Diemo Urbig

About the Power to Enforce and Prevent Consensus by Manipulating Communication Rules

We explore the possibilities of enforcing and preventing consensus in continuous opinion dynamics that result from modifications in the communication rules. We refer to the model of Weisbuch and Deffuant, where n agents adjust their continuous opinions as a result of random pairwise encounters whenever their opinions differ not more than a given bound of confidence e. A high e leads to consensus, while a lower e leads to a fragmentation into several opinion clusters. We drop the random encounter assumption and ask: How small may e be such that consensus is still possible with a certain communication plan for the entire group? Mathematical analysis shows that e may be significantly smaller than in the random pairwise case. On the other hand, we ask: How large may e be such that preventing consensus is still possible? In answering this question, we prove Fortunatos simulation result that consensus cannot be prevented for e > 0.5 for large groups. Next, we consider opinion dynamics under different individual strategies and examine their power to increase the chances of consensus. One result is that balancing agents increase chances of consensus, especially if the agents are cautious in adapting their opinions. However, curious agents increase chances of consensus only if those agents are not cautious in adapting their opinions.

Agents enacting social roles. balancing formal structure and practical rationality in MAS design

We introduce an integrated approach to the conceptualisation, implementation and evaluation of a MAS (multi-agent system) which is based on sociological concepts of practical roles and organisational coordination via negotiations. We propose a middle level of scale, located between interaction and the overall organisational structure, as the starting point for MAS design, with formal and practical modes of coordination to be distinguished over all relevant levels of scale. In our contribution, we present the modelling principles of our MAS, the agent architecture and the implementation. In the next step the approach is extended to a methodology for the investigation of processes of hybridisation, which means the re-entering of artificial sociality in a real-world domain. The integrated approach is intended to contribute to a generalised understanding of the Socionics program, which in our view should be seen as the enrolment of independent, but subsequent steps in an overall interdisciplinary approach.

C-IPS Approach to Negotiating Agents: Specifying Dynamic Interdependencies between Issue, Partner, and Step

Our C-IPS approach to negotiating agents considers the choice of negotiation issues, partners and steps as separate restricted decision processes. Here we focus on specification of dynamic interdependencies between these decisions and illustrate this by modeling sequential dependencies for an idealized scenario. The explicit modeling of all involved decisions and dependencies eases the understanding, evaluation, and comparison of different approaches to negotiating agents.

The C-IPS Agent Architecture for Modeling Negotiating Social Agents

The basic concepts in agent negotiation are negotiation object, negotiation protocol and reasoning process. While aiming to transfer sociological concepts into multi-agent systems, where agents coordinate themselves by negotiation, we recognized the need for a more detailed structure. Therefore we developed C-IPS, which is presented in this article. It distinguishes between external constraints and internal reasoning processes. The reasoning process covers decisions regarding the negotiation issue, the partner, and a particular step. Its modularization supports the development of agents with different degrees of autonomy. The components of the C-IPS architecture are modeled according to the BDI approach. As an example for the application of C-IPS we consider agents that are required for the INKA project.

C-IPS: Specifying Decision Interdependencies in Negotiations

Negotiation is an important mechanism of coordination in multiagent systems. Contrary to early conceptualizations of negotiating agents, we believe that decisions regarding the negotiation issue and the negotiation partner are equally important as the selection of negotiation steps. Our C-IPS approach considers these three aspects as separate decision processes. It requires an explicit specification of interdependencies between them. In this article we address the task of specifying the dynamic interdependencies by means of IPS dynamics. Thereby we introduce a new level of modeling negotiating agents that is above negotiation mechanism and protocol design. IPS dynamics are presented using state charts. We define some generally required states, predicates and actions. We illustrate the dynamics by a simple example. The example is first specified for an idealized scenario and is then extended to a more realistic model that captures some features of open multiagent systems. The well-structured reasoning process for negotiating agents enables more comprehensive and hence more flexible architectures. The explicit modeling of all involved decisions and dependencies eases the understanding, evaluation, and comparison of different approaches to negotiating agents.

Negotiating agents: from full autonomy to dynamic degrees of delegation

The modularization of negotiating agents as proposed by the C-IPS approach provides a sound base for a concept that we call dynamic degrees of delegation. Agents following this concept enable their users at runtime to delegate particular subsets of decisions to their artificial agents. We extend our specification and implementation of fully autonomous agents to capture this concept. We have successfully implemented such a system for application in a sociological experiment.