Introduction: The evolution of description

Copyright 2016 Graham Berrisford. One of about 300 papers at Last updated 16/11/2017 18:27


Enterprise architecture is about business system planning.

It can be seen as applying the principles of general system theory – which we’ll get to later.

This paper is one in a family of related papers – listed at the end.


First, what theories underpin general system theory?

To apply system theory is to describe a system as it is observed now, or envisaged in the future.

What is a system description?

This paper starts an explanation that is continued in other papers.


Systems and descriptions of them... 1

Seven steps in the evolution of description.. 2

What else makes humans special?. 7

A general description theory. 9

A general system description theory. 10

Conclusions and remarks. 12


Systems and descriptions of them

Things change, and in the end, all is unstable.

A system is a transient island of stable behavior in the ever-changing universe.

We can point to any observable thing, small or large, and name that thing.

What differentiates a "system" from any old named thing is that a system has one or more stable roles and rules

For example, a game of chess, a tennis match, a business system, a software system, or anything describable in the form of a causal loop diagram.


Systems are islands of orderly behavior that are describable in terms of:

·         system actors - persistent entities – active structures – locatable in space

·         system activities – behaviors over time – which follow some logic or law – and change the state of the system.


To apply system theory is to describe those regular behaviors of an entity that are relevant to some given aims(s) and/or concern(s).


The description of a natural system

Our sun and its planets interact in the solar system by following the laws of gravity and motion.

It is describable as a system because its behavior conforms to laws.


Russell Ackoff encouraged people to distinguish (but went on to confuse himself) abstract and concrete systems.

1.      The "solar system" is an abstract system description; it is our way of making sense of a reality.

2.      The relevant bodies out there in space are the concrete system realisation.

As will be discussed later, the second is more than the first; the first is a selective abstraction from the second.


The description of a designed system

The solar system is a natural system; it predated the humans who first observed it.

By contrast, consider a business system that is designed and created by humans.

It too can be described as a system because its behavior conforms to some logical rules, for example:


A business system



Place order

Send invoice

Send payment

Send receipt


Some branches of systems thinking are about social entities rather than social systems in the system theory sense.

The behavior of an ever-changing social entity is not describable as a “system”.

However, if the entity that manages the social entity (and directs its behaviour) is orderly, then it might be describable as a “meta system”.

Seven steps in the evolution of description

Only humans have the ability to abstract system descriptions from realities and document them.

Nevertheless, in evolutionary terms, that is an advanced form of what other animals can do.

This section discusses the evolution of description, which leads to a discussion of system description theory.


Before life, the universe was composed of matter and energy that changes over time.

There was physical matter and energy in space; and a continuous process of change to that matter and energy over time.

This process led to some systems, like the solar system, that repeat behaviors in an orderly way.


Before life however, there was no description of the universe or any system in it.

For eons, there was stuff happening, and even some orderly systems; but there was no description of those systems.

(Or else, there was only a metaphysical description by God that is unknowable).


Eventually, biological evolution led to humans with the ability to describe things using words.

But a Darwinian explanation of description cannot start from words, it has to start earlier.

1: Recognition

Organisms evolved to recognise things.

How do we know an organism recognises things that resemble each other, are of the same type?

We ask: Does it repeatedly react to particular things of that type in the same or appropriate way?

If yes, then that knowledge of that type must somehow be encoded/symbolised in the animal’s biochemistry (the “how” is irrelevant).


E.g. sunflowers recognise and follow the sun’s daily repeated passage across the sky.

Probably, even single cell animals may recognise something about the state of their environment, but sunflowers are a more obvious example.

2: Memory (using mental models)

We do remember particular things.

But does the brain work by remembering everything and comparing every new thing with everything remembered?

The evidence suggests brains store and recognise patterns of related sensations, or mental models of loosely-defined types.

And that recognising family resemblances gives organisms an advantage in the struggle for survival.


There are infinite overlapping family resemblances between things.

Some are more recognisable, universal or useful than others.

