Introduction to Complexity Science

1
Introduction toComplexity Science
Working with Systems
2
Systems

• Systems science is clearly concerned with
  systems. Two kinds are relevant here

• We can think of the things in the world that
  provide the demand for systems science problem
  systems
• genomes, cities, corporations, the health
  service
• Sometimes the things that we build to solve these
  problems are also thought of as systems
• databases, simulation models, virtual
  environments, etc.
• The word system is clearly a very general term
  that can be applied to many, many things. What
  do we mean by it?

3
What is a System?

• At root a system is a set of individual
  components that are linked by relationships of
  some kind to form a whole.

Other definitions exist, but this captures the
various levels of description involved studying
any system part and whole, system and its
surroundings, etc.
4
A Biological Example

• Many genes code for proteins that either promote
  or inhibit the transcription of other genes.
  Together, such genes form genetic regulatory
  networks.

• can we infer the structure of these networks from
  micro-array data, samples of transcription factor
  levels?

• is the data too noisy or irregular for this to
  work?
• could we simulate a particular real network?

• could we use simulations to discover how these
  types of genetic regulatory network behave in
  general ?

5
A Geographical Example

• One of the main factors influencing the design of
  built environments such as airports, museums,
  libraries, etc., is the way in which pedestrians
  move through these spaces.

• How can designers structure environments such
  that the behaviour of pedestrians is appropriate
  or desirable?
• ensuring safe timely exit and escape behaviour
• avoiding bottlenecks, congestion, crowding, etc.
• maximising impact of advertising space,
  facilities, etc.
• Virtual environments and models of pedestrian
  movement can provide designers with feedback
  before bricks are laid.
• How can we ensure that these computational
  solutions are fit for purpose usable, accurate,
  flexible?

6
An Engineering Example

• Many engineered products are complex assemblies
  of sub-components manufactured by many different
  companies.

For example, JPL, Lockheed, and Boeing, among
others, collaborated on the design of the Genesis
11 spacecraft.
Effectively tracking design changes and their
ramifications requires understanding how the
relations between components of the system
reflect the relations between the firms
collaborating to build it.
How might complexity science improve the
management of this information?
7
A Medical Example

• In 2001, an outbreak of foot mouth (a disease
  that affects several species of livestock) cost
  the UK 5bn

• within 14 days it had covered the entire country
• at its height nearly 50 cases a day were being
  detected
• millions of animals were slaughtered
• hundreds of farmers lost their livelihoods
• the rural tourism industries lost billions of
  pounds

A poor understanding of livestock transportation,
disease behaviour, and vaccination, coupled with
bureaucratic delays poor co-ordination created
a rapacious epidemic. Could we have reduced the
impact of foot mouth?
8
A Science of Systems

• There have been several attempts to develop tools
  to help solve problems like the ones listed on
  the previous slides.

• Such efforts are kinds of systems science
• cybernetics, systems theory, dynamical systems
  theory, complexity, control theory, information
  theory, etc.

Each of these approaches attempts to provide
frameworks for thinking about and analysing
systems in general. Motivating these endeavours
is an assumption that the systems mentioned in
earlier slides are fundamentally similar their
differences are merely superficial. If this is
true, is there something to be gained from
studying such systems in general, rather than
individually?
9
Levels of Description

• It is important to appreciate the subjectivity of
  a systems perspective there is no single
  correct level of analysis.

• For instance, human DNA can be understood as
• a set of genes, a string of bases, a large
  molecule, etc.
• Each perspective is valid, but each differs from
  the others in important respects, and is relevant
  at different times.
• Often the success of a particular approach is
  crucially dependent on choosing an appropriate
  level of description
• detailed enough to capture critical system
  behaviour
• yet abstract enough to avoid unnecessary
  complexity
• Which aspects to visualize, model, etc., depends
  on the nature of the problem that is being solved.

10
Describing System Structure

• The implications of a systems structure will
  tend to be domain-specific, but there are also
  general considerations

• types of atomic entity how many? what kind?
• e.g., herds of cattle sheep, diseases, vaccines
• types of interaction how many? what kinds?
• e.g., transportation, infection, vaccination,
  etc.
• type of connectivity sparse (rural) vs. dense
  (city)
• e.g., road rail networks, infection vectors,
  etc.
• degree of uniformity homogeneous vs.
  heterogeneous
• e.g., random or grid-like vs. structured in some
  way
• inputs outputs how many? what kind?
• e.g., open system vs. closed system?

