Change: adaptation, evolution and self-organisation

This page is published under the terms of the licence summarized in the footnote.

All free-to-read materials at 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 in whichever social media you use.


Stuff exists and stuff happens, for a while, but not for ever.

Heraclitus of Ephesus was a Greek philosopher known for his doctrine of change being central to the universe.

Plato quoted him as saying “Everything changes and nothing stands still.


Some socio-cultural systems thinkers have suggested that a system should be “adaptive, self-organising and resilient”.

And the term “agile (enterprise) architecture” has been bandied about for some years.

This paper explores what these things mean, or could mean, and exposes (again) the widespread confusion between actors in social groups and roles in social systems.


Preface. 1

Four kinds of system change. 2

State (adaptive) change. 2

Generational (evolutionary) change. 3

More about discrete step evolution. 4

Change in social groups. 5

Adaptive. 5

Self-organising (not self-sustaining) 5

Resilient 5

Can an enterprise be adaptive, self-organising and resilient?. 6

What does Agile Architecture mean?. 7



From “An Introduction to Cybernetics” W. Ross Ashby M.A., M.D.(Cantab.), D.P.M. Director of Research, Barnwood House, Gloucester

First published1956 by Chapman & Hall; later in New York, by John Wiley and son. 440 Fourth Avenue; and by William Clowes and Sons, London and Beccles.


It will be seen that the word "change" if applied to such a [determinate] machine can refer to two very different things.

·         change from state to state, which is the machine's behaviour, and which occurs under its own internal drive, and

·         change from transformation to transformation, which is a change of its way of behaving, and occurs at the whim of the experimenter or some other outside factor.

The distinction is fundamental and must on no account be slighted.


By “change from transformation to transformation” Ashby meant what is called “generational change” below.

Four kinds of system change

The term analogue is used to describe signals or data that vary continuously, as the hands on a clock face do.

The term digital used to describe signals that are chunked into discrete units, as you read from a clock that displays discrete numbers. 

The distinction between analogue and digital signals is related to the distinction between continuous state change and discrete state change.

EA is built on the premise that business operations can be modelled in terms of discrete event-driven state changes.


There is another way that change can be divided into two kinds.

State (adaptive) change is a change to the state of a system, within one system generation; it doesn't change the nature or identity of the system.

By contrast, generational (evolutionary) change is a change to the rules (or DNA) of a system.

Changing the nature or identity of the system is to produce a new system, or new system generation.

EA plans changes that move a system from one generation to the next.


Table 3 presents Enterprise Architecture as being about discrete change rather than continuous change.

Table 3: Four kinds of change

Discrete change

Continuous change

State (adaptive) change

Enterprise Architecture (EA)

System Dynamics

Generational (evolutionary) change

Enterprise Architecture (EA)

No describable system


EA presumes a system is under change control, which implies an old system is replaced, in a discrete step, by a new system.

By contrast, if actors continually change their roles and rules, then the notion of a describable system crumbles.

State (adaptive) change (1st order cybernetics)

Ashby’s concern was homeostatic organisms and machines that maintain property values within ranges fitting to survival.

In this context, adaptive means self-sustaining; a self-sustaining individual gathers resources it needs and adapts to its environment.

Actors/components of the operational system follow rules to act on inputs to maintain system state - but do not change those rules.


Continuous state change

State changes in weather systems and biological entities are continuous.

But you can convert a continuous system model into a discrete event system model.

The basic trick is to divide time into discrete events (seconds, hours, days, whatever time unit is appropriate).

And by using this trick you can build a digital system to simulate an analogue one.


Discrete state change

“We shall assume that [system] changes occur by finite steps in time; any difference is also finite;

the change occurs by a measurable jump, as the money in a bank account changes by at least a penny.” Ashby


EA is similarly concerned with business systems connected to their environments by information feedback loops.

A business must know the state of those entities and activities in its environment that it monitors, supports or directs

It learns of changes to the environment and customer requirements through feedback - the arrival of new inputs.

It responds to inputs and/or state changes by sending outputs to inform or direct entities and activities in its environment.