Animals must distinguish firm and stable things of the type we call “solid” and from fluid and stirrable things of the type we call “liquid”.

Predators must recognise things of the type we call “prey”.

Have you ever engaged a kitten by wiggling a piece of string?

Then you know the type we might call “mouse tail like” is somehow encoded in a cat’s biochemistry.

(It may be relevant that Hubel and Wiesel (1959) showed cat’s brains have cells dedicated to detecting movement in slit-shaped spots of light.)


A family resemblance

Long, thin, wiggly


Naturally, members of the same species tend to conceptualise realities in the same ways.

They share a common bio-chemistry; they have similar experiences and similar needs.

However, their knowledge of the world does not have to be precise or perfect.


Evolution does not require an animal to have a perfect description of anything.

Holding mental models of the world helps animals to manipulate things in the world and predict their behavior.

But all mental models are partial and flawed models of reality.

These descriptions need only be accurate enough to help animals survive and breed.

Animal minds evolved to model reality, not exactly as the world is, but just well enough to sense, predict and direct events that matter to survival.

3. Learning (creating mental models)

Animals evolved to learn about things in their environment

They encode perceptions in memory and recognise when new things resemble old ones.

E.g. A bird learns by trial and error to recognise things that instantiate the type “edible thing”.


Creating and remembering new descriptions helps animals to recognise food, mates, friends and enemies.

Having flexible mental models of the world helps animals to survive, thrive, and pass their genes on.

The more intelligent the animal, the greater its ability to abstract descriptions from realities.


Most animals have no words to describe things and family resemblances between them.

Here, because we are discussing these matters in writing, we have no option but to use words.

We say a cat recognises a long, thin wiggly thing as being of the type we label “mouse tail”.

We learn that a flying thing with feathers and a beak is a “bird”.


A family resemblance

has feathers, has a beak, can fly.


Words are so natural to us that our mental models may take verbal forms.

For more on mental models, read How the brain works.

4. Communication (sharing mental models)

The types of things recognised by animals become evident in the symbols or tokens they use to communicate.

E.g. an animal sounds an alarm call to signal a new instance of the type we call “danger”.


To communicate, animals translate internal/organic mental models into and out of external/inorganic forms of description.

Symbolic languages include: facial expressions, sounds (calls, barks, whistles), smells, movements, gestures and manipulated materials (e.g. nests).

By using such physical symbols, animals share knowledge about food, friends and enemies, and signal mating intentions.


Communication by messages

Many birds communicate by singing.

One may perceive danger and sound an alarm call, another hears that call and internalises it as sense of danger.

Thus, the two birds share an internal mental model of the current situation as being what we symbolise as “dangerous”.


Honey bees must recognise and remember things of the type we might name “pollen location”.

Famously, they also communicate about pollen sources by performing and observing wiggle dances.

The dance of a honey bee encodes particular values for the variables “distance” and “direction” of a “pollen location”.

The proof that bees “know” the symbols for these two types is that bees do find pollen where it was described.


Externally, bees communicate pollen locations by dancing.

Internally, their mental models of pollen locations are somehow encoded in their biochemistry (how does not matter here).

They have to translate internal descriptions of reality into and out of external descriptions of reality.

One bee finds some pollen, encodes its distance and direction in private memory, and later, translates that information into a public dance.

Another bee decodes the information from the public dance, translates the information into private memory and later, finds the pollen.


Beehive communicaiton

Pollen location descriptions

<create and use>                   <idealise>

Honey bees   <observe and envisage> Real world pollen


Communication using shared memory spaces

Animals can leave a smell on a tree to say “This is my territory”.

They can shape materials into a nest and signal to a potential partner “I am ready for mating” (cf. Ashby’s sticklebacks later).

The smelly tree and the nest are persistent external memories; they are shared memory spaces.


For more on communication in social systems, read A communication theory.

5. Verbalisation of communications

There is nothing human-specific above; we were not the first to remember descriptions of things and communicate them.

What sets us apart is the use of words (and graphical representations of words) to remember and communicate information.


