“Self-organising systems”

And third order cybernetics

Copyright 2017 Graham Berrisford. One of about 300 papers at http://avancier.website. Last updated 18/02/2019 03:21

 

System Theory Tutorial in London Saturday March 2nd 2019

 

Systems thinkers speak of self-organising systems, but with two very different meanings.

 

·         Circular causal loops

·         Recursive system definition

 

The second is the basis of what is here called “third order cybernetics.”

Whereas second order cybernetics tends to undermine classical cybernetics, third order cybernetics preserves it.

Contents

Recap. 1

System state change by circular causal loops. 4

System mutation by recursive system definition. 5

Third order cybernetics. 6

So, what is a self-organising system?. 9

Concluding remarks. 9

Footnote: Why did Ashby say that?. 10

 

Recap

Systems (recap)

In “systems thinking” discussions, people often refer to a named entity (say IBM) as a system.

What this means is often unclear or debatable.

It might mean merely that all IBM employees are related via their employment by IBM.

W Ross Ashby urged us to recognise that looking at IBM in different ways can reveal different (even conflicting) systems.

 

Inside a system of interest to us there are usually:

·         active structures or actor(s) that perform

·         behaviors or activities that modify

·         passive structures that form the state (material and/or information) of the system.

 

The actors and activities are orderly in the sense that they conform to some roles and rules.

The roles and rules can be described, and the conformance of a system’s behavior to the rules can be assessed.

 

In his “Introduction to Cybernetics”, Ashby spoke of what is called a Discrete Event-Driven System (DEDS).

In Ashby’s work, there is:

·         An abstract system of rules for responding to events.

·         A concrete system of physical entities that realises the abstract system.

·         State variables whose values are changed in response to events.

 

This table summarise the distinction between abstract and concrete systems.

 

Abstract system

A set of roles and rules (the logic or laws entities follow)

E.g. The stickleback mating ritual.

Concrete system

Entities playing the roles and acting according to the rules

E.g. A pair of sticklebacks.

 

One abstract system can be realised by many concrete systems

Each concrete system contains actors playing the roles, following the rules, and acting on objects or variables.

And note that the same actors may act in other systems. E.g. sticklebacks play a role in the digestion system of kingfishers.

System Dynamics (recap)

Ashby’s discrete event-driven system is transformable into one of the kind modelled in Forrester’s System Dynamics.

The trick is to abstract from singular entities to populations of them (stocks), and from singular events to batches of them (flows).

In System Dynamics, we have:

·         An abstract system: a definition of stocks and inter-stock flows.

·         A concrete system: an animation of the flows impacting the stocks.

·         System state: stock populations that increase and decrease in response to flows.

 

A System Dynamics model is closed, a self-contained model of stocks and flows.

However, each stock may be seen as a system or subsystem in its own right, and the whole model may be regarded as an ecology.

 

In discussing System Dynamics, it is important to distinguish theory, animated theory and reality.

 

Abstract System Dynamics model

A theory: a model of stocks and flows that interact according to fixed laws

Concrete System Dynamics model

An animated theory: a performance of actions precisely according to the laws above

An ecology in the real world

Actors that interact to a greater or lesser extent according to the laws above

 

When a model of a human social system is animated, the actors robotically obey the roles and rules (modelled as stocks and flows).

In the real world equivalent, human actors may ignore or disobey those roles and rules.

They may spend most of their time acting outside of the modelled system and/or playing roles in other systems, even ones with conflicting rules.

Which to say, there may be a mismatch between theory and reality.

Social systems (recap)

Russell Ackoff wrote about social systems, as he saw them.

In his “System of System Concepts” Ackoff spoke of:

·         An abstract system: a system in which the elements are concepts.

·         A concrete system: a system whose physical objects realise an abstract system.

·         System state: the values of a system’s properties at a particular time.

 

However, Ackoff was inconsistent; he started off distinguishing abstract systems from concrete systems.

Later, he spoke of a human organisation as being a system regardless of any abstract system description.

This confusion of a human network with a system runs through much systems thinking discussion.

This table is an attempt to separate three concepts that have become entangled.

