Innovation, Behavior, Emergence & Market
Module 1: Theory of Complex Adaptive Systems
Lecture 1: It’s a jungle out there
System: a set of connected things
Type of systems
1. Simple (bike)
- Well-ordered, predictable cause effect
- “Things” are simple and few
- Relations are simple and stable
- Easily repairable
- Input-output relations are simple
- Structure and functions are clear
- “Things” have no behavior: they don’t change their mind
2. Complicated (airplane, phone)
- “Things are many and can be complex
- Relations are many fold and diverse
- Difficult to design and repair (experts)
- Structures and function are partly hidden; they have their own logic
- Engineered
- Things don’t change their mind
3. Nonlinear (atmosphere: the weather)
- Continuously changing
- Unpredictable
- Many things, but no thinking or adaption
- Input-output relations unclear
- Butterfly effect: a small change may cause a large effect
- Difficult to control and change
4. CAS (from bacteria to Dutch health care)
à Very general conceptual model of a system, it should work for all kind of systems
- Many things (actors, agents); are connected into a network
- Adaptive (at the thing level and at the system level)
- Details are unpredictable, but general laws exist
- Constant change: no fixed equilibrium à multiple equilibria and changing patterns
- Nonlinear
- Butterfly effect
- Resilience (system is resistant against disturbances)
- Evolution, specialization
- Not a single big boss, but distributed control
- System has boundaries, but they are permeable
Examples: Ecosystems, City, Organizations (hospital), Market (business ecosystem), Artificial Systems
,Visible general CAS properties
- Diversity / specialization of actors
à in types of actors (teachers, students etc), types of courses, actors within working groups
- Actors change behavior: genes and learning
- Flows (food chains, info, water, cars)
à money, information, knowledge
- Aggregation
à courses, working groups (self-organization), but also institutes and even the faculties
- Cooperation
à in and between courses, working groups, assignments
- Building blocks: things that are successful can be copied, combined and re-used:
à course slides, exam questions, parts of a report or a presentation
- Boundaries, but they are permeable
à working groups work separately on their assignment, make courses a separate unit
- Adaptation and behavior change (learning)
à align with the rules of the uni, a course or working group; bad grades; feedback
- Tags: a visible code to easily identify an actor (useful for other actors)
à a title like Dr. of Professor in mails, on the door of their rooms
- Struggle and survival / competition between actors
- Reward mechanisms: determine outcomes / behavior
à passing for a course, graduation and the wage of teachers and coaches
- Internal model
à the MPA program and courses are an big attempt to enrich our internal models
- Self-organization
à when you can form your own working groups and divide tasks
- Inequality
à in teacher and student capacities, differences in position and rewards, status of people
Adaptation + Rewards result in:
- Selection: failure of the weak and success for the fittest
- Inequality: a CAS is unfair (some are rich, most are poor)
Example Hospital:
- Specialization of actors
- Tags: white coats and badge
- Competition: at certain levels of profession
- Reward mechanisms: professional recognition and money (doctors), health (patients)
Invisible general CAS properties
- A CAS can have several equilibrium points
- It can switch between these forms by passing through a transition point
- Perturbations (big or small critical events) may cause a jump to a new equilibrium point
- In a CAS, cause-effect relations are non-linear: you can’t calculate the effect of a change,
even if you know everything about the individual actors
- CAS are resistant to change (resilient)
- CAS are usually in a stable form: small changes don’t disturb the system, the system adapts
and stays close to the equilibrium
- At a certain level of perturbation, the system can jump to another stable situation
- CAS tries to understand:
o What makes a system stable?
o How to predict what is the critical point?
o How to change a system in the right way?
,Nonlinearity
Because of the complexity of a CAS and the emergent nature of phenomena
à There is no simple relation between a change and the reaction of the system
à No simple cause-effect relation
A small change may cause:
- No effect (stability)
- Unexpected effect
- Large effect: across a transition point
A large change may cause:
- No effect (resilience: stability, adaptation)
- Unexpected effect (emergence)
- Large effect: across a transition point
Why study a CAS?
- To analyze how the system works as it does
- To understand what happens
- To find similarities and general laws
- To predict and forecast what will happen in a system
- To improve it (smart interventions – WHO uses the CAS approach)
Why and how study CAS?
- Similarities: city suburbs, brain and broccoli all grow in the same way (why?)