Communication by messages

Any matter or energy flow can be used by one actor to send information to another.

Evolution gave us humans a unique and dramatically well-developed communication tool.

We can create and use an infinite variety of sounds to symbolise things and types of things.


We translate internal mental models into and out of external oral descriptions.

We give voice to and hear verbal messages ranging from short and simple to long and complex.

Our messages contain descriptions, decisions and directions

Oral communication was a huge step forward for mankind, and is essential to most peoples’ lives today.


Communication using shared memory spaces

We have a second huge advantage when it comes to sharing mental models.

We have shared memory spaces that far exceed those other animals can use - in scope, complexity and value.

We can record oral descriptions, decisions and directions using that triumph of human invention - the written record.

Thus, we can translate internal mental models into documented models for posterity, for agreement and for testing.


Recalling a biological memory has the effect of modifying it to some extent, if not as far as in "false memory syndrome".

For this and other reasons, the externalisation of our mental models – into publicly shareable written, audio and visual forms - is vital.

Written communication is so important to modern society that schools prioritise the teaching of reading and writing over other subjects.


Formalisation of verbal language

Natural languages give most of us enough communication capability to get by.

But natural languages are biological rather than logical.

It isn’t just that a word can have several different meanings (e.g. “flagging” is used to mean tiring, highlighting or bringing to a halt.)

The grammar we use is loose, and our natural language expressions can be unclear or ambiguous.


“Evolution is the accumulation of small errors that turn out to be advantageous” (Steve Jones?)

Perhaps ambiguities in verbal communication are essential to life, since they lead to innovation.

However, they do limit our ability to extract the full and intended meanings of natural language statements.

So, natural language is not precise and rigorous enough to support all our endeavours.


Our ability to communicate verbally hugely amplifies the intelligence nature gave us.

It enables us to collaborate with others in complex projects to do complex things and create complex objects.

However, to specify complex artefacts, machines and systems we need more formal domain-specific languages.


When formalising systems we have to be more precise about the meanings of words used in those communications.

We use words to fix family resemblances in the form of type definitions.

Note that our type definitions may be polythetic rather than strictly monothetic.

E.g. If this type is polythetic, then penguins can qualify as birds.



Family resemblance


has feathers, has a beak, can fly.


For more, read A language theory.

6. Mathematics and computing

Here (as explained later) what may be called "type theory" is more directly relevant than set theory.

Terms used here include.

Thing: a discrete structure, object, behavior, event or description.

Type: an intensional definition of a thing; a description of the property type(s) that things of that type embody or give values to.

Instance: an embodiment, by one thing, of the properties of a type.


This triangle represents how types idealise particular things.


General Types

<create and use>                <idealise>

Typifiers   <observe and envisage> Particular things


Mathematicians and software engineers often create and use a rigid monothetic type definition.

That is, a type that defines properties necessary and sufficient to be a thing of that type.


Type name

Type properties (Genus: differentiators)


A plane shape that has a boundary equidistant at every point from a fixed central point.


What may be called "type theory" is more directly relevant to system theory than basic set theory.

A type is defined by listing other types/properties that set members share.


Type name

A generic type with differentiating property types


An animal that has feathers, has a beak, can fly.


A plane shape that has a boundary equidistant at every point from a fixed central point.


A bush that is woody, has thorns, has flowers.


A male that is mature, is unmarried.


A body in space that is large, is incandescent, is remote from other stars.


A body in space that is large, orbits a nearby star.


In natural language a type can be polythetic – it defines more properties than are necessary – so a thing need not instantiate all properties of the type.

For more on the fuzziness of types in nature, read A description and type theory.

7. Artificial intelligence (AI)

AI machines can create and use descriptions using neural networks, polythetic types and fuzzy logic

AI overcomes the limitations of monothetic types by imitating biological processes.


Don’t misinterpret news (in 2017) of AI machines talking to each other and “inventing their own language”.

Look at what they actually said to each other, and you’ll see it looks very unlike a language, and far more like the babble of software going awry.

