Agent Dynamics in Complex Multilevel Systems of Systems of Systems Jeffrey Johnson

Agent Dynamics in Complex Multilevel

Systems of Systems of Systems

Jeffrey Johnson

The Open University - UK

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

e.g. a conversation is emerges from interaction

e.g.

did you see the football ?

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

e.g. a conversation is emerges from interaction

e.g.

did you see the football ?

no

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

e.g. a conversation is emerges from interaction

e.g.

do you like Paris ?

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

e.g. a conversation is emerges from interaction

e.g.

do you like Paris ?

yes, a lot

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

e.g. a conversation is emerges from interaction

e.g.

I love the food …

and the wine

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

e.g. a conversation is emerges from interaction

I love the food …

and the wine

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

e.g. a conversation is emerges from interaction

Generally the interaction is computationally irreducible

I love the food …

and the wine

… unpredictable

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

e.g. a conversation is emerges from interaction

Generally the interaction is computationally irreducible

I love the food …

and the wine

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

e.g. a conversation is emerges from interaction

Generally the interaction is computationally irreducible

I love the food …

and the wine

… unpredictable – cannot make point predictions

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

More generally the behaviour of people depends on interaction with many other people.

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

More generally the behaviour of people depends on interaction with many other people.

in networks

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

More generally the behaviour of people depends on interaction with many other people.

in networks

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

More generally the behaviour of people depends on interaction with many other people.

in networks

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

nodes in networks can be anything

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

nodes in networks can be anything

e.g. countries

Taiwan

France

USA

JapanIceland

Ghana

Mexico

Peru

Australia

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

What is there between the micro and the macro level ?

Taiwan

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

What is there between the micro and the macro level ?

Taiwan System

Sub-System

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

What is there between the micro and the macro level ?

Taiwan System

Sub-System

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

What is there between the micro and the macro level ?

Taiwan System

Sub-System

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

Taiwan is a system of systems of systems

Taiwan System

Sub-Systems

System of

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems – dynamics emerge from the

interaction of the parts

Taiwan is a system of systems of systems

Taiwan System

Sub-Systems

System of

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems

How can we:

- represent multilevel systems of system of systems ?

- model their dynamics at any level ?

- integrate their bottom-up and top down dynamics ?

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Complex systems

How can we:

- represent multilevel systems of system of systems ?

- model their dynamics at any level ?

- integrate their bottom-up and top down dynamics ?

What role do agents play in all this ?

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Relations give us networks of agents

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Relations give us networks

Micro-level connectivity structure matters

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

3 binary relations one 3-ary relation

Binary relations are not rich enough

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Relational Structure

Binary relation 3-ary relation 4-ary relation

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Multilevel Systems & Hypernetworks for the dynamics of complex multilevel systems

Polyhedral Connectivity in Hypernetworks

0- near polyhedra

The intersection of two simplices is called their shared face.

They are q-near if their shared face has dimension q

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

1- near polyhedra

0- near polyhedra

The intersection of two simplices is called their shared face.

They are q-near if their shared face has dimension q

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity in Hypernetworks

0- near polyhedra

The intersection of two simplices is called their shared face.

They are q-near if their shared face has dimension q

1- near polyhedra

(and also 0-near)

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity in Hypernetworks

1- near polyhedra

2- near polyhedra

0- near polyhedra

Polyhedral Connectivity in Hypernetworks

Polyhedra can be q-connected

through shared faces

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity in Hypernetworks

Polyhedra can be q-connected

through shared faces

1-connected components

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity in Hypernetworks

Polyhedra can be q-connected

through shared faces

1-connected components

Q-analysis: listing q-components

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity in Hypernetworks

Polyhedral Connectivity & q-transmission

The Bull The Dog The Ship The Royal Oak

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity & q-transmission

My dog’s got

No nose

How does he smell ?

The Bull The Dog The Ship The Royal Oak

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity & q-transmission

Terrible !

How does he smell ?

The Bull The Dog The Ship The Royal Oak

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity & q-transmission

Terrible !

How does he smell ?

The Bull The Dog The Ship The Royal Oak

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity & q-transmission

My dog’s got

No nose

How does he smell ?

The Bull The Dog The Ship The Royal Oak

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity & q-transmission

My dog’s got

No nose

How does he smell ?

The Bull The Dog The Ship The Royal Oak

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity & q-transmission

My dog’s got

No nose

How does he smell ?

The Bull The Dog The Ship The Royal Oak

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Polyhedral Connectivity & q-transmission

Terrible!

How does he smell ?

The joke is q-transmitted through the backcloth

The Bull The Dog The Ship The Royal Oak

Hypernetworks for the dynamics of complex multilevel systems ISCTE Lisbon 12-Jan-2008

Polyhedral Structure

provides a new way of representing

& multilevel systems …

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

e.g. take a set of 3 blocks

Formation of simplices hierarchical structure

{ }AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

{ }R

R

Formation of simplices hierarchical structure

e.g. take a set of 3 blocks assembled by a 3-ary relation

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

{ }R

Formation of simplices hierarchical structure

e.g. take a set of 3 blocks assembled by a 3-ary relation R

The structure has an emergent property

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

{ }R

Formation of simplices hierarchical structure

Level N+1

Level N

n-ary relation

assembles elements

into named

structures at a higher

level

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

{ }R

Formation of simplices hierarchical structure

Archn-ary relation

assembles elements

into named

structures at a higher

level

R

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example:

gin, tonic, ice, lemon

gin

ice

tonic

lemon

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example:

gin, tonic, ice, lemon

Gin & Tonic is a Gestalt !

