Introduction to Complexity Science

1
Introduction toComplexity Science
Science, Models and Theories
2
Would-be Worlds?
A fluke?

• Superbowl XXXII
• Denver 28, 15, 34, 19 24
• Green Bay 24, 17, 21, 18 21

3
Rerunning the Tape
4
Settling Empirical Questions

• Why should a simulation model, any simulation
  model, be able to settle an empirical question?

• Could these empirical questions be settled via
  simulation modelling?
• Do I wear underwear while lecturing?
• Is there life on Mars?
• If these questions cannot be settled via
  simulation modelling, why should the re-running
  the tape thought experiment, or Denver fluked
  it question be any different?

5
What is a Model?

• The word model is applied to many things

Are these the same type of thing?
6
Standing in for..

• Models are often said to stand in for the real
  phenomena or systems that they represent.

Its more accurate to say that models idealise
the systems that they represent.
7
Scientific Modelling

• Physics lays out a role for scientific models

• theory-led
• not driven directly by observations of reality

Theory
8
Making Predictions
In the exact sciences we are used to expecting
our models to make accurate predictions.
Requires strong faith in theory.
Nevertheless, such models are still idealisations.
9
Simulation Models
What of simulation models (of complex systems)?
Are they different or special in some respect?
Should simulation models somehow
be maximally accurate and maximally general?

• Confusion of this type can lead to problems
• validation does it make accurate predictions
• verification is it bug-free?

10
Reasons to Simulate Pt. 1

• Bonabeau Theraulaz (1994)
• Ray (1994)
• Taylor Jefferson (1994)
• Miller (1995)
• Di Paolo (1996)
• Sober (1996)
• Noble (1997)
• etc.

Simulations are better than equations

• more realistic
• more flexible
• easier to construct
• Also in some sense
• clearer
• more explicit
• inter-subjective

Is this really true?
11
Doomed to Succeed
Where models strive for realism, per se, the
modelling cycle is broken.
Repeated tinkering leads to a model that matches
the modellers pre-conceptions, generating only
confirmatory observations.
Tom Ray what we see is what we know
12
Opaque Thought Experiments
A suggestion where simulation models address
complex systems that we do not yet
understand treat them as a kind of thought
experiment.

• Thought experiments
• generate no new facts about the world
• but reorganise ones theoretical commitments

• These ones are
• partially mechanised
• opaque

13
The Lure of Artificial Worlds
Why are overly complicated and mysterious
kitchen-sink simulation models so attractive?
14
Complexity Science Models
So, simulation models of complex systems should
be elegant, beautiful and lean.
But also simple and unrealistic
15
Lies, Damned Lies, and Simulation
Our failure to establish an effective methodology
for simulation modelling may be fatal to the
field. Statistics is crucial to modern
science. Yet stats is typically regarded as
deeply dubious. Its what shysters (e.g.,
politicians) use to sell you their point of
view. Another decade of dodgy simulation models
used to back up important policy decisions could
see simulation modelling tarred with the same
brush.