AI has been hyped for 40 years, and it has recently been joined by “big data”; surely, both are still early on the hype curve.

The ability of a machine to recognise patterns in, or derive types from, data is one thing.

Our ability to create complex system descriptions and then build complex systems is on a completely different level.


An aside on free will.

Robots can do unexpected things, by employing probabilistic fuzzy logic or randomising functions when choosing a response to inputs.

That doesn’t imply they have intention or will, but does prompt the question about whether animals have free will.

My view is that it doesn't matter how animal intelligence works, whether it is deterministic or not.

For almost all practical day-to-day purposes, we have to treat people as making choices of their own free will.

We allow judges and juries to make allowances for cases where people are “out of their minds” or coerced by others to make choices.

A brief recap

A Darwinian explanation of description must start before mathematics and before words.

It starts from the notion that organisms can recognise family resemblances between things.

A family resemblance occurs when the same properties appear in several things, or when a new thing resembles a past thing.


The first half of this paper has told a story of seven steps that lead to increasingly sophisticated creation and use of “types”.

1.      Recognition: sunflowers recognise the sun’s daily repeated passage across the sky.

2.      Memory: a predator instinctively recognises things that instantiate the type “prey”.

3.      Learning: a predator learns by trial and error to recognise things that instantiate the type “edible thing”.

4.      Communication: an animal sounds an alarm call to signal a new instance of the type we call “danger”.

5.      Verbalisation: people may loosely define the type “bird” as having feathers, a beak and the ability to fly. Still, penguins count as birds.

6.      Mathematics and computing: the type “circle is defined strictly, as a round plane figure whose boundary consists of points equidistant from a fixed point.

7.      Artificial intelligence.


Our system theory presumes describers have the abilities at stages 4, 5 and 6 above.

It is presumed we can describe the structures and behaviors of a system by using words to typify particular things.


Read A description and type theory for more on types.

This paper goes on to develop a theory of system description.

What else makes humans special?


Translation between different kinds of description

We can translate between descriptions in three ways.


First, we can translate between internal and external descriptions

It may seem natural to draw a division between internal mental models and external oral/documented models.

A mental model is unconsciously encoded in a biochemical form that is fuzzy, fluid, fragile, prone to decay and be forgotten

A documented model is consciously encoded in much more stable and persistent physical form.


However, we continually translate between internal and external models of reality.

We translate mental models into oral/documented models and vice-versa.

We continually, well nigh automatically, translate between mental models and spoken words or written words.


Second, we can translate between two kinds of external description

We frequently translate between external forms of description: e.g. between speech and writing.

That seems no different in principle from translating between internal and external models: e.g. between translating mental models into and out of spoken words.

And no different in principle from translating up and down the mysterious communication stack from chemistry to consciousness


Third, we can translate between descriptions usable by humans and by machines we make

Computers facilitate human communication of descriptions, directions and decisions.

They require that all descriptions, directions and decisions are translated into patterns of binary digits.

But translating into and out of binary code is no different in principle from translating between other symbolic languages.


Introspection and analysis

Biological evolution led to animals with self-awareness.

Experiments have shown many animals recognise themselves in a mirror, including elephants, apes, dolphins and whales.

However, we are not only more self-aware but also more introspective than other animals.

We analyse what is communicated, we challenge it, we test it.


It is easy to make assertions with no evidence.

Persuasive oratory and/or laziness can lead people to believe false assertions are true.

Historically however, the written record changed the game.

“As soon as writing made it possible to carry communication beyond the temporally and spatially limited circle if those present at a particular time,

one could no longer rely on the force of oral presentation; one needed to argue more strictly about the thing itself.” Luhmann


The written record helps us to examine what is thought, said and written, to challenge it, to test it.

We need tools confirm assertions are true, descriptions are definitions, and hypotheses are knowledge.

We have developed tools to test the truth of assertions: mathematics, logic and the scientific method.



Like everything else in the knowable universe, verbal descriptions can be described.

Logicians seek to describe things using statements that can be tested as true or false.

