Systems thinking approaches

How systems thinkers came to eviscerate the system

Copyright 2017 Graham Berrisford. One of about 300 papers at Last updated 12/03/2018 11:15


This paper analyses a three-way classification of systems thinking approaches.

It challenges the classification and proposes an alternative view.


A three-way classification of system thinking approaches. 1

Class 1: “Hard” systems thinking approaches. 1

Class 2: “Soft” systems thinking approaches. 2

Class 3: “Critical” systems thinking approaches. 6

The real schism: first and second order cybernetics. 8

Footnotes: Afterthoughts. 10


A three-way classification of system thinking approaches

Social systems thinkers like to classify systems: e.g. into organisms, animal societies and machines.

They also like to classify approaches to thinking about systems and solving problems.


Gerald Midgley (2000) presented three classes of systems thinking approach as an evolutionary sequence.

He wrote of three phases of inquiry, which brought with it a new set of methods.

1.      Hard systems thinking: focused on concrete issues of ‘problems’ and problem solutions for issues where there was perceived “unity of purpose”.

2.      Soft systems thinking: began with the wider soft systems perspective on people and their perspective on issues.

3.      Critical systems thinking: presented as the latest development in a historical progression.


This paper challenges using the labels “hard” and “soft” to distinguish classes 1 and 2, and proposes a different, more schismatic, distinction.

Remember, it is easy confuse classifications of systems with classifications of system thinking approaches.

This paper regards people using a systems thinking approach as a meta system to any system they analyse or design.

Class 1: “Hard” systems thinking approaches

Midgely (2000), listed these so-called “hard” systems thinking approaches.

1956 General Systems Theory (Bertalanffy)

1956 Classical (classical) cybernetics (Ashby)

1957 Operations research (Churchman et al.)

1962 Systems engineering (Hall)

1963 Socio-technical systems (Trist et al.)

1965 RAND-systems analysis (Optner)

1971-72 System Dynamics (Forrester; Meadows et al.)


Read System Thinkers for more on some of the authors above, including Ashby.

Note that several System Dynamics models may be built of one reality.

So in one sense of the term, all System Dynamics models are “soft systems”.

Class 2: “Soft” systems thinking approaches

Midgely (2000) listed these so-called “soft” systems thinking approaches

1971 Inquiring systems design (Churchman).

1972 Second order cybernetics (Bateson)

1972 Soft systems methodology (Checkland)

1981 Interactive management (Ackoff)

1981 Strategic assumption surface testing (Mason and Mitroff)

1988 Cognitive mapping for strategic options development and analysis (Eden)


Read System Thinkers for more on some of the authors above, and read on for more on Checkland.

It is common to speak of soft systems as though they are clearly distinguishable from hard systems.

Yet both Ashby (above) and Checkland spoke of systems as input-to-output transformations.

And the term soft system is used in various ways.


Soft system meaning 1: a theoretical/abstract system, a perspective

The notion that systems are abstractions from real world behaviors has been shared and reinvented by many.

W Ross Ashby (supposedly a “hard” systems thinker) pointed out that infinite theoretical systems may be abstracted from a concrete entity.

Russell Ackoff wrote that different observers of the same concrete reality may see it as different systems.

Peter Checkland defined a system as a perspective of a reality, a world view or “Weltenshauung”.


Churchman initially wrote for actors in what was then commonly called the operations research department.

Operational researchers studied regular business operations and proposed how to optimise or otherwise change them.

The operations research department acted as a meta system to business systems.


Operational research

Abstract system descriptions

<create and use>                    <realised by>

Operational researchers <observe and envisage> Business operations


Surely, every system-of-interest is a selective perspective of a reality?

You can describe a concrete motor car from different perspectives.

It is a system for converting fuel into motion and waste gases.

It is a system built (to engineers' drawings) so as to achieve a sale value.

To call the motor car a system, you ought to nominate which system description you have in mind.


Soft system meaning 2: a system with questionable aims

“Churchman, Ackoff and Checkland consider hard systems methodology to be a special application of systems theory in situations where the objectives are not in question” Bausch (2001)

Surely, the aims of any non-trivial designed system are questionable?

Even mechanical engineers are taught to manage stakeholders and trade off between conflicting goals.


Perhaps the functional goals of atomic systems element are indisputable.

E.g. In a church service, the goal of circulating the collection plate is clear.

And in a motor car, the functional goal of a nut and bolt is clear.

However, even the designer of a nut and bolt must begin by asking stakeholders about non-functional goals.

E.g. cost of materials, cost of manufacture, longevity, standards to be applied, reusability in different structures.


Soft system meaning 3: an empirical/concrete system in which human actors play roles

Surely, giving system actors a free hand to determine the actions they perform is to undermine the concept of a system?

This is to confuse the system of interest with the meta system that observes and changes that first system.

See second order cybernetics below.


Soft system meaning 4: a situation rather than a system

Some say Checkland’s methodology is “situation thinking” rather than “systems thinking”.

However, Checkland uses the term “transformation” for a system, much as Ashby did.

His concern is to model regular business systems as transformations, and change them from a baseline (problematic) state to a target (improved) state.