- Find the hidden structure in social networks (using FB data to fight alcoholism)
- Diffusion of viruses, ideas and innovations is determined by the (social) structure of a CAS
- Critical points: instability and system shift
- Understand how coalitions form … and implode
- Simulate event (crowd behavior)
- Use adaptation / reward for computer programming
Debriefing Niche Game
Why is it realistic?
- There is no time to form coalitions or make a deal with your competitors
- It is difficult to find a strategy that works because you don’t know the strategies of the others
- There is no second round, just like in the business world
The main problem
You don’t have insight in what your competitors will do. Are they risk taking, or avoid it?
, Lecture 2: Hidden order
Internal models
à An actor’s model of its environment in a form that describes how to behave
An internal model can change
- By coincidence (mutation)
- By design (programming)
- Purposeful (learning from experiences)
To be adaptive, you need something that remembers what you did and how it worked (John Holland)
You = [generalized] an actor in the CAS
An internal model is the carrier of adaptivity
- Forms of internal models
- Why internal models are useful
- What is the difference between survival and internal models
How to store?
- Brain (neurons) à explicit knowledge / experience, cultural norms and instinct
- DNA (coding for proteins)
- Text (recipe, business plan, Bible, surgery protocol)
- Software / algorithm (artificial intelligence)
à can be copied and modified (experience and learning)
Structure of internal models: “rules”
How to act (medical protocol)
IF (Symptom 1) THEN (treatment 1)
IF (S1 and S2) THEN (T2)
IF (S1 and S@ and age <5) THEN (T1+T3)
IF (S2 and low Hb) THEN (T2 + T4)
Internal models are nested model (general rules – specific rules – exceptions)
A rich internal model allows more precise decisions and more diverse reactions, but you need more
memory space
Learning =
- Changing an existing rule
- Adding and modifying exception rules
- Add exceptions and exceptions to exceptions
Summary
- Internal model is the basis for learning
- It makes an actor better equipped for survival
- Actors with a failing (too simple/wrong rules) model have a competitive disadvantage
- A better model will survive and spread (through it’s owner)
- A rich model helps it owner to do better
- In a CAS, we always see working internal models, not perfect internal models
- We see a snapshot of internal models in various stages
Module 1: Theory of Complex Adaptive Systems
Lecture 1: It’s a jungle out there
System: a set of connected things
Type of systems
1. Simple (bike)
- Well-ordered, predictable cause effect
- “Things” are simple and few
- Relations are simple and stable
- Easily repairable
- Input-output relations are simple
- Structure and functions are clear
- “Things” have no behavior: they don’t change their mind
2. Complicated (airplane, phone)
- “Things are many and can be complex
- Relations are many fold and diverse
- Difficult to design and repair (experts)
- Structures and function are partly hidden; they have their own logic
- Engineered
- Things don’t change their mind
3. Nonlinear (atmosphere: the weather)
- Continuously changing
- Unpredictable
- Many things, but no thinking or adaption
- Input-output relations unclear
- Butterfly effect: a small change may cause a large effect
- Difficult to control and change
4. CAS (from bacteria to Dutch health care)
à Very general conceptual model of a system, it should work for all kind of systems
- Many things (actors, agents); are connected into a network
- Adaptive (at the thing level and at the system level)
- Details are unpredictable, but general laws exist
- Constant change: no fixed equilibrium à multiple equilibria and changing patterns
- Nonlinear
- Butterfly effect
- Resilience (system is resistant against disturbances)
- Evolution, specialization
- Not a single big boss, but distributed control
- System has boundaries, but they are permeable
Examples: Ecosystems, City, Organizations (hospital), Market (business ecosystem), Artificial Systems
,Visible general CAS properties
- Diversity / specialization of actors
à in types of actors (teachers, students etc), types of courses, actors within working groups
- Actors change behavior: genes and learning
- Flows (food chains, info, water, cars)
à money, information, knowledge
- Aggregation
à courses, working groups (self-organization), but also institutes and even the faculties
- Cooperation
à in and between courses, working groups, assignments
- Building blocks: things that are successful can be copied, combined and re-used:
à course slides, exam questions, parts of a report or a presentation
- Boundaries, but they are permeable
à working groups work separately on their assignment, make courses a separate unit
- Adaptation and behavior change (learning)
à align with the rules of the uni, a course or working group; bad grades; feedback
- Tags: a visible code to easily identify an actor (useful for other actors)
