Hauptinhalt
Topinformationen
31st Systems Science Colloquium
Winter semester 2024/25
The colloquium takes place on Wednesdays from 16:15 to max. 17:45, in room 35/E16 (biology building, Barbarastraße 11).
Program
30 October 2024
Preparatory meeting for students
27 November 2024
Dr. Hugo Martin, School of Public Health, University of Rennes (France)
04 December 2024
Prof. Dr. Ezio Venturino, University of Turin, Department of Mathematics (Italy) and University of Franche-Comté, Chrono-environment Laboratory, Besançon (France)
An overview of modeling case studies in current mathematical biology
08 January 2025
Kai Neumann, Consideo GmbH, Lübeck (Germany)
Practical scientific application of systems thinking and modeling
15 January 22025
Dr. Felix May, Institute of Biology, Free University of Berlin (Germany)
22 January 2025
Dr. Thomas G. Preuss, Environmental Effects, Bayer AG (Germany)
29 January 2025
Prof. Dr. Nico van den Brink, Division of Toxicology, Wageningen University (The Netherlands)
05 February 2025
Dr. Johannes Zimmermann, Junior Group Leader: Mechanisms of Microbial Metabolic Interactions, Cluster of Excellence: Balance of the Microverse, Friedrich Schiller University Jena
Microbiome Life History
12 February 2025
Prof. Dr. Sebastian Musslick, Institute of Cognitive Science, Osnabrück University (Germany), and Brown University (USA)
Automated scientific discovery of the human mind
Abstracts of presentations
27 November 2024
Dr. Hugo Martin, School of Public Health, University of Rennes (France)
Wearing face masks to protect oneself and/or others: Counterintuitive results from a simple epidemic model accounting for selfish and altruistic human behavior
We study a simple SIS (susceptible-infected-susceptible) epidemic model accounting for human behavior. Individuals can decide at each instant of time whether or not they adopt prophylactic (hereafter protection) measures such as mask wearing or social distancing. These measures decrease susceptibility and/or transmission. We consider a situation in which individuals are unaware of their current health status (infected or not), but can perceive disease prevalence at the population level. This assumption fits situations in which tests are not widely available. Thus, personal decisions depend first on disease prevalence, as a proxy for the risk of being infected or infecting others, and second on the fraction of the population complying to the protection measure, which people can observe in their every day life. Human behavior is assumed to be driven by imitation dynamics (Bauch, 2005; Poletti et al., 2009).
When the disease does not naturally die out, the model has three types of endemic equilibria: no-protection, mixed-protection, and full-protection. Which endemic equilibrium is stable depends on the parameter values. We assume that the efficiency of the protection measure is positively correlated to its individual cost. Increasing the efficiency of the protection measure and therefore its individual cost can make the system switch from full protection to mixed-protection. This way, increasing the efficiency of a protection measure may increase disease prevalence at equilibrium. In other words, disease prevalence is minimized for intermediate efficiency, and individual cost, of the prophylactic measure. The rational is that when the prophylactic measure is too effective and therefore costly, part of the population free-rides on the effort of others and drops protection, resulting in increased prevalence.
Altogether, our results show that the interplay between epidemiology and human behavior may lead to counterintuitive but nevertheless intelligible outcomes, that should be anticipated when designing public health policies.
This work is shared with François Castella and Frédéric Hamelin.
4 December 2024
Prof. Dr. Ezio Venturino, University of Turin, Department of Mathematics (Italy) and University of Franche-Comté, Chrono-environment Laboratory, Besançon (France)
An overview of modeling case studies in current mathematical biology
The role of models in science is known since centuries. However, the application of mathematical tools to life sciences is much more recent, dating back about a century with seminal works in interacting populations and epidemiology. Here, the attention is directed to some important environmental applications.
We present how to deal with pests in agriculture via biological controls, rather than using chemicals. Unfortunately the latter end up on our tables through the food web and slowly poison us as well. Among pests, nowadays also alien species play an essential role. These are imported from exotic places, generally in a-non intentional way, via global trading. They find suitable living conditions at our latitudes and prosper at the expense of the native species. In some unlucky cases, their negative impact on the native populations entails even their eradication. These effects are also favored by the ongoing climatic changes. A quick look at the latter shows pitfalls in our naive way of thinking, for which the impact of these phenomena is underestimated among the ordinary people.
We will also show how to take advantage of fungi for specific applications, ranging from their use in agriculture as pest antagonists, to the depuration processes of industrial wastewaters.
The fight against a disease affecting goats, discovered only fifty years ago, is illustrated as a final application of mathematics to farming issues.