To do this, they use a generalised form of description called a predicate statement, which takes the form: subject <verbal phrase> object.

For example: “The Lawn Tennis Association <are responsible for defining> the laws of tennis.”


For more on the use of predicate statements, read A language theory.


The scientific method

The word science is rooted in an old word for knowledge.

Science might be simplified to three propositions.

1.      Scientists <observe and predict> the universe – meaning the behavior of entities observable in reality

2.      Hypotheses <conceptualise> the universe – meaning the behavior of entities observable in reality.

3.      Scientists <create and use> hypotheses - which are regarded as knowledge if they successfully predict real-world behavior that is measured.


The three propositions are arranged in a triangle below.


Hypotheses & Knowledge

<create and use>                 <idealise>

Scientists    <describe and predict>   The universe


To generate a hypothesis is easy, and it does not amount science.

As Thomas Edison said “Genius is 1% inspiration, 99% perspiration.”

The inspiration, creating a hypothesis, is 1% of the effort; the other 99% lies in analysing and testing the hypothesis.

For more on the limits to our knowledge, read Knowledge and truth.


There is a lot of pseudo science about – not least in the world of “systems thinking”

For more on that, read “Seven signs of shamanism”.

A general description theory

Describer: a thing (organism or machine) that can create and use a description.

Description: a thing that idealises another thing by recording or expressing some of its properties.


A premise here is: there was no description before life.

There could be no concept before there was an actor able to conceive it.

And there could no description before there was a describer.


Moreover, the survival of describers depends on their ability to create and use descriptions of reality.

Humans use unique tools – verbalisation, the written record and science – to stretch what evolution gave us.


These papers use the term “idealise” rather than “conceptualise”.

And express the three propositions in a more graphical form, as a triangle.

Description theory


<create and use>               <idealise>

Describers <observe & envisage> Realities


Note: this triangle does not match some other triangles you may have come across, like the semiotic triangle.

And there is recursion; the universe of realities includes describers and descriptions.


For more description theory read A description and type theory.

A general system description theory

This table maps system theory to other ways of describing the world.



create and use

Descriptions (logical)

which idealise

Realities (concrete, or relatively so)

Type theory


create and use


which idealise

Set members



create and use


which idealise




create and use

Hyphotheses & knowledge

which idealise

The universe

System theory

System describers

create and use

Abstract system descriptions

which idealise

Concrete systems


System describers create and use abstract system descriptions which idealise concrete systems.

A concrete, running, system is made of individual things, somewhat different from each other, which realise/instantiate types in an abstract system description

Conversely, an abstract system description is made of types that idealise similar things by naming properties they instantiate.


This table lists several applications of system theory.

General system theory

System describers

Abstract system descriptions

Concrete system realisations




“The Dewey Decimal System”

which idealises

Sorting books on library shelves




“The Solar System”

which idealises

Planets in orbits

Lawn tennis



“The laws of tennis”

which idealise

Tennis matches

Classical music



Symphony scores

which idealise

Symphony performances

Business systems

Enterprise architects

create and use

Business models

which idealise

Business actors and operations

Software systems

Software architects

create and use

Software models

which idealise

Software objects and operations


Ashby said in 1956: “our first impulse is to point at a concrete entity repeating a behavior and to say "the system is that thing there".

But to apply system theory is to select and describe those behaviors of an entity that are relevant to some given aims(s) and/or concern(s).

The describer must abstract from the infinite describable facts that could be found in observing the activities of the concrete entity.

And notice, moreover, that our concern is activity systems rather than passive structures.


"Cybernetics does not ask "what is this thing?" but ''what does it do?" It is thus essentially functional and behavioristic.”

“[It] deals with all forms of behavior in so far as they are regular, or determinate, or reproducible.” Ashby 1956


Abstraction in system thinking

This table specialises our description theory for systems.

System description theory

Abstract system descriptions

<create and use>                            <idealise>

System describers <observe & envisage> Concrete system realisations


This table outlines four examples.