Suppliers   Customers

v          ^



Inputs     Outputs

v          ^


performed by actors

according to values of

state variables


Thus a business system continually adapts a business to changes in its environment, according to business rules.

Each adaptation changes the state of a system and/or produces outputs that change the state of the environment.


All EA frameworks are based on two presumptions that few make explicit.

Happily, the UML 2.1 standard does make them explicit:

·         “all behavior in a modeled system is ultimately caused by actions executed by so-called “active objects”

·         “behavioral semantics only deal with event-driven, or discrete, behaviors” 


Adaptive state changes in formalised social systems (including business systems) happen in response to discrete events.

EA treats the enterprise as a system in which the state of the business changes in response to discrete events.

This is what enables business roles and processes to be “digitised”.

In fact, I believe you cannot digitise a system that you cannot model as a Discrete Event Dynamics Systems (DEDS).


Reader: Oh my god, so this is what digitization implies!

Digitization occurs at the conceptual level, not just the physical level.

And part of what EA does is to determine what is digitizable.


That is an interesting way to put it.

I'd rather say that the world out there is fuzzy.

In our perception, memory and description of the world we divide it into discrete entities and events.

That is what enables us to recognise things we perceive as being similar to things we remember.

The discreteness of things is assumed in the natural language we use in social transactions to describe entities and events.

It is formalised in the data types we use in the more formal communications of business transactions.

And this discreteness is what enables us to a) model systems as composed of discrete entities and events and b) digitise them.


Moreover, governing an operational system against a system model has an implication at a higher level for the discreteness of system evolution.

Generational (evolutionary) change (2ndt order cybernetics)

Evolution is a process in a meta system by which an individual entity’s nature is changed, or it is replaced by a different entity.


Continuous evolutionary change

This means the properties of an entity or social group, its very nature, changes continually.

If an entity evolves continually, if its properties are ever-changing, then its properties can be little or not all described.

You may name it, you may know the date it was born, and perhaps some other properties.

But if you cannot define the properties that make it a system, then it cannot be mapped to a system description or be regarded as a system.


Human actors have personal goals or purposes outside of a group they belong to. (These conflicts of interest may prove a hindrance to the group.)

A social group may have few rules and/or its members may act outside of roles given to them. (This flexibility may prove helpful to the group.)

In so far as the roles and rules of a group’s members can be described, there is a system that can be placed under change control and governed.

Beyond that, where behaviour cannot be described, there is no system that can be placed under change control or governed.


All businesses and human social groups are in continual flux.

Change control does not prevent continual change to all the actors and activities of a business.

But it does prevent continual to change to those roles and rules that are under the governance of EA.


Discrete step evolutionary change (and the meta system that drives it)

This means an entity evolves in discrete steps, from one generation to the next.


Biological species change in discrete steps, as each child is born.

There is a meta system in which environment forces determine which adults get to produce children.

Evolution favours changes that increase the efficiency, effectiveness, suitability or survival of children.

When a child is born, it is given a new name.


Business systems also change in discrete steps, under change control.

There is a meta system in which change requests steer enterprise architects to define a new system generations.

Architects look to make changes that increase the efficiency, effectiveness, suitability or survival of the business.

When exactly does old system A become new system B?

Given a system with thousands of variable types and rules, changing a few of them is barely noticeable, and we assume continuity of identity.

Rather than call it by a new name, we give it a new version number.

More about discrete step evolution

A biological species or social group evolves when individual members are replaced by somewhat different, better-adapted, members.

“Since offspring tend to vary slightly from their parents, mutations which make an organism better adapted to its environment will be encouraged and developed by the pressures of natural selection, leading to the evolution of new species [eventually] differing widely from one another and from their common ancestors.”


Members who join a social group tend to be different from their predecessors

Changes to group members that make a social group better adapted to its environment will be encouraged by the pressures of natural selection.

This leads to the evolution of new, different, social groups.

This kind of evolution is not continuous, it happens in discrete steps, as new members join and older members leave.


New business systems tend to be different from their predecessors.

Changes to business systems that make an enterprise better adapted to its environment will be encouraged by the pressures of natural selection.

This leads to the evolution of new, different, enterprises.

This kind of evolution is not continuous, it happens in discrete steps, as new business systems are deployed and older ones are decommissioned.


