“Complex
adaptive systems”
Is IBM a Complex Adaptive System or a Complex Evolving Entity?
Copyright 2017 Graham Berrisford. One of about 300 papers at
http://avancier.website. Last updated 11/09/2019 11:49
The term Complex Adaptive System (CAS) is a strange compound term.
It is used where the meanings of its component terms are ambiguous or not defined.
This paper explores ambiguities and incoherencies in the use of the term.
Contents
A
first definition of “complex adaptive system” – from the Sante Fe institute
A
second definition of “complex adaptive system” – from MIT
Read social systems thinking discussion and you will come across “complexity science”, “complexity theory” and “complex adaptive systems”.
Authors use these terms glibly, without being clear what they mean.
What is a complex adaptive system? How are complexity and adaptivity defined and measured?
What kinds of adaptation or change are possible? How is a change measured? How is the ease of change measured?
As in so much systems thinking discussion, there are ambiguities .
Complex?
In cybernetics, a system is complex if the
system description is complex; the
roles and rules are complex
To social systems thinkers, a system is
complex if the reality is complex,
the actors are complex (their roles and rules may be lightly prescribed, if at
all.)
Adaptive?
In cybernetics, a system adapts to feedback from its environment
by changing state – which may be called self-regulating.
To social systems thinkers, a system mutates as actors change its roles,
rules or aims - which may be called self-defining.
System?
In cybernetics and system dynamics, a system
is a collection of repeated or repeatable activities.
In social systems thinking, a system is a
collection of actors, who interact as
they choose (a social network).
In his 2003 book, Michael Jackson proposed as follows.
"Social systems are not just ‘complex adaptive systems’ bound by the fixed rules of interaction of their parts.
Rather, they are ‘complex evolving systems’ that can change the rules of their development as they evolve over time."
He could more clearly have written:
“Social networks are not systems, in which actors are bound by roles and rules
Rather, they are “complex evolving entities” that can change the roles or rules of any system(s) they realise.”
A complex evolving entity (or ever unfolding
process) - in which no behavior is regular, or determinate, or reproducible –
is not a system in the ordinary sense of the term.
This table distills some
ambiguities.
|
Term |
School |
Meaning |
|
Complex |
Classical cybernetics |
The
measurable complication of an
abstract system description |
|
Sociological thinking |
The
un-measurable disorder or
unpredictability of a real-world situation |
|
|
Adaptive |
Classical cybernetics |
System
state change – updating the values
of system variables |
|
Sociological thinking |
System
mutation - changing the roles and
rules of the system (evolving) |
|
|
System |
Classical cybernetics |
Actors
playing roles and acting according to
rules |
|
Sociological thinking |
A
group of self-aware actors who inter-communicate and act as they choose, or a problematic
situation (an entity) |
|
|
Emergence |
Classical cybernetics |
A
property arising from coupling
subsystems into a larger system |
|
Sociological thinking |
Not
seen before, new, or surprising. |
The Sante Fe
institute’s courses (at complexityexplorer.org) look like a
wide-ranging exploration of “complexity” from different angles.
There is no question here about the excellence of these
as individual courses
The challenge here
is to the linking of “complex”, “adaptive” and “system” into a compound term.
The institute defines the three-part compound term thus:
“A
complex nonlinear, interactive system which has the
ability to adapt to a changing environment.
Such
systems are characterized by the potential for the emergence of new structure
with new properties.
Complex
adaptive systems (CASs) can evolve by random mutation, self-organization, the
transformation of their internal models of the environment, and natural
selection.
Examples
include living organisms, the nervous system, the immune system, the economy,
corporations, and societies.
In
a CAS, semi-autonomous agents interact according to certain rules, evolving to
maximize some measure like fitness to their environment.”
https://www.complexityexplorer.org/explore/glossary/418-complex-adaptive-system
This might make sense to
its author.
But it is impossible to tell;
because it is a jumble of ambiguous terms and disparate ideas.
“A complex, nonlinear,
interactive system which has the ability to adapt to a changing environment.”
What
is complexity? How is it
measured?
What
does non-linear mean? What is non-linear? The system’s processes? Or the
trajectory of the system’s state variable values?
What
does interactive mean? Inter-subsystem interaction? Or system-environment
interaction?
