A Complexity Primer for Systems Engineers
November 2015
Dr. Jimmie McEver Senior Systems Scien st, JHU Applied Physics Laboratory Chair, INCOSE Complex Systems Working Group [email protected]
1 Agenda
• Brief overview of Complex Systems WG
• Complexity Primer for Systems Engineers
2 Complex Systems (CxS) WG Overview
• Organized in 2006 from Systems Science WG • Co-chairs: Dr. Jimmie McEver, JHU/APL Mr. Michael Watson, NASA/MSFC • 46 INCOSE members par cipated in discussions at IW 2015
https://connect.incose.org/WorkingGroups/ComplexSystems/
3 Purpose and Objec ves
• The purpose of the Complex Systems Working Group is to enhance the ability of the systems engineering community to deal with complexity. • The CxS Working Group works at the intersec on of complex systems sciences and SE, focusing on systems beyond those for which tradi onal systems engineering approaches and methods were developed. • Complex Systems Working Group objec ves – Communicate the complexity characteris cs to systems engineering prac oners – Provide knowledge and exper se on complex systems in support of other INCOSE working groups working in their systems engineering areas – Facilitate the iden fica on of tools and techniques to apply in the engineering of complex systems – Provide a map of the current, diverse literature on complex systems to those interested in gaining an understanding of complexity. • Although analysis is important, the goal is to make a difference in synthesis (crea on of new systems).
4 Candidate Ac vi es
• Expand on Mat French MBSE in Complexity Contexts work begun in 2014 • Expand on selected Primer topics; evolve into wiki to facilitate evolu on and access; develop annotated complexity bibliography • Work with US Government partners to iden fy WG ini a ves to help address current systems challenges (digital engineering, engineering resilient systems, open architecture based approaches for engineering complex systems) • Iden fy rela onships and strengthen interac ons with other INCOSE WGs (and with complexity groups in other organiza ons) • Workshop or other joint mee ng with AIAA Complex Aerospace Systems Exchange organizers • Maturity Model for the Complex SE capability of an organiza on or endeavor • Organize webinars to discuss WG ini a ves and topics of interest
5 A Primer on Complexity for Systems Engineers
• The INCOSE Complex Systems Working Group has dra ed a primer to introduce complexity concepts and approaches to prac cing systems engineers • Members of the primer development team – Sarah Sheard – Joseph Krupa – Eric Honour – Paul Ondrus – Jimmie McEver – Robert Scheurer – Dorothy McKinney – Janet Singer – Alex Ryan – Joshua Sparber – Stephen Cook – Brian White – Duane Hybertson
6 Mo va on: The Implica ons of Complexity
• Systems engineering has evolved to improve our ability to deal with scale and interdependency though the life cycle of engineered systems • Systems are becoming increasingly dynamic and interdependent, with growing emphasis on adap veness
Source: Sarah Sheard, Complexity and Systems Engineering, 2011 Source: Monica Farah-Stapleton, IEEE SOS conference, 2006 • BUT, complexity places new demands on systems engineers that require more than extensions of “classical” SE prac ce
7 Classes of Systems Problems: Kurtz and Snowden’s Cynefin Framework
Cynefin domains Complex systems, Massively- characterized by complicated systems interdependence, present challenges of self-organization, and scale and interface emergence – new accounting – tools and approaches extensions of are needed traditional methods/ tools can help
Chaotic systems Simple and can only be complicated systems reacted to; can are straightforward to attempt to deal with using transform into traditional systems another domain engineering / mgmt approaches
Source: Kurtz and Snowden, “The new dynamics of strategy: Sensemaking in a complex and complicated world,” IBM Systems Journal, 42 (3), 2003.