Evolution can be seen as a meta system that replaces individuals by different individuals (better suited to their environment), and so changes a social group or species.

Enterprise architecture can be seen as a meta system that replaces business systems by different systems (better suited to their environment), and so changes an enterprise.

Note that IBM as a system today is different from IBM as a system yesterday.

The named entity has adapted to circumstances by replacing and radically changing the business systems it uses (and sells).

Change in social groups

People talk about social groups needing to be adaptive, self-organising and resilient.

Is this true of all social systems? And what does that mean anyway?


The table below shows a classification of social “system” types used by some systems thinkers.

A so-called “complex adaptive system” is partly an ordered system and partly a chaotic one.

Social system type

The behaviour of actors

It is

Ordered system

is fully constrained to follow the given rules.

fully describable

Complex adaptive system

Or self-organising system

is constrained by some given rules,

but actors work flexibly outside those rules

partly describable

Chaotic system

is unconstrained; there is no system

not describable


An ordered system in which the behaviour of human actors is constrained to follow roles defined by the rules of the system, is described somewhere.

Enterprise architecture is much about the standardisation and integration of systemisation of orderly business roles and processes.


A chaotic system, in which the behaviour of human actors is unconstrained, cannot be described as a system

Though the way actors interact and behave might be susceptible to some kind of analysis, statistical or other.


In a so-called complex adaptive system, the system actors are allowed or encouraged to act as system architects.

The shapes and rules of operational structures and behaviours are continually changing.

Only a simple description - showing “core” roles and processes that are stable and repeatable – is possible.

The system shows those roles and processes, and hides what can change.


Adaptive (evolutionary)

In this context, adaptive means modification of a system through what is called evolutionary change above.

A system that fits its environment well may last unchanged for years, decades or centuries.

But clearly, humans live in a rapidly changing social environment.

And a social group may be empowered to change the roles and rules of a social system.

But to say the social system is adaptive is to confuse the social group with the social system.

The actors in the social group adapt the roles in the social system; the system does not change itself.


Human actors can play a role in a system and a role its meta system at the same time.

·         As system actors, they can perform processes in a social system.

·         As systems architects, they can observe, redesign and change how things are done on the fly.


This table classifies social groups according to what system actors are allowed to change.

Social group type

Actors can change system goals

Actors can change processes and/or

organisation structure





A proper system; its structure and behaviour can be described as per general system theory




See below




Not a system; it cannot be described as a system, because all its properties are in flux


What about the “goal-driven social group”?

It has actors and goals – you might be able to measure the achievement of the goals.

But its roles or rules can change continually, and so there is no describable system

Actors may work in an ad hoc way, may do nothing, may not co-operate or even undermine others’ efforts.

Actors may decide to outsource the achievement of the given goals to a different group of actors.


However, the classification above is naïve, since a social group may well be a hybrid of types.

It may be partly ordered, partly goal-driven and partly chaotic.

In so-called “complex adaptive system”, only the ordered part is really a “systemthe rest is just actors doing things.



This is not a generally-required system property,

An oyster, a motor car, an ERP or CRM system is composed of parts that play their given roles.

These systems do not organise themselves, their organisation is changed by discrete step evolution from one generation to the next.

And the reorganisation happens in a meta system rather than in the system itself.


If customers, suppliers or employees ignore or change the roles or rules of a business system, this is disorganising rather than self-organising.

To abandon roles and rules is to abandon the notion of systematic behaviour; it undermines the notion that there is a system in any useful sense.

Hence, the term “systems thinking” is ill-fitting when applied to social group with no stable roles or rules.


Is a golf club self-organising?

Its committee is a meta system that maintains the rules of the golf club.

One person can play a role in the meta system (as a committee member) and in the operational system (as a player).


Is an agile system development project a self-organising team?

The daily stand-up meeting (that sets or changes the roles and priorities of team members) is better separated from the development work itself.

One person can play a role as a manager in the meta system that plans work, and as a developer in the operational system.


Is an enterprise with an enterprise architecture team self-organising?

The enterprise architecture team is not involved in performing day-to-day business operations.

