Evolution in Biological and Social Systems

Size: px
Start display at page:

Download "Evolution in Biological and Social Systems"

Transcription

1 Evolution in Biological and Social Systems 41. Wilhelm and Else Heraeus-Seminar January 2-24, 28 at the Physikzentrum, Bad Honnef Prof Peter M. Allen Complex Systems Research Centre, School of Management Cranfield University,

2 Physics and Evolution in Biological and Social Systems Physics is about the laws governing interacting matter and in many physical systems repeated experiments can be performed because they involve well defined, closed systems. In physics, discovering the patterns of behaviour and communicating this knowledge does not CHANGE the behaviour Evolution is about systems with changing TAXONOMY qualitative change and new variables, where communicating the patterns of behaviour observed can lead to responses that change that behaviour. In physics elements have stable internal structures, while people, actors and agents in social systems are actively engaged in up-dating their interpretive frameworks which they use to decide what to do and how to do it. Page 2

3 Emergence: forms, features, capabilities Page 3

4 From the Complex to the Simple: Complexity Reality Practice Soft Systems Not Science Heuristics Intuition Literature, History, Descriptions... Boundary Classification X Y Z Strategy Successive Assumptions Time Evolutionary Models CAS Structural Evolution Organizational Change New Variables Emergence, Innovation Learning Multi-Agent Models Creativity + Selection Structural Stability Fixed Variables X Y Probabilistic Non-linear Dynamics Dynamic, non-gaussian Probabilities, Master Equation, Multi-Agent Models Fixed Variables but Different spontaneous regimes Or configurations Contingency Z Simplicity Stationarity Equilibrium Average Dynamics Self-Organized Criticality Power laws, sand piles, Firm, income, city Sizes.. X Quantity Y Attractors Z Mechanical Non-Linear Dynamics Equilibrium Deterministic System Dynamics, Chaos,. Price Operations Page 4

5 So how does Evolution occur?: Freedom Constraints Reality Post Modernists Adaptive Evolutionary Model Darwinians Evolutionary: Ecology, Biology, Economics, Social Systems, Organisatons, et al Bottom-uppers Self-Organising Model With Noise X Y Mechanical Model Z Spontaneous changes of Regime, System Adaptation Assumptions Ecosystems evolve because some new types successfully invade The system can only be invaded by things to which it is unstable X Y We can calculate what can invade an ecosystem: Allen, 1976, Evolution, Population Dynamics and Stability, PNAS, Z Newtonians Chaos Theory, System Dynamics Prediction??? Economists Population Dynamics USA, Col 73, No 3, pp It tells us the effect of evolution of an ecosystem over time The Stability Matrix Page 5

6 Micro-diversity production - leads to coherent system diversity The maths of the evolutionary criterion of invadability (Allen, 1976) allow us to calculate the niche width degree of specialization that is best nν = Lc/(εσ 2 ) Robert May calculated that the species separation would depend on the resource fluctuations So, together we can calculate predict - the evolved diversity for a given resource spectrum. Galapagos Evolutionary Drive Allen & McGlade, 1987, Evolutionary Drive: The Effect of Microscopic Diversity, Error Making & Noise Darwin s Finches Page 6

7 The Probability of take-off : Freedom Constraints Reality Post Modernists Adaptive Evolutionary Model Darwinians Evolutionary: Ecology, Biology, Economics, Social Systems, Organisatons, et al Bottom-uppers Self-Organising Model With Noise If a new type arises then instead of being lethal or successful probabilistic analysis shows that: - good types do not necessarily survive -bad types survive for some time - Evolution is MESSY! X Y Z Spontaneous changes of Regime, System Adaptation Mechanical Model Assumptions Evolution is the cumulative effects of: -The probability of different new types - the probability of their survival X Y Z Newtonians Chaos Theory, System Dynamics Prediction??? Stochastic Population Dynamics Economists Allen & Ebeling, Biosystems, 16 (1983) Page 7

8 Hard Complexity Modelling: Paper Manufacturer Freedom Constraints Reality Post Modernists Adaptive Evolutionary Model Darwinians Evolutionary: Ecology, Biology, Economics, Social Systems, Organisatons, et al Bottom-uppers Self-Organising Model With Noise X Y Z Spontaneous changes of Regime, System Adaptation X Y Mechanical Model Z Newtonians Chaos Theory, System Dynamics Prediction??? Stochastic Model Fixed Taxonomy Economists Assumptions Distribution Centre Agents Customer Agents Unpredictable, Fluctuating Demand Factory Agent Central Warehouse Agent People inside the organization cannot learn! Page 8