8 What do we mean by complexity?
• No easy, agreed-upon defini on • We o en call something complex when we cannot fully understand its structure or behavior: it is uncertain, unpredictable, complicated, or just plain difficult (see Silli o (2009): Subjec ve Complexity) • Concepts that seem essen al – Emergence: Features/behavior associated with the holis c system that are more than aggrega ons of component proper es – Mul -scale behavior: System not describable by a single rule, structure exists on many scales, characteris cs are not reducible to only one level of descrip on. – A system with self-organiza on, analogous to natural systems, that grows without explicit control, and is driven by mul ple locally opera ng, socio-technical processes, usually involving adapta on
9 Complex vs Complicated
Complex system Complicated system
Traffic jams exhibit: Modern transit vehicles exhibit: • Self-organization and Emergence – • Large numbers of interacting systems – local actors and decisions interact to fuel, electrical, engine, transmission, create larger patterns safety, etc. • Memory – even after an obstacle is • Aggregate properties, but not emergence removed the jam can persist for hours – e.g., range, top speed… • Counter-intuitive outcomes – e.g., • Unexpected outcomes still possible, Braess’ Paradox – adding capacity to the but origin is different network can degrade network • Transit systems may be complex performance The opposite of “complex” is “decomposable”, not “simple”
Source: Alex Ryan 10 Hallmarks and Implica ons of Complexity How complexity makes it different
Hallmarks of complexity Impact on Decision Maker Interdependence Cannot treat by decomposition Nonlinearities Extrapolation of current conditions à error Open boundaries Cannot focus only on processes inside boundary Multi-scalarity Have to address all relevant scales Causal & influence networks Challenge: develop ‘requisite’ conceptual model within time and information resource constraints Emergence Unknown risks and unrecognised opportunities Complex goals Goals may change, be unrealistic, vague Adaptation & innovation ‘Rules’ change, interventions stimulate adaptation Opaqueness Many possible hypotheses about causal paths, insufficient evidence to discriminate
Adapted from Grisogono, Anne-Marie and Vanja Radenovic, The Adaptive Stance –Steps towards Teaching more Effective Complex Decision-Making, International Conference on Complex Systems, June 2011.
11 Sources of Complexity
• System – Large number of components, many intricate interdependencies – Adap ve components, interac ons – Interfaces with human users or other complex en es – Evolving technology • Environment – Opera onal environment • Problem you’re trying to solve changes • Condi ons under which you’re trying to solve a problem change • Way that you elect to solve the problem changes – System environment • Changes to elements of the larger SoS • Design/management – Large number of people and organiza ons involved
12 Sarah Sheard, Complexity and Systems Engineering, 2011. Used with permission System Complexity as SE Challenges
System Characteristics Cognitive Characteristics (Objective Complexity) (Subjective Complexity)
Many pieces Uncertain Emergent Difficult to understand Nonlinear Unclear cause and effect Chaotic Adaptive Unpredictable Tightly coupled Complexity Unstable Self-Organized Uncontrollable
Decentralized Unrepairable, unmaintainable
Political Open Takes too long to build
Multi-Scale Costly
13 Complexity and Systems Development
• You need complexity to respond to Ashby’s Law of complexity Requisite Variety – Provides degrees of freedom to needed to deal with/work in complex, volatile and uncertain environments • But developing and using complex systems have their own challenges – Less ability to plan and predict – Problems/systems not neatly decomposable – interdependencies A model system or controller between subsystems grow beyond ability can only model or control to deal with them in traditional ways something to the extent that it has sufficient internal – Uncomfortable phenomena arise, such variety to represent it. as normal accidents, black swans, catastrophic fragility
14 Candidate approaches: Selected Guiding Principles
v
Think like a gardener, Take an not a watchmaker adap ve stance • Grow, don’t build • Enable and improve • Focus on the ecosystem adapta on capabili es – Extensible substrate – Observe system behavior – Rules and feedback for – ID and create varia on adapta on – Selec ng the best versions • Influence and intervene – Amplify the fit of the vice design and control selected versions
15 Candidate approaches: Selected Guiding Principles
Think at mul ple levels and from mul ple perspec ves • System behavior may be • Systems may create different fundamentally different at value at different levels and from different levels different perspec ves • The decomposed problem is a • Need to look at systems and different problem requirement from diverse perspec ves
16 Candidate approaches: Selected Guiding Principles
• Combine courage with • Focus on desired regions of humility the outcome space rather • Use free order than specifying detailed • Iden fy and use pa erns outcomes • See through new eyes • Understand what mo vates autonomous
• Collaborate agents
• Achieve balance • Maintain adap ve • Learn from problems feedback loops • Meta-cogni on • Integrate problems
17 Candidate approaches: SE Methods Environmental Complexity
Requirements Solution Elicitation and Architecture and Development Derivation Trade Studies Design Process Environment Use power laws Make Design for Resilience susceptible to to characterize dimensions resilience to analysis unpredictable relevant of agility key “beyond- but high- phenomena trade-space design- Enterprise consequence attributes envelope” development: events and/or Focus elicitation events to study how recursive on agility vice Use trades to provide enterprises or complexity optimizing to ID aspects of robustness societies particular the problem and timely survive assumptions space that recovery to a catastrophes. will drive the minimally system functional architecture state.