11
Sub-systems Coupling

• It is often useful to consider an entire system
  as divided into parts, e.g., because they seem
  relatively independent.

E.g., the eye brain can be considered to be a
single cognitive system, or to be distinct
sub-parts of the system.
They exchange signals via nervous tissue, the eye
supplying sensation while the brain controls the
ocular muscles. Likewise, robot and environment
interact in many ways, e.g., via sensors and
motors.
Sub-systems that influence one another in this
way are said to be coupled.
12
The Eye of the Beholder

• Like choosing a level of description, deciding
  what counts as inside or outside a system, or how
  to divide one into sub-systems is a subjective
  issue. For example

• it may be useful to consider a tool to be part
  of the agent, or a robots wheels part of the
  environment, etc.
• We often treat an external system as a part of
  our body
• prosthetic devices such as eye-glasses,
  pacemakers
• tools (e.g., a hammer), vehicles (e.g., a
  bicycle, a car)
• or sometimes consider a body part to be external
  to us
• e.g., when a body part is anaesthetised or fails
  somehow
• Indeed, sub-systems tend to be noticed as
  separate only when they fail in some manner

13
Hierarchy vs. Anarchy

• An important distinction separates systems that
  exhibit a hierarchical structure from those that
  are disordered.

Many man-made systems feature central
controllers, or higher authorities that organise
lower-level entities.
Structures like these are intended to generate
well-ordered behaviour. In contrast, many natural
systems are not structured in this way, yet are
still capable of generating well-organised,
coordinated behaviours.
For example, the self-organisation of ant
colonies, or traders at the New York stock
exchange.
14
Describing System Behaviour

• It is typically more important to characterise a
  systems actual behaviour, rather than its
  structure.

• Some systems have no behaviour (e.g., a fixed
  classification system), but most do.
• some systems are static until acted upon in some
  way
• many man-made computational systems, for instance
• in contrast, some systems have an intrinsic
  dynamic
• a brain? an ant-colony? economy? online community?

• Like structure, what behaviour is attended to is
  subjective.
• long-term, short-term, low-level, high-level,
  etc., etc.
• In what ways can we classify system behaviour?

15
Stability

• In many cases, we wish to understand under what
  conditions a system will remain stable.

• will a newsgroup remain robust as new users are
  added?
• will an ecology remain stable as new species are
  added?
• is the economy crashing? will a stock retain its
  value?
• is the market for our product changing? how?
• In fact, it is often not stability per se that is
  of interest, but the extent to which a system has
  departed from stability.
• How is the system being perturbed? How is this
  perturbation being coped with? What results from
  it?
• Are we able to alter the system? Can we
  effectively change it in desired ways?

16
Emergence

• The behaviour that a system exhibits at one level
  of description may be very different from that at
  other levels.

Systems that appear disordered at a low level
(such as ant colonies, crowds, economies, etc.)
may never-the-less exhibit ordered behaviour at a
higher level. For example

• Jupiters Great Red Spot a huge gas cloud
• termite mounds, bee hives, wasp nests
• efficient market prices the invisible hand
• traffic jams, crowd behaviour, fashion cycles

When high-level ordered behaviour arises from the
un-coordinated actions of lower-level entities,
it is termed self-organization or emergent
behaviour.
17
Adaptation

• Adaptive systems change over time such that they
  come to suit their environment they adapt to
  their surroundings.

Evolution by natural selection is the primary
example of an adaptive process, but many other
types exist

• Learning e.g., shaping an organisms tastes,
  fears, etc.
• Imitation e.g., trading behaviour on a stock
  exchange
• Competition e.g., pop bands compete for an
  audience

18
Behaviour from Structure?

• So far we have talked about structure and
  behaviour separately, but it is clear that they
  are intimately linked.

• How does a companys management structure
  influence its behaviour? Its flexibility,
  quality control, creativity?
• How does Sotons traffic network influence rush
  hour?
• How does the human genome influence
  morphogenesis?
• Will a particular virtual reality encourage
  collaboration?
• Can we confidently make changes to a systems
  structure in order to bring about desirable
  changes in behaviour?

In order to answer this type of question, we need
to do more than just describe a systems
structure behaviour.