What
does adapt mean? The system’s state changes, as in homeostasis? Or the system
mutates, as in biological evolution by sexual reproduction?
Does the term CAS embrace both
SAS (simple adaptive systems) and CEE (complex evolving entities)?
What
changes in the environment? The states of the entities in it? Or the types of
the entities in it?
“Such systems are characterized by the
potential for the emergence of new structure with new properties.”
Is
the system a CAS before new properties emerge? If it never happens? Or only
after it happens?
What
kind of new structures and properties can emerge? New structures and properties
of the same type? Or of different types?
If
different types emerge continuously, there is never any describable or testable
system.
If
different types emerge in discrete step changes, it implies a new system has
been created from an old one.
“Complex adaptive
systems can evolve by random mutation, self-organization, the transformation of
their internal models of the environment, and natural selection.”
This
appears to embrace changes as diverse as state variable value changes, state
variable type changes, and the births of children
(For discussion of
self-organisation, see the conclusions below.)
“Examples include living organisms, the
nervous system, the immune system, the economy, corporations, and societies.”
This
confuses biology with sociology (a confusion deprecated by Ackoff and Bausch).
And
confuses the “real machine” with any “system” we say it realises (a confusion
deprecated by Ashby).
See
the section on IBM below.
“In a CAS, semi-autonomous agents interact
according to certain rules, evolving to maximize some measure like fitness to
their environment.”
In what sense are the organs of
the body autonomous at all?
If they are partly autonomous,
then in what sense is any subsystem of any simple system not also autonomous?
Does “according to certain rules”
mean the rules are fixed?
What is it that evolves? The
states of the agents? The types of the agents? Or the “certain rules?
If the agent types and rules are continuously changing, then how to distinguish the so-called system from chaos?
The definition of a CAS fits everything from the simplest homeostatic system to the most complex ever-mutating entity.
It aggregates different ideas; but aggregation is not generalisation; so the definition does not say what all CAS share.
It defines a chimera – part one thing, part another.
"Complex Adaptive Systems are dynamic systems able to adapt in and evolve with a changing environment.
It is important to realize that there is no separation between a system and its environment in the idea that a system always adapts to a changing environment.
Rather, the concept to be examined is that of a system closely linked with all other related systems making up an ecosystem.
Within such a context, change needs to be seen in terms of co-evolution with all other related systems, rather than as adaptation to a separate and distinct environment."
This definition doesn’t tell us what a “system” is, or how to distinguish it from any other kind of entity in the universe.
It uses the term “adaptive” in a way that may be interpreted to mean different things.
And it gives us no hint as to why or how any system, adaptive or not, should be considered “complex”.
Let us analyse the definition, dividing it into six clauses.
Several of the clauses are ambiguous, and all can be applied to simple systems.
"Complex Adaptive Systems are dynamic systems”
The term dynamic may be read as changing or changeable.
Simple systems are dynamic in the sense they change state in response to external events.
They are also dynamic in the very different sense that they can change via generational system mutation.
“able to adapt in and evolve with a changing environment.”
By adapt and evolve do MIT
mean state change and/or system mutation?
Again, does the term CAS embrace both SAS (simple adaptive systems) and CEE (complex evolving entities)?
“there is no separation between a system and its environment in the idea that a system always adapts to a changing environment.”
At a glance, this makes a nonsense of the system concept.
If a system has no boundary, you cannot distinguish it from its environment.
You cannot describe or test it, and to call it a system is an empty assertion.
Also, for a system adapt to a changing environment, the two must be connected.
Clearly MIT must allow that a system is connected to its environment – logically in description, and physically in reality.
The connection may be a uni-directional flow or a bi-directional feedback loop.
Perhaps what MIT mean is that the actors who play roles in the system also play roles outside the system?
(For discussion of self-organisation, see the conclusions section below.)
“a system always adapts to a changing environment.”
Again, does the term CAS embrace both SAS (simple adaptive systems) and CEE (complex evolving entities)?
And does mutation mean generational change or continuous change?
If a social network or other entity changes continually, it can never be described and tested as a system.
And so, to call it a system is an empty assertion.
Moreover, it undermines Bertalanffy’s first concept of a system, since
it appears disorganised rather than organised.
(For discussion of
self-organisation, see the conclusions section below.)
