Managing and Controlling a Complex System

1
Managing and Controllinga Complex System

• Nicky Grigg Fabio Boschetti

2
Outline

• Adaptive talk
• Cause-effect relations and CSS
• Recycling some old results

3
Difference between a Complex and Non-Complex
System
A complex system displays emergence and
self-organisation

• Not sure what emergence and SO are
• Global
• Too descriptive, not clear what a CS does
• Not sure what components/agents do
• We can only observe, not control

4
Difference between a Complex and Non-Complex
System
In a complex system cause-effect relations are
not really clear Not clear who does what

• Cause-effect relation is what our science deals
  with
• It is essential for control and manipulation
• It is essential for communication

5
Causality

• Serious philosophical problems in discriminating
  between causality and mere correlation (Hume,
  Russell, etc..)
• Operative definition of causality (Pattee, Pearl)

Useful causation requires control. . Clearly it
is valuable to know that malaria results from
Plasmodium parasites. .. For example, while it
is true that bacteria and mosquitoes follow the
laws of physics, we do not usually say that
malaria is caused by the laws of physics (the
universal cause). That is because we can hope
to control bacteria and mosquitoes, but not the
laws of physics Pattee, H.H. (1997)
6
Causality
7
Causality
8
Connections open and closed
Configuration Interaction
Analytical Sensitivity
matrix M solution measure
open
feedback loops
9
Causality
10
Causality
11
Example
Option 2
Option 1
12
Example
Option 2
Option 1
?
?
?
?
?
?
?
?
?
?
13
Example
Option 2
Option 1
?
?
?
?
?
?
?
?
?
?
14
Example
Option 2
Option 1
?
?
?
?
?
?
?
?
?
?
15
Example
16
Example

• Non trivial interactions
• Emergent behaviour / Patterns
• Adaptation
• Self-Referentiality
• Self-Organisation

17
Example

• Majority always loses
• Optimal when minority as large as possible
• Fluctuations wasted resource
• Supposedly hopeless problem

18
Example
Why is the problem hopeless?

• No director no instruction to individuals
• Each individual acts selfishly
• Conflicting Constrains
• Too many individuals
• Tragedy of the Commons-like problem

19
Example
is it REALLY hopeless?

• No, if the agents understand the problem better
• Collective Intelligence
• This can be achieved while agents still behave
  selfishly
• Difference between Contribution and Impact

20
Managing a Complex System
Contribution versus Impact

• Contribution part a vessel plays in the final
  outcome.
• Impact how the agent directly affects the
  outcome or,
• what the outcome would be without the agents
  intervention.

21
Contribution 1
-

0
Impact 0
22
Contribution 2
-

2
Impact 2
23
Contribution versus Impact
Impact
Contribution
24
Managing a Complex System
Zone 2
25
Managing a Complex System
26
Test cases
27
Test cases
Zone 2
Zone 1
28
Future work
Zone 1
29
Test cases
30
Managing a Complex Systemdoes not need to be
complex

• all is needed in order to implement a COIN
  strategy are some basic bookkeeping and a handful
  of elementary operations (,-,,/).
• Basically, the entire approach could be performed
  with pen and paper by agents with only primary
  school training.

31
Discussion
In a CS cause-effect relation unravel in time
32
Connections open and closed
Configuration Interaction
Analytical Sensitivity
matrix M solution measure
open
feedback loops
33
Discussion
In a CS cause-effect relation unravel in time