9 The Model be run a million times to see what works: With data on past unpredictable demand we can find the current agent behaviour, and then try all kinds of variations By running the model many times we can explore how the different possible behaviours of the agents leads to the best performance. It reduces stock-outs completely, and requires less inventory. The total number of product changeovers is 8 as compared to the current 132 The time lost in changeovers is reduced to 22 days from 47 Interacting Agents form a Complex System Interacting agents cannot learn easily! Probably every business needs this kind of model. It cannot know how well it could function. Page 9

10 Schumpeter - Markets are evolutionary Systems: Average life of S&P firms has fallen from 65 years ( ) to 12years (2) In the last 55 years only 17 firms survived the period, but all but one had a return on investment less than the overall market gain It seems that companies either fail rapidly or if successful, fail later, creating an identity and locking into it. Usually, after some time as a successful company the innovative firms that are invading the market overtake them, and they either are acquired or simply cease operation Average Time in S&P list Foster and Kaplan, 21 The real task is to transform the company as fast as the market is evolving! Page 1

11 Evolution as Creative Destruction: Paul Ormerod modelled the life expectancy of firms under different hypotheses about their capacity to learn: He finds that the model that fits best is the one corresponding to random extinction and very little learning Past Instability The Market: Creative Destruction Extinction Present System Present Future Firms Time Most companies do not use complexity thinking they just try to make the most of opportunities and survive threats. Long term: Divergence Short term: Convergence Discontinuity Continuity What you in a market at a Given time is an accumulation Of things that have not yet died, Are mature and are still growing!

12 A Multi-Agent Economic Market Model: Revenue + Net Profit - Agent 1 Total Jobs Inputs+Wages Fixed Costs Production Staff COSTS PRICE PROFIT MARGIN FIRM 1 Sales Staff STOCK Production QUALITY Attributes of Supply SALES Potential Market Relative Attractivity OTHER FIRMS Interaction Attributes of Customer Demand Other Agents Customer Agents Customers with a product Type 1 Type 2 Type 3 POTENTIAL CONSUMERS Strategy on Profit Margin, Quality, R&D, Design. Page 12

13 What evolves in Price/Quality Strategy Space? Freedom Constraints Reality Post Modernists Adaptive Evolutionary Model Darwinians Evolutionary: Ecology, Biology, Economics, Social Systems, Organisatons, et al Bottom-uppers Self-Organising Model With Noise X Y Mechanical Model Z Spontaneous changes of Regime, System Adaptation 6 Darwinian Firms random re-launch Model 1 X Y Z Newtonians Chaos Theory, System Dynamics Prediction??? Evolution and Economists Assumptions Learning 6 Firms as before, but Firm 1 is an imitator. Model 2 All imitators... Model 4 6 Firms but 5 imitators. Model 3 Firm 1 can Learn Mixed Strategies All Firms learn Model 6 The Game Model 5 Model 7 Model 8 Page 13

14 uccess of MARKET with different Strategies and Luck Darwin 6 Seeds Average Final Vlaue = -114,161 St Dev = 1,134,154 No Learning Luck Seed = 3 Darwin Seed 1 Darwin Seed 2 Darwin seed 9 Darwin seed 6 Darwin seed 5 Darwin 3,, 2,5, 2,, 1,5, 1,, 5, All Imitate 6 Seeds Average Final Value = 847,954 St Dev = 1,255,568 Imitate the Winner , Luck seed 3 AllImitate seed 1 All Imitate seed 2 All Imitate seed 9 All Imitate seed 6 Allimitate seed 5 All Imitate -2-1,, All Learn Average Final Value = 1,348,471 St Dev = 99,531 All Learn by Experiment 3,5, 3,, Luck 35 3 Mixed Strategies Average Final Value = 1,61,629 St Dev = 1,142,622 Diverse Strategies Luck 2,5, 2,, 1,5, 1,, 5, Seed 3 All Learn Seed 1 All Learn Seed 2 All Learn Seed 9 All Learn Seed 6 All Learn Seed 5 All Learn Mixed 1 Mixed 2 Mixed 9 Mixed 6 Mixed 5 Mixed , Page 14