 

Abstract social system

A set of roles and rules (the logic or laws actors follow)

Concrete social system

Actors playing the roles and acting according to the rules

Human network

Actors who inter-communicate and act as they choose

 

A human network is usually both less and much more than any given system.

·         Less in that it does not implement the system perfectly, so the concrete system only approximates to the abstract system.

·         More in that it does very much more than is described by the abstract system, and may realise several other systems.

 

Obviously, human actors are much more than the “parts” of a system

They live, breath, speak and otherwise act outside of any system that could ever be described.

The say and do things that cannot possibly be regarded as systematic, and sometimes defy any given rules.

One actor can belong to several human networks.

And the actors in one human network may realise several distinct activity systems.

Some are complex, others simple; some are adaptable, others inflexible

Some are purely social, others socio-technical; some are cooperative, others in conflict.

 

By the way, a social cell is a social network in which actors find that realising the roles and rules of an abstract system (in a concrete system) is attractive.

So attractive that they resist any change to the abstract system.

System change (recap)

In System Dynamics, we set the system in motion then observe how inter-stock flows change the volumes of stocks over time.

But the system cannot change its very nature; only the modeller can change the inter-stock flows, add or remove stock types.

To change the rules is to change the system itself, in a discrete evolutionary step, from one generation to the next.

"Change the rules from those of football to those of basketball, and you’ve got, as they say, a whole new ball game.” Meadows

 

Similarly, Ashby considered it essential to distinguish two kinds of system change.

“The word "change" if applied to [an entity repeating a behavior] can refer to two very different things.

1.      change from state to state, which is the machine's behavior, and which occurs under its own internal drive, and

2.      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.” Ashby 1956

 

In other words, Ashby insisted we should on no account confuse two kinds of change.

1.      System state change: changing the values of given state variables

2.      System mutation: changing the state variable types, or the rules that update their values.

 

An earlier paper explored four varieties of system change with reference to Ashby’s ideas.

 

System change

System state change

System mutation

Discrete

system state change

Continuous

system state change

Discrete

system mutation

Continuous

system mutation?

 

Classical cybernetics and System Dynamics are about system state change.

The term adaptation is used to mean homeostatic state change – regulating the values of state variables.

Second-order cybernetics is about system mutation

The term adaptation is used to mean changing the state variable types, or the rules that update their values.

Moreover, the system is self-organising.

 

Corresponding to the two kinds of change, systems thinkers speak of self-organising systems with two very different meanings.

·         System state change by circular causal loops

·         System mutation by recursive system definition.

System state change by circular causal loops

Ashby said a system's behavior “occurs under its own internal drive”.

In a homeostatic system, two subsytems may interact such that each keeps the state of the other within certain bounds.

This kind of self-organisation has been recognised in biology for c200 years.

Cybernetics and System Dynamics are also much about circularity.

They feature causal or feedback loops that may act to dampen or stabilise system state change.

As for example in a solar system or a weather system.

 

System Dynamics gives us a way to model a system as a set of laws governing “stocks” and the “flows” between them

A causal loop is formed when two stocks are connected by flows in both directions

A causal loop may have a balancing/dampening or reinforcing/amplifying effect on the populations of the stocks

 

Setting the model in motion reveals the trajectory of quantitative state changes over time.

The animation may reveal that the state of a system

·         changes in a linear or non-linear, orderly or chaotic, manner.

·         changes continually in one direction, or oscillates back and forth.

·         settles into a steady cyclical pattern or state (as in a homeostatic system)

·         periodically moves from one steady state to another (as a weather system or solar system may do).

 

Steady and periodic states may be “attractive” - meaning the system, when in a nearby state, likely moves towards them.

Thus, the system appears to be “self-organising”.

 

Nothing said above implies changing the system’s behaviour or state variables.

Whatever the state change trajectory turns out to be, it is an inexorable result of behaving according to given rules.

While the state of a weather or solar system may change in orderly or chaotic was, the laws of physics do not change.

This first “self-organisation” is an inexorable result of the system’s laws; it does not change those laws.

System mutation by recursive system definition

The question to be addressed is how the rules of a system are defined and changed.

Ashby said: “a change of its way of behaving occurs at the whim of the experimenter or some other outside factor”.