“the concept to be examined is that of a system closely linked with all other related systems making up an ecosystem.”
Systems are both decomposable and composable.
And the environment outside a given system of interest may be divided into separately describable systems.
Together, the given and external systems can be seen as subsystems interact in a wider system - an eco-system if you like.
“change needs to be seen in terms of co-evolution with all other related systems, rather than as adaptation to a separate and distinct environment."
Even the simplest open system, when it changes state, may trigger related systems to change state.
And when it mutates, related systems may have to mutate also, if they are to survive.
Again, not only are several of the clauses are ambiguous, but all can be applied to simple systems.
Systems thinkers often speak of a social network organisation as a CAS.
But
a system is less than the reality it describes.
It
is a selective abstraction that hides the infinite complexity of the entity
itself.
It
was suggested above that: “Social networks are not systems, in which
actors are bound by roles and rules
Rather, they are “complex evolving entities” that can change
the roles or rules of any system(s) they realise.”
E.g. IBM can be described as a social network
in which employees communicate.
It may be seen as realising a simple system
that pays suppliers and receives payments from customers.
It can also be described as realising
countless other - parallel,
disparate and sometimes conflicting – systems.
What does it mean to assert that IBM as a whole is one CAS?
Of four possible answers, the last two seem the best.
IBM as a whole is one system that is changed frequently under change
control.
This is a poor answer, because it confuses description and reality.
It confuses the entity (the social network) with the countless different systems it realises.
(In practice, it would be impossible to describe or design everything that happens in IBM as a system.
If we did it, the necessary change control and testing procedures would disable any attempt to change the system.)
IBM changes its
organisation/management structure frequently.
This may well be true, but is a poor answer to the question, because it does not necessarily change what IBM does.
The primacy of behavior is central to most general system theory.
IBM has a high-level meta system for changing what its actors do.
OK, this might be true.
IBM realises countless distinct business systems, some of them likely to be conflict with each other.
There may be a central enterprise architecture team, struggling to govern changes to these business systems.
This overarching meta system is small compared with the countless distinct operational systems.
IBM is a complex adaptive social network rather than a system.
OK, IBM is a society in which the individual human actors are inherently complex and adaptable.
The network of actors employed is large and complex.
They play roles in many distinct business systems, some more flexible than others.
Actors may change those systems in response to changes in the environment of – and related systems outside of - IBM.
There is complexity in the intelligence of IBM’s employees.
There is adaptability in their ability to work with little by way of fixed roles and rules.
Much of what they do is creative, ad hoc and not prescribed by IBM or any manager or system it employs.
To call IBM a CAS is to say it is flexible social network that is not describable and testable as a system.
It is complex evolving entity (CEE) rather than a complex adaptive system (CAS).
Social systems thinkers often discuss what they call Complex Adaptive Systems (CAS).
The trouble is, it isn't clear they agree
· why they call thing they are talking about a system
· why they call it complex, or how they could measure that
· in what ways they expect it to adapt and
· when if ever they would consider an adaptation to have changed the system into a different one.
People use the term CAS to label entities that are unstable, disorderly and unpredictable.
That is, to describe things that are not systems in a natural language or general system theory sense.
Self-regulation (as in homeostatic organism or machine) maintains variable values within a desired
range.
This kind “self-organisation” is an inexorable
mathematical result of a system
following the rules that define it.
By contrast, generation-to-generation system mutations change the variables, roles or rules of a
system
By using adaptation to mean mutation, social systems thinkers have undermined the value of the system concept.
For sure, a
concrete system may adapt to its environment by changing state, but a mutation makes it a different thing.
An ever-mutating entity (or ever-unfolding process) in which no behavior is regular, or determinate, or reproducible – is not a system in the ordinary sense of the term.
System theory gurus Ashby and Maturana rejected the concept of a self-defining system; said it makes no sense, it is impossible.
System designers do apply the principles of general system
theory and cybernetics to design role-and-rule-bound systems
But the system designers (be they actors in the system or not) themselves are a social network that are more goal-bound than rule-bound.
How to extend general system theory to embrace “self-organisation”?
Read second order cybernetics for the start of an answer to that question.
If your interest is the application of these ideas to agile
architecture, try https://bit.ly/2UKnE8e.
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..