Abstract system description

“Solar system”

The laws of tennis

The score of a symphony

The roles in a radio play

Concrete system realisation

Planets in orbits

A tennis match

A performance of that symphony

Actors playing those roles


By necessity, a system description abstracts from most aspects of what it describes.

It is impossible to describe the infinite potentially describable feature of a concrete system realisation.

E.g. A definition of the solar system says nothing about life on earth.

E.g. The score of a symphony says nothing about the personalities of orchestra members.

E.g. Human role definitions say almost nothing about the individual actors that play those roles.


Remember our concern is activity systems,

The next table subdivides the concrete system realization into the specified behavior and the entity that performs it.


System thinking levels

System 1

System 2

Abstract system description

The specification

A symphony score

The design-time code of a computer program

Concrete system realisation

The specified behavior

A performance of the above

A run-time execution of the above

The entity that performs it

The orchestra members in a concert hall.

A computer in a data centre.


Remember, one entity can realise many systems.

E.g. one computer can perform many unrelated programs at once

E.g. you (as a person) may realise at least two of the three systems in the table below at the same time (three at once is not advisable).



System thinking levels

System 3

System 4

System 5

Abstract system description

The specification

“Oxygen to carbon dioxide”

“Gene reproduction”

“System modelling”

Concrete system realisation

The specified behavior

Person breathing

Person making love

Drawing ArchiMate diagrams

The entity that performs it

A person – you for example


Remember Ashby’s warning: “our first impulse is to point at a concrete entity repeating a behavior and to say "the system is that thing there".

An entity is rightly called a system only when and where it conforms to a system description.

No entity is rightly called a system without reference to a specific system description or model – one to which the entity demonstrably conforms.


The interaction between system describers, descriptions and realisations is especially intimate in the case of human and/or computer activity systems.

If you consider “friendliness” to be an important property of your run-time system, then you should put it in your role definitions.

The only features of the real-world entity that count as part of the system are the features included in your system description.


Concretion of system realisations

In practice, it is normal to regard the concrete entity as part of the system realisation.

Because the remainder of that entity, and whatever it does outside the system of interest, is out of scope.


The describer’s premise is that a concrete system will realise (gives values to) variables in an abstract system description.

For example, the members of an orchestra will give values to the notes in a musical score.

We test this by observing and measuring the values the concrete system gives to those property types.

To run at all, the system must employ actors who can read the types in a description and assign values to them. 


Level of abstraction?

The level of abstraction in any system architecture is a matter of choice.

The architecture of a computer program can be described at several levels, from program code up to very abstract architectural models/diagrams.

The “system of interest” to a system architect extends only so far as the architect chooses to describe the structures and behaviors of that system.

For more on abstraction of description from reality, read A philosophy.

Conclusions and remarks

If every entity we name is a system, then the term has no meaning.


E.g. Some glibly refer to a business, such as IBM, as a system.

In system theory terms, IBM comprises infinite describable systems, and much that is not systematic.

The system that produces IBM's annual report proceeds at a level of abstraction disconnected from most individually identifiable actors and activities in ground-level operations.

Some business operations don't interact, some overlap, some compete with each other.

Much employee behavior is beyond the bounds of a describable system.

Much is ad hoc, unpredictable and down to the personal motivations and characteristics of individual humans.


Some say that all adds up to "complex system", but this is to abuse the term system.

Really, it adds up to a "complex entity" that features complex human beings working to their own ends, and some systems can be usefully discerned and described.


Business systems and software systems can be seen as products of the biological evolution that created social animals.

The later papers on general system theory treat both as formalised versions of social systems.


This paper is one in a family of related papers that precede discussion of system theory.

1.      The nub of our philosophy

2.      Introduction (which leads to How the brain works)

3.      A communication theory.

4.      A language theory.

5.      A description and type theory (which leads to Realism or Idealism?)

6.      A philosophy (which leads to Other triangular philosophies)

7.      Knowledge and truth



All free-to-read materials on the Avancier web site are paid for out of income from Avancier’s training courses and methods licences.

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