Enterprise architecture is better regarded as a meta system to the business systems it observes, envisages and plans.


The systems we observe usually appear resilient, since systems that are not resilient soon disappear from view.

But still, resilience is not a generally-required property of a system.

The fact that a system is short-lived doesn’t mean it is no good, or no use.

And what does resilient mean anyway?


Resilient can mean an object will return to shape after being pressed out of shape.

It could mean homeostatic, maintaining system state within a range fitting to survival.

But social groups do not have to be homeostatic; they can grow, shrink or change direction.


Resilient can mean a system will withstand or recover quickly from damage,

Disaster recovery resources and processes are supposed to restore the same system that failed, not change it.


Resilient is better interpreted here as meaning a business (a legal entity) can change in response to unpredictable events.

It can respond to unpredictable changes in its environment, customers, suppliers, market or government regulations.

A biological species proves resilient when individual organisms are replaced by somewhat different, better-adapted, organisms.

An enterprise proves resilient when individual business systems are replaced by somewhat different, better-adapted, business systems.

Can an enterprise be adaptive, self-organising and resilient?

EA is about systems that involve many different kinds of actor/component, rather than a homogeneous population of identical actors/components.

EA is about designing macro-level systems that act to achieve given goals.

EA is activity-centric rather than actor-centric.

EA is about formalised social systems rather than informal social groups.


All enterprises are adaptive (evolving), self-organising and resilient to a degree.

Their flexibility is mostly down to the informal behaviour of human as individuals and in groups.

Obviously, you can employ capable, self-aware humans, give them some goals and free reign to act as they see fit.

That is the most flexible system you can devise, but it barely counts as a “system”.


Business systems are formalised social systems with defined roles and rules.

Defining and the required roles and rules in the first place is difficult.

The challenge of testing and deploying formal business systems is substantial.


A business system realises the roles and rules set out in a system description.

Changing the roles and rules of a business system creates a different business system.

The systems share some goals and properties, but are different systems versions or generations.

Changing a business system from one generation to the next can be as challenging as building it.

And designing systems so that they can be changed (within designed bounds) is even more challenging.

What does Agile Architecture mean?

This grand-sounding term has no agreed meaning.

Does it mean Guerilla EA? Agile architecture documentation? Architecting for a system to be agile?


Guerrilla EA

EA can be about making small steps towards the vision of digitised, standardised, integrated business processes and data.

“Guerrilla EA” does what it can, when and where it can, to these ends.

It may socialise solution architects into a group that cooperates by aligning solutions in the directions of standardisation and integration.

If Agile Architecture means that, then OK.


Agile architecture documentation (aka evolutionary design)

Of course you can work iteratively and incrementally, develop your architectural documentation alongside your system.

Years ago, Scott Ambler called this Agile Model-Driven Development, but his guidance is for software development teams.

Today it is called evolutionary design, but is it design at all?

Isn’t it agile documentation rather than agile architecture?

Surely, an enterprise architect must describe a system's structure and behaviour at a level of abstraction that is as stable as possible?

If that highest-level architecture is continually changing, doesn’t that undermine the notion of architecture?


Architecting for a system to be agile

The notion of an agile system is not a simple or single idea, since a system can be flexible in at least four ways.

A system that is


yet might also be

its architecture is likely to be


do many different things

rigid and hard to change or extend.

large or complex


be readily changed to a new version

limited and not versatile

slight or simple


be extended with extra features

not malleable (the Open-Closed principle in OO design).

fixed (must be right first time)


change its own roles and rules

reliant on human ability, barely a system at all

slight and simple


Building flexibility into a non-human system can be difficult and costly, and make the system more complex.

So, agile system architecture is challenging for all these reasons.


Related topics

Goals, Purposes and Choice

Adaptation and Evolution - Analysis for further exploration of system change.





Footnote: Creative Commons Attribution-No Derivative Works Licence 2.0       07/01/2016 16:01

Attribution: You may copy, distribute and display this copyrighted work only if you clearly credit “Avancier Limited:” before the start and include this footnote at the end.

No Derivative Works: You may copy, distribute, display only complete and verbatim copies of this page, not derivative works based upon it.

For more information about the licence, see