Holism and emergent properties

Copyright 2014-17 Graham Berrisford.

One of about 300 papers at http://avancier.website. Last updated 21/02/2017 17:32

 

“The concept [emergent property] has been used to justify all sorts of nonsense.” Gerald Marsh.

The discussion below leads to the conclusion there are several ways to define “emergent property”.

There is the correct definition and various others.

 

Holism

 “General System Theory… is a general science of wholeness… systems [are] not understandable by investigation of their respective parts in isolation.” Bertalanffy

Bertalanffy wanted to shift attention from the parts of a system to the whole.

·         how bodily organs cooperate to sustain the whole organism

·         how human actors cooperate to meet the goals of a social group.

·         how business processes cooperate to meet the goals of a whole enterprise.

 

Holism is the first of many words in system theory that are as often abused as used.

To take a holistic view does not mean looking at everything in a system.

It means rather looking at how its essential parts cooperate to the benefit of the whole.

 

The concept of a whole implies something divided into parts.

Systems thinkers often promote "holism" and deprecate "reductionism", but these are two side of the same coin.

To think holistically is to think how parts cooperate in the processes of a whole. 

You can't take a holistic view of how parts cooperate if don't know what the parts are.

 

So holistic and reductionist views are complementary, not in opposition.

A system designer has to take both views and flip between them.

You can't see any property of a macro-level system as “emergent” if you have never considered subsystems separately.

 

Interpretations of holism include the following.

 

Parts are interrelated so strongly they cannot exist independently.

The peripheral parts of a biological entity cannot exist independently.

But the core parts can do this (you can live without your limbs).

A bicycle and its rider exist independently.

IBM could be subdivided into smaller independent businesses.

 

A part cannot be understood without reference to the whole.

OK, but the boundary of the whole is a choice made by a describer.

Systems are recursively composable and decomposable.

You can understand a system or subsystem at any level of decomposition you choose to.

 

The whole is greater than the sum of its parts.

This means parts in connection do things they cannot on their own.

This comes as no surprise to system designers and so-called “reductionist” system engineers.

Because such emergent properties are the primary objective of their design and engineering

 

Encapsulating a system hides its parts.

Sometimes, holism is interpreted to meaning looking at a system as a black box, definable by is interface and its goals.

This – encapsulation - means an external observer cannot see any internal parts of the system.

And therefore, cannot see any of its properties as emerging from parts.

Encapsulation of internal actors and activities is essential in the description of business systems.

System architects start with a vision that encapsulates the system of interest; they define properties required of whole,

The purpose of system design is to obtain properties from the whole that cannot be obtained from isolated parts.

You could rather say properties of parts emerge during the design process from the properties of the whole!

 

Emergent properties

Emergent properties emerge from the interaction of actors or components in a wider or higher process.

Yes, the concept really is that simple!

 

These discrete things

When coupled in a process can

Person and bicycle

Cycle forwards

Bridge and wind

Wobble

Three geese

Fly in a V shape

Person, screwdriver, screw and wood

Drive the screw into wood

Board, CIO and EA team

Standardise and integrate business systems

 

A person and a bicycle are discrete entities.

The process of pedalling the bicycle produces the emergent properties of balance and motion.

In this case, the emergent property was deliberate and wanted.

 

The famous wobble of the Tacoma Narrows Bridge (which led to its collapse) has been called an emergent property of the bridge.

Wrong! It was an emergent property of the wider process in which the bridge and the wind were interacting components.

In this case, the emergent property was accidental and unwanted.

 

You cannot see a property of a whole as being emergent until you see its parts (in isolation) do not have the same property.

The biologist Bertalanffy expressed the idea of emergent properties as “the whole is more than the sum of its parts”.

He said the whole system has properties that appear as new or emergent because the parts do not have those properties.

Implying that the properties emerge from the interaction of actors/components previously considered separately.

 

The definition offered by Laszlo and Krippner (1998) seems to distil most of what most systems thinkers agree about the concept.

“An emergent property is marked by the appearance of novel characteristics exhibited on the level of the whole ensemble, but not by the actors/components in isolation.” Laszlo and Krippner

 

Ashby (1956) was clearer: he pointed out that properties emerge not from merely aggregating parts, but from processes (not found in the parts) that couple the parts.