gin

ice

tonic

lemon

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example: Road Accidents

upset

rain

speed

tired

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example: Road Accidents

upset

rain

speed

tired

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

The accident is a whole

the individual parts may not cause an accident

Example: Road Accidents

upset

rain

speed

tired

can weight importance

of the vertices

0.6 contribution

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example: Road Accidents

upset

rain

speed

tired

can weight importance

of the vertices

0.6 contribution

0.8 contribution

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example: Road Accidents

rain

speed

tired

can weight importance

of the vertices

0.6 contribution

0.8 contribution

0.3 contribution

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

upset

Example: Road Accidents

upset

rain

speed

tired

can weight importance

of the vertices

0.6 contribution

0.8 contribution

0.9 contribution

0.3 contribution

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example: Road Accidents

upset

rain

speed

tired

can weight importance

of the vertices

0.6 contribution

0.8 contribution

0.9 contribution

0.3 contribution

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Fuzzy 0.6 contribution

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

‘scoring a goal’

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

‘the old 1-2 move’

‘scoring a goal’

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

‘The old 1-2’ as a trajectory in multidimensional space

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

‘The old 1-2’ as a trajectory in multidimensional space

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

‘The old 1-2’ as a trajectory in multidimensional space

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

‘The old 1-2’ as a trajectory in multidimensional space

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

‘The old 1-2’ as a polyhedral trajectory in multidimensional space

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

formation of a polyhedron is a structural event

Passing ball event

Goal scoring event

Example: robot soccer

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Multilevel Structure and Set Definition

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Multilevel Structure and Set Definition

Complex systems generally have many ‘levels’

cells → organs → animals

chips → boards → computers

people → departments → institutions

nurses → … hospitals … → health trusts

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Multilevel Structure in Human Systems

Clipboard challenge

you are studying a hospital, how many

things can you see ? € 1.00 for each !

e.g.

bandages, desks, nurses, uniforms, aspirin, light-switches, door knobs, beds, pencils, pills, doctors, trolleys, windows, doors, wards, flowers, canteens, children, teddy bears, waste bins, curtains, computers, car parks, cups, …

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Multilevel Structure in Human Systems

Clipboard challenge

bandages, desks, nurses, uniforms, aspirin, light-switches, door knobs, beds, pencils, pills, doctors, trolleys, windows, doors, wards, flowers, canteens, children, teddy bears, waste bins, curtains, computers, car parks, cups, …

This is a multilevel mess !

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Multilevel Structure in Human Systems

Sorting out the hierarchical mess …

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Multilevel Structure in Human Systems

Lifting coherent vocabulary from the soup

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Multilevel Structure in Human Systems

Lifting coherent vocabulary from the soup

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Multilevel Structure in Human Systems

Lifting coherent vocabulary from the soup

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

AND and OR aggregations in multilevel systems

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Representing Complex Systems & Dynamics

Backcloth, Traffic & Type-1 Dynamics

Traffic on the multi-dimensional backcloth

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Representing Complex Systems & Dynamics

Backcloth, Traffic & Type-1 Dynamics

Traffic on the multi-level backcloth

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Representing Complex Systems & Dynamics

Backcloth, Traffic & Type-1 Dynamics

Traffic on the multi-level backcloth - coherence

Cannot look at just one level

Traffic must aggregate coherently over the structure

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Unpacking High-level Structure in Human Systems

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

System time and System Events

System dynamics as traffic on a fixed multilevel backclothAESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Representing Complex Systems & Dynamics

Backcloth, Traffic & Type-1 Dynamics

Towards representing systems of systems of …

Cannot look at just one level

Traffic must aggregate coherently over the structure

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Agents, Climate Change & Systems of Systems of Systems

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Agents, Climate Change & Systems of Systems of Systems

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Government Policy

?

?

Agents

Conclusions

- complex systems - systems of systems of systems

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Conclusions

- complex systems - systems of systems of systems

- need new multilevel mathematics to integrate

hypernetworks are necessary if not sufficient

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Conclusions

- complex systems - systems of systems of systems

- need new multilevel mathematics to integrate

hypernetworks are necessary if not sufficient

- hypernetworks give a theory of aggregation

- mappings aggregate over hypernetworks

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Conclusions

- complex systems - systems of systems of systems

- need new multilevel mathematics to integrate

hypernetworks are necessary if not sufficient

- hypernetworks give a theory of aggregation

- mappings aggregate over hypernetworks

- governments think at the macrolevel (it’s easier!)

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Conclusions

- complex systems - systems of systems of systems

- need new multilevel mathematics to integrate

hypernetworks are necessary if not sufficient

- hypernetworks give a theory of aggregation

- mappings aggregate over hypernetworks

- governments think at the macrolevel (it’s easier!)

- agents are needed at the microlevel

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009

Conclusions

- complex systems - systems of systems of systems

- need new multilevel mathematics to integrate

hypernetworks are necessary if not sufficient

- hypernetworks give a theory of aggregation

- mappings aggregate over hypernetworks

- governments think at the macrolevel (it’s easier!)

- agents are needed at the microlevel

- new opportunities for agent based modelling !

AESCS'09 Pacific-Asian Association for Agent-based Approach in Social Systems Sciences November 13-14, 2009