à a title like Dr. of Professor in mails, on the door of their rooms
- Struggle and survival / competition between actors
- Reward mechanisms: determine outcomes / behavior
à passing for a course, graduation and the wage of teachers and coaches
- Internal model
à the MPA program and courses are an big attempt to enrich our internal models
- Self-organization
à when you can form your own working groups and divide tasks
- Inequality
à in teacher and student capacities, differences in position and rewards, status of people
Adaptation + Rewards result in:
- Selection: failure of the weak and success for the fittest
- Inequality: a CAS is unfair (some are rich, most are poor)
Example Hospital:
- Specialization of actors
- Tags: white coats and badge
- Competition: at certain levels of profession
- Reward mechanisms: professional recognition and money (doctors), health (patients)
Invisible general CAS properties
- A CAS can have several equilibrium points
- It can switch between these forms by passing through a transition point
- Perturbations (big or small critical events) may cause a jump to a new equilibrium point
- In a CAS, cause-effect relations are non-linear: you can’t calculate the effect of a change,
even if you know everything about the individual actors
- CAS are resistant to change (resilient)
- CAS are usually in a stable form: small changes don’t disturb the system, the system adapts
and stays close to the equilibrium
- At a certain level of perturbation, the system can jump to another stable situation
- CAS tries to understand:
o What makes a system stable?
o How to predict what is the critical point?
o How to change a system in the right way?
,Nonlinearity
Because of the complexity of a CAS and the emergent nature of phenomena
à There is no simple relation between a change and the reaction of the system
à No simple cause-effect relation
A small change may cause:
- No effect (stability)
- Unexpected effect
- Large effect: across a transition point
A large change may cause:
- No effect (resilience: stability, adaptation)
- Unexpected effect (emergence)
- Large effect: across a transition point
Why study a CAS?
- To analyze how the system works as it does
- To understand what happens
- To find similarities and general laws
- To predict and forecast what will happen in a system
- To improve it (smart interventions – WHO uses the CAS approach)
Why and how study CAS?
- Similarities: city suburbs, brain and broccoli all grow in the same way (why?)
- Find the hidden structure in social networks (using FB data to fight alcoholism)
- Diffusion of viruses, ideas and innovations is determined by the (social) structure of a CAS
- Critical points: instability and system shift
- Understand how coalitions form … and implode
- Simulate event (crowd behavior)
- Use adaptation / reward for computer programming
Debriefing Niche Game
Why is it realistic?
- There is no time to form coalitions or make a deal with your competitors
- It is difficult to find a strategy that works because you don’t know the strategies of the others
- There is no second round, just like in the business world
The main problem
You don’t have insight in what your competitors will do. Are they risk taking, or avoid it?
, Lecture 2: Hidden order
Internal models
à An actor’s model of its environment in a form that describes how to behave
An internal model can change
- By coincidence (mutation)
- By design (programming)
- Purposeful (learning from experiences)
To be adaptive, you need something that remembers what you did and how it worked (John Holland)
You = [generalized] an actor in the CAS
An internal model is the carrier of adaptivity
- Forms of internal models
- Why internal models are useful
- What is the difference between survival and internal models
How to store?
- Brain (neurons) à explicit knowledge / experience, cultural norms and instinct
- DNA (coding for proteins)
- Text (recipe, business plan, Bible, surgery protocol)
- Software / algorithm (artificial intelligence)
à can be copied and modified (experience and learning)
Structure of internal models: “rules”
How to act (medical protocol)
IF (Symptom 1) THEN (treatment 1)
IF (S1 and S2) THEN (T2)
IF (S1 and S@ and age <5) THEN (T1+T3)
IF (S2 and low Hb) THEN (T2 + T4)
Internal models are nested model (general rules – specific rules – exceptions)
A rich internal model allows more precise decisions and more diverse reactions, but you need more
memory space
Learning =
- Changing an existing rule
- Adding and modifying exception rules
- Add exceptions and exceptions to exceptions
Summary
- Internal model is the basis for learning
- It makes an actor better equipped for survival
- Actors with a failing (too simple/wrong rules) model have a competitive disadvantage
- A better model will survive and spread (through it’s owner)
- A rich model helps it owner to do better
- In a CAS, we always see working internal models, not perfect internal models
- We see a snapshot of internal models in various stages