15 We can model LIMITED Creative Destruction: For given POTENTIAL market the market structure, apparent demand, scale and profits of an industry/market that actually emerges depend on LUCK The success of a firm depends on LUCK, and diverse strategies work better than similar ones. Learning (by experiment) is most consistently successful Learning by imitation is dangerous The future does not exist you are part of its creation Such models can be used to explore probable consequences of different strategies. Page 15

16 Soft Modelling: Using surveys of beliefs Interviews with supply chain managers led to 27 possible practices being seen as possible. The different dimensions of selection that operate on the supply chain are: Quality of fabrication Cost efficiency Reliability of Delivery Level of Technology and innovativeness Degree of Vision in the conception of a product Page 16

17 Aerospace Supply Chain Practices Characteristics Success criteria factors Rate of characteristic to successfactor criteria Product Cost Delivery Techn./ Vision for High (9), None ()) quality efficiency precision innovation the future 1. Outsourcing competitive advantage Outsourcing what is easily imitated High level of collaborative relationship Arms length relationships 5. Long-term relationship Formal partnership Subcontracting whole systems and sections Flexibility of operations Risk-sharing Sharing knowledge Offsets as part of sales contract Culture of continuous improvement Ability to handle cutural differences High level of dominance over supplier High level of planning and control Easy dialogue with supplier IT system integration High levels of integration of chain Responsive to market change Transparent organisation 21. TQM procedures Just-in-time delivery Lean practice Explorative learning practices Investment in training Supplier development Monitoring supplier Page 17

18 Pair Matrix Data: Characteristics Strongly synergetic (+5), indiffernt effects 1. Outsourcing competitive advantage (), strongly conflicting (-5) 2. Outsourcing what is easily imitated 2. Outsourcing what is easily imitated High level of collaborative relationship 3. High level of collaborative relationship Arms length relationship 4. Arms length relationship Long-term relationship 5. Long-term relationship Formal partnership 6. Formal partnership Subcontracting whole systems and sections 7. Subcontracting whole systems and sections Flexibility of operations 8. Flexibility of operations Risk-sharing 9. Risk-sharing Sharing knowledge 1. Sharing knowledge Offsets as part of sales contract 11. Offsets as part of sales contract Culture of continuous improvement 12. Culture of continuous improvement Ability to handle cultural differences 13. Ability to handle cultural differences High level of dominance over supplier 14. High level of dominance over supplier High level of planning and control 15. High level of planning and control Easy dialogue with supplier 16. Easy dialogue with supplier IT system intergration 17. IT system integration High levels of integration in chain 18. High levels of integration in chain Responsive to market change 19. Responsive to market change Transparent organisation 2. Transparent organisation TQM procedures 21. TQM procedures Just-in-time delivery 22. Just-in-time delivery Lean practices 23. Lean practices Explorative learning practices 24. Explorative learning practices Investment in training 25. Investment in training Supplier develop 26. Supplier development Monitoring suppliers Page 18

19 The Pair Interactions between the 27 Practices: The (27x27 pair matrix)x(transpose) = Map of total synergy/conflicts Synergy Conflict Airframe L S27 S25 S23 S21 S19 S17 S15 S13 S11 S9 S S5 S S1 Page 19

20 The Real Performance of a bundle of Practices: The direct effects of practices on different dimensions of output performance are given in first table The effects of one practice on another are given by the 27x27 pair matrix The real output of a weighted sum of the different dimensions of performance are given by: (27x27 Pair matrix)x(27x1 Column) = (27x1)Real Output This calculates the real, observed output performances for any combination of practices Page 2

21 Practices that Maximise weighted sum of Outputs Quality Cost Trade-Off Weights 1,5,1,1,1 Delivery Technology Vision Real!st Order Quality Practices Cost Practices Quality Cost Vision Technology Delivery Practices Vision Practices Technology Practices Delivery Page 21

22 Evolutionary Learning Model of Adding Practices: Different Sequences of Quality Innovation Series1 Series2 Series3 Series4 Series Practices and Quality Performance 4 After t = 4, Performance = Performance Practices Added Page 22