“Parts can be coupled in different ways to form a whole.

The defining of the component parts does not determine the way of coupling.

From this follows an important corollary: that a whole machine should be built of parts of given behaviour is not sufficient to determine its behaviour as a whole.

Only when the details of coupling are added does the whole's behaviour become determinate.”

 

The so-called “removal of parts” test is misleading.

“Emergent properties are lost when the system breaks down to its components.

The property of life, for example, does not inhere in organs once they are removed from the body.

“When a component is removed from the whole, that component itself will lose its emergent properties.

A hand, severed from the body, cannot write, nor can a severed eye see.” Laszlo and Krippner

 

First, you can indeed remove parts from a homogenous system (remove a fish from a school of fish) without changing the qualitative properties of the system.

Second, you can remove some parts from a multi-purpose system (remote a font from a word processor) without most observers noticing.

Third, writing is not in the hand, seeing is not in the eye, these emergent properties were never in the parts; they required processes that connect those parts to the brain.

Conclusions and remarks

There are various ways to define “emergent property”.

There is the correct definition and several other oft-used definitions.

 

Correctly: a property of the whole that depends on and is derived from interactions between its actors/components.

That is merely a side effect of system nesting and system boundary drawing.

Properties that look at emergent at fine-grained level are not emergent at a coarse-grained level.

If the properties of the whole are designed and unsurprising, then what does “emerge” mean?

 

Commonly: surprising behaviour

This is behaviour that (though perhaps theoretically predictable) only becomes evident when system parts or whole systems act together

The definition becomes more interesting when refined to mean what appears to be mysterious to an observer who has knowledge of the system’s parts.

When an apparently complex property of a whole is derived from surprisingly simple rules followed by individual actors/components.

 

Other meanings of the term

It is obvious that two subsystems cannot (separately) do everything that they can do when coupled in a wider system.

Emergent properties are the very purpose of engineering.

Today however, people use the term “emergent property” with various other meanings.

Discussion would be clearer if people said what they really mean, say:

·         essential-to-purpose property.

·         accidental property.

·         run-time-only property.

 

Read the analysis below for a deeper exploration of such alternative meanings.

 

 

Emergent properties – older analysis for eager readers only

Surely it is obvious that two subsystems cannot do everything (separately) that they can do when coupled in a wider system? Barely worth saying?

Today, people use the term emergent property loosely, with various meanings, and sometimes to express a value judgement.

 

Questions to be explored below include:

·         Does an emergent property differ from any other property?

·         Does an emergent property require all of a system’s actors/components?

 

Before we explore these and other questions; a few notes on outputs and outcomes.

·         Outputs are what a system produces; they are properties of the system.

·         Outcomes are the result of what external entities do with outputs; they are properties of a wider system.

 

Possible outcomes include:

·         Welcome outcomes.

·         Unwelcome outcomes (cf. “the law of unintended consequences”)

·         Unpredicted and welcome outcomes caused by external entities using system outputs in an unexpected way.

Many ways the term “emergent property” is used

Think how you may use the term emergent property, then see if you can match your example(s) to the cases listed below.

 

Suppose you are asked to design a clock to help people tell the time.

Your clock must display an ever-changing number that stays in step with standard time.

You design and build the clock. What next?

You might bury your clock in the ground; it never works and is never used to tell the time.

Else, your clock could be used in various ways that illustrate a variety of behaviours some have called emergent properties.

 

A behaviour triggered by a condition or time passing

E.g. your clock’s alarm is set and goes off in response to the arrival of the set time.

That is a predictable, designed, deterministic behaviour that occurs when the systems advances to a new state.

 

A behaviour just now noticed by an observer

E.g. Long after buying the clock, the purchaser notices the clock has an alarm feature.

 

An unpredicted failure of the system

E.g. your clock’s parts do combine as they are designed to, but they work too fast, so display a number that almost never accurately represents the standard time.

 

An unwelcome outcome

E.g. your clock is used as the timer for a bomb.

That is an unpredicted and unwelcome outcome caused by an external entity using a system output in an unexpected way.