18 Candidate approaches: SE Methods System/Solu on Complexity
Requirements Solution Elicitation and Architecture and Development Derivation Trade Studies Design Process Complexity in Use multi-scale modeling Emphasize Use Systems- System (linking macro- and micro- selection of of-Systems Design & level models), including robust and methodologies Development exploratory analysis and adaptive to synchronize (General) agent-based modeling, and elements and constituent experimentation: structures over systems; • To generate insight into the optimizing to incentivize implications of derived meet current collaboration. requirements requirements • As the basis for trade Ensure studies and to inform trade- prototyping and off decisions experimentation are used.
19 Applicability of Generalized Analy c Methods: The Cook Matrix
• Lists and briefly describes modeling and analysis methods from a wide range of disciplines • Candidate approaches to consider in modeling complex systems problems
Analyze Diagnose Model Synthesize Data Mining Algorithmic Complexity Uncertainty Modeling Design Structure Matrix Splines Monte Carlo Methods Virtual Immersive Modeling Architectural Frameworks Fuzzy Logic Thermodynamic Depth Func onal / Behavioral Models Simulated Annealing Neural Networks Fractal Dimension Feedback Control Models Ar ficial Immune System Classifica on & Regression Trees Informa on Theory Dissipa ve Systems Par cle Swarm Op miza on Kernel Machines Sta s cal Complexity Game Theory Gene c Algorithms Nonlinear Time Series Analysis Graph Theory Cellular Automata Mul -Agent Systems Markov Chains Func onal Informa on System Dynamics Adap ve Networks Power Law Sta s cs Mul -scale Complexity Dynamical Systems Social Network Analysis Network Models Agent Based Models Mul -Scale Models
20 Candidate approaches: Analy c tools for the Complex SE Toolkit? • Some are emerging for “massively complicated” SE – MBSE tools and methods – Emerging thinking on T&E in large test spaces • For complex SE, some exist – but are not designed for or aligned with SE community • Complex SE (or CxSE) requires more than just tools – Not a “run the tool, get the answer” problem – Requires understanding of the nature of complexity and how to deal with vola lity and deep uncertainty • Mindset, experience (and educa on) – Tools needed are those that will enable this understanding • A key role of the systems engineer is to facilitate this understanding across his/her stakeholder communi es
21 Next Steps (for the Primer and the WG)
• Present Primer for INCOSE • Leverage primer (and other) technical review approaches to address (or • Con nue to socialize primer at least generate new as a resource for INCOSE insight into) community members “hard problems” • Expand on selected Primer – Cybersecurity topics; evolve into wiki to – Evolu onary Acquisi on facilitate evolu on and – Systems of systems access; develop annotated • Engage with other working complexity bibliography groups to iden fy – Construct Concept Map for intersec on topics we can complex systems knowledge work together
22