·         Planets don’t define the roles and rules of a solar system – the laws of physics do that.

·         Termites don’t define the roles and rules of termite nest – a process of reproduction and DNA do that.

·         Tennis players don’t define the laws of tennis – the Lawn Tennis Association do that.

 

Think of systems layered on top of each other, at different levels of abstraction.

There can be a hierarchy of process control: a control system at level N throws an exception up to a control system at level N+1, and awaits direction.

Here, a system at one level can be redirected by a higher level system.

There can be a hierarchy of system definition: the rules of a system at level N are state variables that can be manipulated by a system at level N+1.

Here, the system at one level can be redefined by a higher level system.

 

In an enterprise, the two kinds of hierarchy may be combined in the sense that one management body plays both roles.

Or else, line-of-business managers direct processes and some kind of “enterprise architecture” function defines processes.

 

Thus the roles and rules of one system are shaped and changed by a higher process or meta system.

Importantly, one actor may alternate between a role in a base system and a role in a meta system.

After all, the actors in a social system have insight into it, and may have a choice about how it works.

So, they can act as a system definer, who changes the roles and rules of any social system they play a role in.

But one action is in one or the other system – not in both.

 

This second kind of “self-organisation” changes the very nature of the system; it changes its laws.

By successive changes of this kind, an entity or human network (say IBM) can realise many generations of a system over time.

 

Why must system mutation be incremental?

If the rules of a game change in a continuous and unpredictable manner, this undermines the very concept of a game.

When the rules change - how do all players get to hear of them? Who tells them to start using the new rules?

Can some players use the new rules while others are still using the old rules?

Down this road you have a disorderly situation or disorganised entity, not a system at all.

 

Certainly, the actors in a social group can change the roles of a system they play roles in.

But you cannot change the laws of tennis while you are playing a rally.

You have to stop the game, agree new laws, and restart the game.

To maintain the integrity of the system concept we must insist its rules are changed incrementally – generation by generation.

Because if the rules change continually, there is never any describable or testable system, and to call the entity a system is meaningless.

There is instead a disorderly situation or disorganised entity.

Third order cybernetics

Third order cybernetics presumes a base system does not organise itself.

It distinguishes system at two levels.

·         In a base system, actors advance the state of the system according to roles and rules.

·         In a meta system, actors define or change the roles and rules of the base system.

 

Meta system examples outlined below are:

·         The Lawn Tennis Association (define the roles and rules of tennis matches)

·         A Constitutional Convention (define the roles and rules of US governments)

·         A daily stand up meeting (define the roles and rules of software development projects)

·         An enterprise architecture function (define the roles and rules of business systems)

·         A cooperative (define the roles and rules of actors sharing a resource)

·         Sexual reproduction (define the roles and rules of cells in organisms).

 

The definition of a base system is the state of a meta system.

So, every mutation from one base system generation to the next represents a state change in the meta system.

Each mutation adds a new member to the set of systems generated by the meta system.

Where the actors are human, there is no constraint on the variety of systems producible by a meta system.

 

Importantly, an actor may play a role at both levels – a role in a base system and a role in a meta system.

An actor can switch between following rules in a base system and defining rules a meta system.

However, one action is in one or the other system – not in both.

 

Meta system examples

A meta system acts to define another system or transform it from one generation to the next.

 

The Lawn Tennis Association

The base system a tennis match, in which tennis players act to advance the score - rather than to advance the laws of tennis.

The meta system is the Lawn Tennis Association that acts to advance the laws of tennis, the definition of a tennis match.

 

A Constitutional Convention

The base system is the US government, whose behavior is constrained by the US constitution.

The meta system is the processes of any constitutional convention by which the US constitution may be amended

To make a change, people may step outside their role in the base system and into the meta system.

 

A daily stand up meeting

The base system is the processes followed by software development team.

The meta system is the “daily stand up meeting”, which defines and refines those software development process.

To make a change, people step outside their role in the base system and into the meta system.

 

Agile development

Software development practices

<define>                                     <idealise>

Daily stand up meeting    <observe and envisage>    Software development

 

An enterprise architecture function

The base system is a regular business systems.