23 Outputs from the Learning Model: Performance in all criteria Practices and Innovative Performance 4 After t = 6, Performance = Quality Cost Efficiency Series3 Technology & Innovation Vision Practice Number Time Performance Practices Added Practices and Cost Performance 4 After t = 6, Performance = 6.21 Practices and Reliable Delivery Performance 4 After t = 6, Performance = Performance Practices Added Practice Number Performance Practices Added Time Page 23

24 Sustaining Business Performance: In a competitive market you cannot afford to relearn everything that others know But, you cannot afford to simply limit your business to what you do now This therefore shows that you need to accept what is known, but continue to do experiments at the edge: New Materials New Technologies New Opportunities New Practices Page 24

25 Evolutionary Complexity - Automobile Production: Organizational Cladistics: McKelvey, Ridgway, McCarthy, Allen, Strathern, Baldwin Standardization of parts 19 Automation (machine paced shops) 37 Job enrichment 2 Assembly Time Standards 2 Multiple sub-contracting 38 Manufacturing Cells 3 Assembly Line Layout 21 Quality Systems 39 Concurrent engineering 4 Reduction in Craft Skills 22 Quality Philosophy 4 ABC Costing 5 Automation 23 Open Book Policy with Suppliers 41 Excess capacity 6 Pull Production System 24 Flexible Multi-functional workforce 42 Flexible Automation of product versions 7 Reduction of Lot Size 25 Set-up time reduction 43 Agile automation for different products 8 Pull Procurement System 26 Kaizen change management 44 In-Sourcing 9 Operator based machine maintenance 27 TQM sourcing 45 Immigrant workforce 1 Quality Circles 28 1% inspection sampling 46 Dedicated automation 11 Employee innovation Prizes 29 U-shaped layout 47 Division of Labour 12 Job Rotation 3 Preventive Maintenance 48 Employees are system tools 13 Large Volume Production 31 Individual Error correction 49 Employees are system developers 14 Mass Sub-Contracting by sub-bidding 32 Sequential dependency of workers 5 Product focus 15 exchange of workers with suppliers 33 Line Balancing 51 Parallel processing 16 Training through socialization 34 Team Policy 52 Dependence on written rules 17 Pro-active Training Programme 35 Toyota Verification Scheme 53 Further intensification of labour 18 Product Range reduction 36 Groups vs Teams Different Practices Identified: Automobile Production Evolutionary Tree Organizational Forms 1. Ancient Craft System 2. Standardised craft System 3. Modern craft System 4. Neocraft systems 5. Flexible Manufacturing 6. Toyota production 7. Lean producers 8. Agile producers 9. Just in time 1. Intensive Mass producers 11. European mass producers 12. Modern Mass Producers 13. Pseudo lean producers 14. Fordist Mass producers 15. Large Scale producers 16. Skilled large Scale producers 23 Organizations are different synergetic BUNDLES of these Page 25

26 Complexity Model of Organizational Evolution: Can we use these ideas to explore organizational change? Performance Can run multiple Simulations of firms: Patterns of potential Synergy or conflict. Manufacturers survey, Baldwin, 24. Extinctions Time Japanese Production Fordism 53x53 Matrix of synergy/conflict between practices Synergy per individual 17 conflicting factors Page 26

27 Internal Instability - beliefs are ambiguous: Our interpretive framework results from our experiences which are guided by our interpretive framework! Continue Modify Beliefs Actions, Experiments (Noisy, Probabilistic) Decision, Choice (not unique) Values given By beliefs Beliefs Knowledge Aims, Goals Values Individual World Modify, Update (not unique) Ideas confirmed Expectation Deny/Confirm Free Will? Page 27

28 rom Physics to Evolution in Biology and Social Science? In physics with elements of fixed taxonomy the invariance of the laws of nature allow prediction. Even after the news has spread, the laws do not change. In Biological and Social Science, what we see is an intricate accretion of earlier events at different levels in the system. Most have failed and disappeared. It is difficult to know (fully) why something is working or why it may fail. Our models are attempts to propose simple explanations that may help us to explore possible futures. Models do not make predictions. They are our interpretive frameworks. They may warn us about possible futures (e.g. climate change, limits to growth). The more credible predictions are, the more likely they are to NOT happen. The Physics of Closed Systems has fixed taxonomy but for Open Systems can have changing taxonomies and become Evolutionary. Page 28