AI with artificial neural networks turns out to profit substantially from the biological role model. The centre of cognitive sciences (ZKW) in Bremen promotes collaborations among a range of disciplines including computer science, psychology, and even philosophy. A central challenge here are simplified access to data and sharing of models. Within the context of NFDI4Phys, we aim to promote literacy of theoretical approaches across all groups, with an emphasis on the methodology of physics.
- Prof. Dr. HG Döbereiner
- Dr. Christina Oettmeier
Basal cognition is a new field concerned with the characteristics and the development of cognitive abilities, such as intelligence, in non-neuronal and comparatively simple neuronal systems on the single cell and organism level. Model species of interest in this regard are the slime mold Physarum polycephalum and the freshwater polyp Hydra. Although cognitive functions are based on biology, no definition of cognition exists that reflects and acknowledges this fundament. Many definitions are inevitably anthropocentric and therefore impede efforts to see these qualities in other species. We generally regard intelligent behavior as the capacity for solving problems. It seems to be inextricably linked to evolutionary fitness. Intelligence and cognition have evolved independently several times in organisms with a central nervous system, such as in primates, birds, and octopuses. It can be argued, however, that this convergent evolution also lead to the development of cognitive abilities based on non-neural systems. Several instances of learning, memory and intelligence have been found in unicellular and very simple organisms. Clearly, not every behavioural trait is a sign of intelligence, but in an unpredictable environment regarding nutrients, resources, or the presence of predators, purely reflex-based and stereotypical behavior is a threat to survival and maladaptive. Emerging from these observations is the idea that the process of cognition is a universal biological mechanism.
Mechanobiology and Morphogenesis
- Prof. Dr. Martin Grube, Biology, Karl-Franzens-Universität Graz
- Bernat Corominas-Murtra, Biology, Karl-Franzens-Universität Graz
Mechanic forces play roles for the growth behavior and complexity of biological systems. These ranges from unicellular protists or stem-cells to organic hierarchies or symbiotic organisms. In all of those, recent research detected feedbacks of mechanic cues with biochemical processes. Using selected examples, such as embryonic tissue, slime molds, fungal mycelia, or lichen symbioses, we want to explore the role of mechanic regimes for morphogenesis. For example, lichens are complex fungal systems which essentially represent growth chambers for energy-providing photosynthetic algae. Mechanical forces are typical for lichen biology due to their hygroscopic motions under changing hydration conditions, yet they have hardly ever been studied. The possibility of mechanic forces for guiding morphological evolution was already hypothesized for lichens earlier (Grube & Hawksworth (2007). In another system, embryos, dynamic changes in tissue material properties were shown to guide development (Petridou et al. 2021). The study of mechanic cues on the behaviour of living matter is at the crossroads of physics and biology, and best be studied by a blend of theoretical and practical approaches. Our approach will thus combine modelling and simulation, as well as biological observation and experimentation.
Physics of AI
- Prof. Dr. Döbereiner, Semantic Physics, Universität Bremen
- Prof. Dr. Dr. h.c. Frank Kirchner, Robotics, Universität Bremen
Robots need to autonomously navigate unknown terrain in order to be fully functional in general situations. Deep neural networks in concert with an ever increasing computing power have been employed with great success. It has been possible to train neural networks to a degree of sofistication thought impossible just a decade ago. Special methods, like re-enforcement learning have contributed to this. Yet, a child recognizes a dog just after a few encounters. Clearly, we miss yet important ingredients as discussed in (Kirchner 2020). One of these missing links leading to human-like general AI is certainly the dynamics of structure formation. How is information encoded, memorized, and recalled? What distinguishes a simple system distribution of states from a functional unit or intentional planing. The physics of these processes has been largely neglected, but is starting to be investigated rigorously, see the stochastic thermodynamics of Learning (Goldt & Seifert). Physics of AI holds great promises to unravel the mechanisms of creativity and free will (George Ellis) by looking at systems exhibiting basal cognition (Lyon et al.), see section on that field above.
- Prof. Dr. Philipp Slusalleck, DFKI Saarbrücken
- Dr. Tim Dahmen, DFKI Saarbrücken
Deep neuronal network perform exceptional well in feature extraction from general data sets when they are sufficiently trained. However, the availability of training data is often a problem out of various reasons. There is thus a need to simulate reality to create digital training data mimicking reality. Simulating basic physical properties of materials is relative straight forward. In contrast, incorporating real life behavior into Digital Human Models has proven to be challenging. We will explore the reusability and interoperability of the rich data sets available from physical modeling of human social behavior in our domain socioeconomic system. Further, single human behavior is increasingly amendable to a quantitative description, especially in well-defined situations as given in training scenarios. We will test suitability in a number of specific environmental settings and contexts. Finally, one regularly encounters the situation that heuristic models with an effective dimension smaller than the general model by orders of magnitude are already sufficient for feature extraction with comparatively low error rates. This phenomenon has been discussed within the framework of causal emergence in the hierarchical level model of nature, see above and domain . Understanding the physical dynamics of neuronal network is at the core of explainability.
Natural Language Processing
- Prof.Dr. Philipp Wieoer, Deputy Director GWDG, Göttingen
- Alireza Zarei, GWDG, Göttingen
For centuries, writing has been the main tool for scholars to communicate ideas, methods and results . However, the ever growing amount of text in different scientific disciplines has not made it easier to evaluate its content and come up with new inputs in a short time while having an understanding of the state-of-art in the same and even other scientific disciplines. In the last years, interest increased to extract knowledge from text in general and to create structured forms of knowledge with the help of machines. The possibility to first extract knowledge from unstructured text i) in different disciplines ii) in a matter of seconds iii) in a structured way, and then perform semantic evaluation of different sources on a topic from different points of view, can open doors for scholars which have been (almost) closed before. Working transdisciplinary, i.e., with the help of respective experts, based on available textual sources, we will create probabilistic knowledge graphs indicating what we know so far and how certain we are about this knowledge. Such evaluation of human knowledge with regard to different contributors and different disciplines has not been possible before, but given the advances in natural language processing, now is the time for it.
- Prof. Dr. Malte Lochau, Universität Siegen
Adherence to formalized quality standards for representation, management, storage and maintenance of research data, as demanded by the NFDI initiative of the DFG, will soon become obligatory in many scientific domains. Physical experiments should be conducted and documented in a way such that the results are available, replicable – and even reproducible – for the research community. The FAIR data principles propagate four essential quality criteria for research data, namely: Findable, Accessible, Interoperable, Reusable. In order to establish FAIR data management in a research domain, a unified conceptual and technical infrastructure is required for experimental data management. We develop an agile data description framework for transdisciplinary physical experiments, based on the resource description framework (RDF). RDF defines a comprehensive methodology for ontology-based knowledge and data management, having a powerful graph database and data inference engine as a backbone. Our goal is a fully-fledged experimental data management tool being useable by practitioners:
• An adapter framework for bidirectional data format transformation between off-the-shelf experimental tools used in physics research and RDF models.
• A modeling framework for editing, composing, validating and evolving experimental knowledge and experimental instance descriptions based on RDF ontology and data construction principles and corresponding composition operators and constraint-verification capabilities.
• Integrated visual support for graph-based RDF knowledge and data management based on existing tools.
- Prof. Dr. John Bateman, U Bremen
- Prof. Dr. Rainer Malaka, U Bremen