The meta system is the enterprise architecture framework as it used by enterprise architects

 

Enterprise architecture

Business system descriptions

<define>                                       <idealise>

Enterprise architects  <observe and envisage>  Business systems in operation

 

A cooperative

The base system is a group of people who share access to resources.

Such as fishermen who share fishing grounds, or farmers who share an irrigation system.

How to avoid “the tragedy of the commons” by which competition exhausts the common resource?

The meta system is the cooperative in which the fishermen or farmers agree their rules.

For a while, the fishermen must stop fishing, and farmers stop farming, to define the rules of their social system.

(Elinor Ostrom (1990, 2010) defined eight generic conditions for such a cooperative.)

 

Sexual reproduction

The base system is the cells in a biological organism, whose behavior is constrained by its DNA.

The meta system is the sexual reproduction processes by which that organism mutates in the next generation.

To make a change, an organism steps outside its day-to-day role as a living entity and into a role in the meta system.

 

(As Ackoff noted, biological analogies can be misleading.

One might say evolution is the general process via which entity N mutates into entity N+1.

Sexual reproduction is a subtype of evolution, and a meta system that acts on entity descriptions.

The meta system manufactures a description of a new entity by merging the descriptions of two mature entities.

Mating is the process in which two entities realise that meta system.

Separately, each of those entities acts in potentially infinite different systems.

That is Ashby’s point; to call something a system with no perspective, no system description, is meaningless.)

 

Meta meta systems

A meta system may defined and re-defined by an even higher level meta meta system.

 

In a base system

actors advance the state of the system according to roles and rules

a game of golf.

In a meta system

actors define or change the roles and rules of the base system

the committee of the club whose course is used to play the game.

In a meta meta system

actors define or change the roles and rules of the meta system

the R&A – who define activities of golf club committees.

 

https://www.randa.org/en/rog/2019/pages/committee-procedures

“Committee Procedures contain practical guidance for those involved in running day to day play at golf courses

Section 8 also provides Model Local Rules that the Committee can adopt to meet local needs.”

So, what is a self-organising system?

A base system does not organise itself; nor does a meta system.

Shall we say a self-organising system = one or more base systems + a meta system?

The question challenges our understanding of the concepts, how we think of them and name them.

 

Does the sum of all computer game playing, games design and distribution = a self-organising game industry?

Does the sum of all organisms + sexual reproduction = a self-organising biomass?

Does the sum of all tennis matches + the Lawn Tennis Association = a self-organising world of tennis?

Does the sum of all business systems + enterprise architecture = a self-organising enterprise?

 

Arguably yes, to the extent that the base systems depend on the meta system.

And the named aggregate does nothing but what is described in those systems.

 

Arguably no, to the extent that the base systems do not depend on the meta system.

And the named aggregate is an ecology or human network that does more than what is described in those systems.

E.g. business systems are created outside the remit of an enterprise architecture team using an enterprise architecture framework

And the enterprise is an ecology or human network that does much more than what is described in its business systems.

 

We might speak of the enterprise as a self-organising ecology or human network.

Perhaps a “resilient adaptive self-organising system” would better be called a “resilient continuously evolving ecology or human network”?

Concluding remarks

This paper advances “third order cybernetics.”

Whereas second order cybernetics tends to undermine classical cybernetics, third order cybernetics preserves it.

It treats the description of one (base) system as the state of a meta system.

In a base system, actors advance the state of the system according to roles and rules.

In a meta system, actors define or change the roles and rules of the base system.

 

Importantly, one actor may alternate between a role in a base system and a role in a meta system.

But one action is in one or the other system – not in both.

 

Much so-called "systems thinking" is to do with the management and/or leadership of human networks, and/or the democratisation thereof.

Arguably, it is better called sociology or management science rather than system theory.

Some is more human network thinking than systems thinking.

A satisfactory system theory has to distinguish the three concepts in this table.

 

Abstract social system

A set of roles and rules (the logic or laws actors follow)

Concrete social system

Actors playing the roles and acting according to the rules

Human network

Actors who inter-communicate and act as they choose

 

The next paper questions what the term “complex adaptive system” means.

 

 

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..