 

An unpredicted but welcome outcome

E.g. your clock is used to teach somebody how to tell the time.

 

A welcome output

E.g. In system testing, your clock’s parts cooperate to display a number that accurately represents the standard time.

 

A welcome outcome in operation

E.g. your clock is bought and used successfully to tell the time.

 

Note that time telling is a welcome outcome of the clock's output being used by a human being.

The time telling is a property of a clock-human system, not the clock alone.

 

Of course, different observers have different ideas about what the welcome or essential properties or purposes are. Take a car for example:

This observer

Wanting this outcome

Sees this as an essential property of a car

Pragmatist

travel to destination

point to point transport

Social climber

status in society

impress the neighbours

Manufacturer

make a profit

profit margin

Cyclist

avoid death on the road

wing mirrors, rear view

 

Below are more ways I have heard the term emergent property applied.

 

A run-time only property

A property observable in an operational system, but not found in its system description.

Perhaps a non-linear or “chaotic” outcome, like a hurricane, or a population crash in an ecology.

 

An essential-to-purpose property

A property necessary to a desired outcome.

This is subjective of course: What is essential in a daffodil, a word processor, a flight of geese, a bicycle, or IBM?

 

An accidental property

A side effect of what has been designed, not purposefully designed for, perhaps revealing of a design error.

 

A surprising property

One that (by comparison with reductionist inspection of a system’s parts) is unexpected or unpredictable.

 

A new-after-change quality

A qualitatively new property, an after-effect of an evolutionary step change, a new property of a new system generation not found in past system generations.

E.g. the first appearance of thorns on a plant.

 

A new-after-change quantity

A quantitative change in an old property, a higher or lower value of a variable.

E.g. Add gears to a bicycle, there is a reduction in pedalling effort.

Merge two companies, their turnover (separately or combined) goes up or down.

Emergent systems

Naturally, when two or more systems cooperate, the wider system may exhibit properties the subsystems cannot exhibit on their own.

And system designers start by defining the properties that will emerge when subsystems are assembled and deployed.

The reason for designing any system is to obtain properties that cannot be obtained from its isolated parts.

 

All said above seem to imply we know what system we are talking about.

And the term “emergent property” implies there is a system to which the property is attached.

Yet, when the term is used, it is often not clear what "system" is being discussed.

So which emerges first? The property or the system?

 

Sometimes the logic goes as follows.

Something happened we didn’t expect; we’ll call that an "emergent property".

Then we’ll define whatever chain of events produced the property to be a "system".

 

Are you familiar with "root cause analysis diagrams" or “cause and effect diagrams”? You can find some on the web.

People draw such a diagram to uncover and show the curious collection of entities and events that have accidentally conspired to cause an unexpected thing to happen.

They might look for root causes under pre-defined headings such a people, management, materials, equipment, measurement and environment.

 

You can of course characterise an event (a problem, an incident, an accident) as “emergent” and then analyse to find some actors/components that interacted in some processes to cause that event.

This root cause analysis can reveal a “system” that nobody was aware of before; it was never designed; its scope was defined only in retrospect.

 

Lo, a system emerges from analysis of a property! However, this system was not understood to be a system in advance of the event happening.

And using this logic, everything that ever happens can be subjected to root cause analysis and then described as the emergent property of its own unique system.

 

Would von Bertalanffy or Ashby have called this accidental conspiracy a "system"? Wouldn’t they rather call it a disorderly and unsystematic process?

Emergent property test questions

Suppose we return to the simple definition that an "emergent property" is a feature of a whole system but not a part of it.

 

The following two test questions have been suggested to distinguish an emergent property from any other property.

The idea is that if the answers are “no”, then the property can be called emergent.

 

Q) Can you observe the property if less than the whole system is present? Can you remove parts and still observe the property?

 

Below are four test cases.

The first two pass the test; the last two fail the test; but it is possible to manipulate the analysis so they pass also.

 

Example 1: The Tacoma narrows bridge, which wobbled so much in a strong wind it collapsed.

Q) Can you observe the property if less than the whole system is present? Can you remove parts and still observe the property?

No. You need both the bridge and the wind.

The emergent property was not the resonant frequency of the bridge, or the collapse.

