Wintersemester 2018/2019
Renke Reinstorm, Universität Bremen, Philosophie, AG Prof. Dagmar Borchers
Abstract
Renke Reinstrom erstellt eine Typologie des Nicht-Entscheidens. Während das Entscheiden in seinen unterschiedlichen Formen und Praktiken erforscht wird, ist das Nicht-Entscheiden bis jetzt wenig untersucht worden: So basiert das Projekt auf der Annahme, dass es Formen und Verfahrensweisen gibt, die darauf ausgerichtet sind, Entscheidungen nicht zu treffen. Hierbei kann an die Entscheidungsvermeidung oder den Entscheidungsaufschub gedacht werden, aber auch an manipulative Praktiken wie die Beeinflussung von zur Wahl stehenden Alternativen. Das Forschungsprojekt ist interdisziplinär ausgerichtet, so gibt es in einigen Disziplinen bereits vereinzelt Forschungsansätze zu Nicht-Entscheidungsprozessen, eine zusammenhängende Aufarbeitung des Themas ist jedoch noch nicht vorhanden.
Hans Hohenfeld, DFKI, Universität Bremen, AG Prof. Dr. Frank Kirchner
Abstract
Over the past few years Deep Reinforcement Learning based on Artificial Neural Networks, attracted significant research to learn and solve tasks in computational as well as robotic environments with a multitude of applications. Given the huge progress in the field of robotics in e.g., autonomous driving, logistics, search and rescue as well as human assistance systems and medical applications, the necessity for scenarios with multiple learning agents in a shared environment arises. While single agents learn based on a simple utility maximization scheme, Multi Agent Systems pose additional challenges. Each agent has to observe, predict and incorporate other agents’ behaviors and possible intentions in its own decision-making while simultaneously exploring hierarchy and structure of a given task. Furthermore agents in a shared environment have to decide on the subdivision and distribution of subtasks, requiring means of coordination and possibly communication. During the talk the state of the art in Multi Agent Learning and decision-making as well as open research questions to be addressed from an interdisciplinary perspective will be presented.
Ann-Marie Parrey, Institut für Theoretische Physik, Universität Bremen, AG Prof. Dr. Klaus Pawelzik
Abstract
What causes an event? How can we quantify that causation?
Ever since the beginnings of natural philosophy, the concept of Causality has fascinated humans. But the quantification of that concept has only come much later: in 1956, Norbert Wiener introduced the idea that when the inclusion of a variable (X) into the prediction of another (Y) helps in improving that prediction, then X ‘causes’ Y.
Nowadays, there are many different ways to calculate causal influence between systems. Two are introduced in this talk: Granger Causality, introduced by Clive Granger, which is a mathematical formulation of Norbert Wiener’s idea of causality, and Topological Causality (TC). TC was only recently introduced by Klaus Pawelzik et al and takes a rather different approach than Granger Causality. The two approaches will be explained, compared and their differences in various aspects discussed.
Jonghyun Lee, Institut für Biophysik, Universität Bremen, AG Prof. Dr. Hans-Günther Döbereiner
Abstract
While the concept of intelligence may be intuitively clear, its concrete definition is still elusive to this day. With the emergence of artificial intelligence as well as continued research on the cognitive abilities of nonhuman subjects, more and more organisms are referred to as ‘intelligent’. What does this really mean? How does one measure and test for this quality, and how does one compare this between organisms?
In this talk I discuss the behaviours of a single-celled organism called ‘Physarum polycephalum’, and whether these capacities can be considered ‘intelligent’, or at least in part a cognitive ability. I compare its behaviours with previous studies of animal cognition and see whether any parallels can be drawn.