08 January 2025
Kai Neumann, Consideo GmbH, Lübeck (Germany)
Practical scientific application of systems thinking and modeling
25 years of experience from systems thinking and modeling as a business consultant and scientist bring a lot of practical insights not just in how to apply the methodologies but also why they are barely used. It can be very frustrating to be the only systems thinker in the room whether in the business world or in science. I will explain the difference between qualitative (from CLD [causal loop diagrams] to FCM [fuzzy cognitive maps]) and quantitative (SD [system dynamics], ABM [agent-based models], NN [neural networks]) modeling and the crucial implications from distinguishing descriptive and explorative modeling, the latter requiring not just modeling but also systems thinking. I will ask the question if models can be validated at all and explain when they are scientifically sound. I will bring three practical applications for the German Environment Agency (UBA): the transformation of society, the global potential to feed the world, and the need for resources for the global energy transition. The transformation of society was first explored qualitatively and then quantitatively with a huge system dynamics model the features the different parts of society, the economy and jobs, the state of the environment and resources (as index values), the development of the economy, welfare (using the national welfare index, NWI) and happiness (as an index). It is the kind of model needed for the current polycrises in the world. But it is not having a model - it is its communication.
15 January 2024
Dr. Felix May, Theoretical Ecology, Institute of Biology, Free University of Berlin (Germany)
Synthesis of fragmentation-biodiversity relationships across scales, study designs and ecological contexts
The effects of habitat fragmentation on biodiversity have been debated for decades, with studies reporting positive, negative, and neutral relationships. We argue that much of the confusion and (apparent) contradictions arise from three key issues: (1) inconsistent terminology, (2) inappropriate comparisons across different spatial scales and sampling designs, and (3) different ecological contexts. In this seminar, I will address these three issues and present our research on fragmentation and biodiversity, which includes a synthesis of empirical studies, mathematical modelling and process-based modelling. We introduce a framework that distinguishes between geometric and demographic fragmentation effects, providing new insights into the scale- and context-dependence of biodiversity responses. Using a process-based metacommunity model, we illustrate how these two effects interact to shape fragmentation-biodiversity relationships. We conclude that this framework reconciles seemingly contradictory findings and provides a foundation for advancing the debate on biodiversity change in human-modified landscapes.
22 January 2024
Dr. Thomas G. Preuss, Environmental Effects, Bayer AG (Germany)
Emerging properties of mechanistic models - essential feature for robust predictions for environmental risk assessment
Environmental risk assessment is increasingly relied on mechanistic models to predict the impacts of various stressors, particularly pesticides, on ecological systems. The talk explores the concept of emerging properties within these models. Emerging properties refer to the complex characteristics and behaviors that arise from the interactions of simpler components within a system, which cannot be predicted by examining the components in isolation. These properties are crucial for robust predictions in environmental risk assessment, as they encapsulate the nonlinear and often unpredictable nature of ecological interactions.
Individual-based models (IBMs), also known as agent-based models (ABMs), serve as a powerful framework for studying these emerging properties. Unlike traditional population models that aggregate individuals into a collective, IBMs focus on the behaviors and interactions of individual agents within a defined environment. Each agent, representing an individual organism, operates based on specific rules that govern its movement, reproduction, and interaction with other agents and the environment. This bottom-up approach allows for the emergence of complex population dynamics from the simple rules governing individual behavior.
The integration of toxicokinetics and toxicodynamics (TK-TD) modeling into individual-based frameworks provides a comprehensive approach to assess the effects of pesticide exposure at the individual level. Toxicokinetics describes how an organism absorbs, distributes, metabolizes, and excretes a toxic substance, while toxicodynamics focuses on the biological effects of the substance at the target sites within the organism. By incorporating TK-TD principles into IBMs, researchers can simulate the fate of pesticides within individual organisms and evaluate the resultant biological effects over time. This modeling approach allows for the exploration of how variations in exposure levels, timing, and individual susceptibility contribute to population-level outcomes.
By utilizing individual-based modeling approaches that incorporate toxicokinetics and toxicodynamics, researchers and risk assessors can gain insights into the complex interactions and behaviors that define ecological systems. Understanding these emergent features is essential for making robust predictions regarding the impacts of pesticides and other stressors on populations and ecosystems, ultimately informing regulatory decisions and conservation efforts.
12 February 2025
Prof. Dr. Sebastian Musslick, Institute of Cognitive Science, Osnabrück University (Germany), and Brown University (USA)
Automated scientific discovery of the human mind
Automated scientific discovery represents a transformative approach to empirical research by automating the continuous cycle of data collection, modeling, and experimental design, all aimed at generating new scientific knowledge. In this talk, I will introduce automated scientific discovery as a paradigm for the study of mind and behavior, detailing its potential to advance our understanding of human cognition. I begin by formalizing the behavioral research process as an iteration between data collection, computational inference, and experimental design. I then introduce AutoRA, an open-source framework designed for automating various steps of empirical research, and showcase its utility for advancing our understanding of the human mind, both in terms of discovering novel computational models of cognition and identifying novel behavioral phenomena. Specifically, I will introduce a discovery method that combines recurrent neural network modeling with sparse dynamical system identification to infer latent dynamics of cognitive processes from human choice and reaction time data. I conclude by addressing the specific challenges that automated scientific discovery encounters in the cognitive sciences and by outlining prospective paths for research, with the aim of refining closed-loop discovery systems and underlying AI methods to more effectively contribute to our understanding of human cognition and behavior.