It was the wobble caused by a process involving the bridge and wind as interacting actors/components in a wider system.

 

Example 2: a rider’s balance on a bicycle.

Q) Can you observe the property if less than the whole system is present? Can you remove parts and still observe the property?

No. You need to both the bicycle and the rider.

Balance is not a property of the bicycle alone.

For the property to exist and be observed, you have to widen the system boundary to embrace bicycle and rider.

 

Example 3: using groupware to send an email

Q) Can you observe the property if less than the whole system is present? Can you remove parts and still observe the property?

Yes, if you can remove parts of your email system that do not contribute to sending an email.

 

Example 4: the speed of a car

Q) Can you observe the property if less than the whole system is present? Can you remove parts and still observe the property?

Yes if you remove parts of the car that do not contribute to its motion.

Remove one wheel from my wagon, and I still keep rolling along.

 

In the last two examples, you could define the property as an emergent property by logically-bounding a system that contains only those parts necessary to enable the property.

But that is a self-referential definition, which you could apply to any property.

Again, using this logic, everything that ever happens can be analysed by root cause analysis and then described as the emergent property of its own unique system.

Mysterious properties of a homogenous system

A heterogenous system comprises many kinds of part.

A homogenous system comprises many parts of the same kind, in which each subsystem instance follows the same rules.

For example:

 

  • Ants in an ant colony
  • Geese in a flight of geese
  • Fish in a school of fish
  • Players in a football team
  • People in a mob on a rampage
  • Buses and drivers in a bus company

 

In some homogenous systems, a subsystem instance has very simple rules to follow.

But when subsystems combine to act in a group, the group may exhibit a behaviour that looks complex – deceptively so – since the part’s rules are so simple.

Such mysterious emergence seems relatively normal in biological, evolved, systems.

 

Consider the V shape of a flight of geese.

Apply the emergent property test questions.

Q) Can you observe the property if less than the whole system is present? Can you remove parts and still observe the property? Yes, provided at least three geese remain in the flight.

You can remove some geese and the flight looks much the same.

 

The example fails the test for an emergent property.

Yet it is often used as an example, surely because it has an element of mystery about it.

 

Here are some quotes from “The demystification of emergent behavior” Gerald E. Marsh, Argonne National Laboratory (Ret) 5433 East View Park, Chicago, IL 60615

 

·         “The concept has been used to justify all sorts of nonsense.

·         It is really more open than simply the combination of two "subsystems" whose design and interfaces are specified before they come together.”

·         there is [no] universally acknowledged definition of emergence.” [However it usually means] "Emergent behavior transcends a mere increase in the behavioral degree of complexity."

·         “Complex behavior can arise from the underlying simple rules”

·         "The emergent properties do not in any transparent way derive from the underlying rules governing the interaction of the system’s components.”

·         "Associated with this phenomenon is a sense of the mysterious.”

 

http://www.gemarsh.com/wp-content/uploads/EMERGENT BEHAVIOR-arXiv.pdf

Further reading?

1 Alex Ryan, “Emergence is coupled to scope, not level”, arXiv:nlin/0609011 v1 (2006).
2 Alexander K. Dewdney, “Insectoids Invade a Field of Robots” Scientific American (July 1991).
3 M. Mitchell Waldrop, “Fast, Cheap, and Out of Control”, Science 248 (1990): 959-961.
4 Randall D. Beer, Hillel J. Chiel and Leon S. Sterling, “An Artificial Insect”, American Scientist 79 (1991): 444-452.
5 Deborah M. Gordon, “The Development of Organization in an Ant Colony”, American Scientist 83 (1995): 50-57.
6 Arun V. Holden, Ed., “Chaos” (Princeton University Press, Princeton 1986).
7 Elizabeth Pennisi, “Searching the Genome’s Second Code”, Science 306 (2004): 632-635.
8 Epigenetics, Nature Insight: Nature 447 (2007): 395-440.
9 Arthur Köestler, “The Ghost in the Machine” (The Macmillan Co., New York 1968).

 

 

All free-to-read materials at http://avancier.website are paid for out of income from Avancier’s training courses and methods licences.

If you find the web site helpful, please spread the word and link to avancier.website in whichever social media you use.