FINAL PAPERMASSACHUSETTS INSTITUTE OF TECHNOLOGYSYSTEMS ARCHITECTURES:WHAT CAN PRODUCT ARCHITECTURE ANALYSISLEARN FROM COMPLEX ADAPTIVE SYSTEMS?December 15, 2000Sebastian Fixson, [email protected], Management and Policy Program, MITCourse: ESD.83 Research Seminar in Engineering SystemsProf. David A. MindellSystems Architectures: What can product architecture analysis learn from complex adaptive systems?“Complexity is how much more the whole is than just the sum of its parts.”(Steward 1981, p.2)Systems Architectures: What can product architecture analysis learn from complex adaptive systems?Table of Contents1. INTRODUCTION___________________________________________________________ 12. WHEN IS PRODUCT A SYSTEM? ______________________________________________ 23. BRIEF OVERVIEWS OF TWO CONCEPTS_______________________________________ 33.1 Product Architecture ___________________________________________________ 43.2 Complex Adaptive Systems ______________________________________________ 54. ANALYSIS FRAMING: NEW INSIGHTS FOR PRODUCT ARCHITECTURE ANALYSIS_______ 74.1 System Boundary: Extensions in Size and Time _____________________________ 84.1.1 Base case: Elements and relations remain constant _________________________ 94.1.2 Extending system size: Elements change, relations remain constant ___________ 124.1.3 Extending system time: Elements remain constant, relations change __________ 164.1.4 Extending system size and time: Elements and relations change______________ 194.2 Meaning of Adaptation_________________________________________________ 204.2.1 Create perspective__________________________________________________ 214.2.2 Make perspective __________________________________________________ 234.2.3 Use perspective____________________________________________________ 244.2.4 Retire perspective __________________________________________________ 265. CONCLUSION AND FUTURE RESEARCH_______________________________________ 28Systems Architectures: What can product architecture analysis learn from complex adaptive systems?- 1 -1. INTRODUCTIONUsually, products are seen as artifacts, produced and consumed by humans, but dead andinanimated by themselves. This is particularly true for assembled products, hardware products.The structures of these products and their implications, however, have been attractingresearchers’ attention for quite some time. In recent years, a number of influential contributionshave been made to this area of product architecture. By decomposing existing productarchitectures, it focuses mainly on structural differences and how these relate to functionalitydistributions.In contrast, an entirely different field of research has developed based on the finding that thereare systems that can be only be understood through their behavior. Their structural appearancedoes not allow conclusions on the systems’ behavior and emergence. Researchers in this fieldhave developed computer models with which they try to mimic these complex adaptive systems.Characteristic feature of these models is that they are built by aggregating the (oftenstereotypical) behavior of relatively simple agents.It seems that the work on complex adaptive systems allows a number of insights impossible forthe view restricted on structural descriptions. This paper attempts to show how basic features ofthe agents and the concept of adaptation, both taken from of the complex adaptive systems, canbe used as tools to improve the understanding of how to interpret product architectures and theirimplications. Although the theory of complex adaptive systems describes systems as diverse asnervous systems or cities, this paper focuses on man-made systems.The remainder of the paper is organized as follows. Chapter 2 relates products to a number ofsystems definitions. Next, the third chapter provides brief overviews on the two concepts underSystems Architectures: What can product architecture analysis learn from complex adaptive systems?- 2 -consideration. Chapter 4 forms the core of the paper discusses additional insights for productarchitectures around the themes boundary definition and adaptation. Finally, chapter 5 concludesand offers ideas for further research.2. WHEN IS A PRODUCT A SYSTEM?Although it seems as if the use of the term system has been increasing in recent times, the worditself has always been served as a description for something that seems to exist in different fieldsas diverse as human anatomy (e.g. nervous system), man made artifacts (e.g. transportationsystem), or astronomy (solar system).iIn its everyday-use, systems usually are associated with some level of complexity. The questionof what complexity actually is, is quite complicated, as various attempts to measure complexityshow, ranging from the amount of information to describe a system to the total amount ofthermodynamic and informational resources required by the physical construction process (Lloyd1990). For the purpose of this paper, I will start of with rather simple systems and increase theirlevels of intricacy stepwise.On the most generic level, a system has to provide at the minimum two distinct features. First,for anything to be a system it has to consist of more than one element, for however broad thedefinition for a system might be, once it is a single entity the meaning of system is lost. Second,the several elements together display characteristics the single elements cannot. Just placingcoffee mugs next to each other, does not make it a system of coffee mugs. This points to aninherent system characteristic that emerges through the existence of interfaces.iiSystems Architectures: What can product architecture analysis learn from complex adaptive systems?- 3 -Both of these features provide direction on what to look for when one wants to describe asystem. Two groups of questions emerge from thinking in these directions. One group isconcerned with the structure of a system, the other with its behavior. Neither group is perfectlyaligned with either the term static or the term dynamic, although some overlap can be found.When can we consider products as systems, given these considerations concerning theirelements, their interfaces and their dynamics? Most products consist of more than one elements,so at least this basic hurdle they would clear. In fact, products can consist of hundreds orthousands of parts. High number of components almost automatically increase also the numberof interfaces, although