767 Copyright © 2010, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited. Chapter 2.22 Building Complex Adaptive Systems: On Engineering Self-Organizing Multi-Agent Systems Jan Sudeikat 1 Hamburg University of Applied Sciences, Germany Wolfgang Renz Hamburg University of Applied Sciences, Germany AbstrAct Agent oriented software engineering (AOSE) proposes the design of distributed software sys- tems as collections of autonomous and pro-active actors, so-called agents. Since software applica- tions results from agent interplay in multi-agent systems (MASs), this design approach facilitates the construction of software applications that exhibit self-organizing and emergent dynamics. In this chapter, we examine the relation between self-organizing MASs (SO-MASs) and complex adaptive systems (CASs), highlighting the result- ing challenges for engineering approaches. We argue that AOSE developers need to be aware of the possible causes of complex system dynamics, which result from underlying feedback loops. In this respect current approaches to develop SO- MASs are analyzed, leading to a novel classifca- tion scheme of typically applied computational techniques. To relieve development efforts and bridge the gap between top-down engineering and bottom-up emerging phenomena, we discuss how multi-level analysis, so-called mesoscopic modeling, can be used to comprehend MAS dynamics and guide agent design, respectively iterative redesign. INtrODUctION AOSE (Weiß, 2002) is a prominent approach to the development of complicated distributed software systems. Agents, that is, autonomous and pro- active entities, are proposed as a basic design and development metaphor. Since highly dynamic and