Multi-Agent Systems

How can we as humans interact with intelligent software agents? What opportunities and challenges does collaboration with such systems offer? In what way do the different personality traits, skills, and knowledge resources of these agents influence the collaboration? How should we shape our interaction with such systems?

For many, large language models (LLMs) are tools of everyday use and terms such as “prompt” could soon find their way into the dictionary alongside “google” or “surf”. Multi-agent systems mark a further step in this development by introducing the paradigm of the “chain of thought” - a multi-stage process that is comparable to a dialog between human experts. Different AI agents, based on LLMs, negotiate with each other about the optimal solution to a problem. 

Such processes of negotiation shape our world, whether on a social, political or economic level. By dividing the workload, multi-agent systems can solve complex tasks. They use step-by-step chains of reasoning to develop solutions that offer high reliability through repeated review and adaptation of outputs. Agents communicate using natural language, thus enabling human involvement in these processes and providing new opportunities for creative problem solving in human-machine interaction.

The event Multi-Agent Systems. Let's Talk? explores the interface between humans and AI in both artistic and technical terms. A first prototype of a multi-agent system will be presented during the event, which forms part of an interactive media art installation by Daniel Heiss (ZKM | Karlsruhe) and Yannick Hofmann (Fraunhofer Institute for Industrial Engineering).

The installation simulates negotiation processes between groups consisting of both human and non-human agents. In a live simulation, participants will experience how different AI agents work together in solving tasks. The focus here is on the possibilities for interacting with such systems, as well as their configurability. How can the artistic, technical, and ethical aspects of human-machine interaction be shaped to meaningfully influence the development of these relationships? We invite you to share your perspectives in an open discussion and collectively seek answers to these complex matters.

Digitalization in Dialogue – digilog@bw

This event is taking place within the framework of the project digilog@bw, a research network of Baden-Württemberg universities, research institutions and ZKM | Karlsruhe, in which more than 50 scientists have been involved since 2019: from the humanities, social sciences, law, economics, media and communication sciences, ethics and computer science, as well as interdisciplinary technology.