The table below suggest his concept of a system is much the same as in hard system thinking approaches.


Generic system description

Ashby’s design for a brain

Boulding’s social system

Checkland’s Soft System (Transformation)

An object-oriented software system

A collection of active structures

that interact in regular behaviors

that maintain system state and/or

consume/deliver inputs/outputs

from/to the wider environment.

A collection of brain cells that

interact in processes to

maintain body state variables by

receiving/sending information

from/to bodily sensors/motors.

A population of individuals that interact by

processing information in the light of

mental images they remember, and

exchange messages to communicate meanings

to others and related populations.

A coherent entity, a structure of components

that interact in mechanisms that

maintain the integrity of the entity and

consume/deliver inputs/outputs

from/to each other and external entities.

A population of objects that interact by

processing information in the light of

state data they remember, and

exchange messages to communicate

with other objects and system users.


All four specific cases fit the generic system description reasonably well.

In every case, the real world entity is regarded as a system only in so far as it realises a system description.


It may be clearer to think of soft systems thinking as subtype of hard systems thinking, which is specific to:

·         Open (rather than closed) systems, which have external customers and suppliers.

·         Designed and purposive (rather than naturally evolved) systems, which have owners and stakeholders.

·         Human (rather than naively mechanical) entities, which employ human actors.


Checkland’s soft systems methodology

Ashby said infinite theoretical/abstract systems may be abstracted from any empirical/concrete entity.

Checkland defined a system as a perspective of a reality, a world view or “Weltenshauung”.

“Implicit here is the notion that an observer engaged in systems research will give an account of the world, or part of it, in systems terms;

·         his purpose in so doing;

·         his definition of his system or systems;

·         the principle which makes them coherent entities;

·         the means and mechanisms by which they tend to maintain their integrity;

·         their boundaries, inputs, outputs, and components;

·         their structure.” Checkland (1981, p. 102.)


The table below lists the seven steps of the methodology against the steps in a typical system engineering methodology.

The steps do not correspond one for one, but the difference are mostly cosmetic, a matter of emphasis or of particular technique.


A “hard” system engineering methodology

Soft Systems Methodology

Study the context: goals, constraints, stakeholders, their concerns, problems and requirements

Enter the problem situation.

Outline optional solution visions, along with value propositions for stakeholders

Express the problem situation (a rich picture)

Analyse trade offs between solution options with respect to the context

Formulate root definitions of relevant systems (validated using the CATWOE analysis below)

Describe a target system that compromises between conflicting goals and constraints

Represent human activity systems as conceptual models (business activity models)

Plan the work to move from the current to the target system

Compare the models with the real world.

Follow the plan to build and test the target system

Define changes that are desirable and feasible

Roll out the target system.

Take action to improve the real world situation.


Checkland’s root definitions and CATWOE

Drawing up root definitions of a business entity can uncover different viewpoints

E.g. Who is the customer of the system that distributes money from various European Union agricultural schemes to French farmers?

The farmers? The French government? The French taxpayer? The European agency whose existence depends on the scheme?


This source offers this example root definition.

“A company owned system to market the products and services of the company to existing and future clients by the most appropriate cost effective means.”

Then analyses that root definition using Checkland’s CATWOE acronym as below.

The “system” is primarily the Transformation, and the roles of Actors in that.


Root definition analysis



Entities that receive outputs from the transformation

“existing and future clients”



Entities that do or could perform activities in the transformation

“the company”



Activities that transform input to outputs

“Market the products and services of the company”


World view

The belief that makes sense of the root definition

Providing appropriate marketing to a particular client will promote company products and services



The decision maker concerned with system performance

“the company”



Constraints outside the system significant to the system

“appropriate cost effective means”


A truly general system theory (think, the solar system) does not feature Customers or Owners.

And for business systems, CATWOE may underplay the importance of Objectives, Outputs, Inputs and Suppliers.


Checkland’s conceptual model

Checkland proposed modelling the Transformation (the system) in an abstract and informal “conceptual model” or “business activity diagram”.

This model shows transformation activities connected by dependency arrows.

Two layers may be added to this very abstract model of business operations,

First, regular operations management: a control system to set aims for, monitor and direct the transformation activities.

Second, overarching executive/governance: a meta system to set aims for, monitor and control the regular operations management.


Checkland’s design pattern

Checkland proposed a generic design pattern for structuring a business in terms of these essential functions.

·         Defining targets (system aims)

·         Operations (system behaviors)

·         Monitoring and Controlling Operations (system regulation as in classical cybernetics).


Similar ideas can be found in other approaches.


Is there any difference between hard and soft systems?

After several decades, Checkland wrote "Soft Systems Methodology: A Thirty Year Retrospective" (2000)

In his review, he said the hard-soft distinction had proved slippery; people grasp it one week and lose it the next.

He said the term "soft" was intended to describe not systems, but an approach to solving problems in human activity systems.

Which brings us to his approach - his methodology.


Checkland proposed his Soft Systems Methodology starts not with spotting a system to be reengineered but with spotting a confusing or complex situation.

OK, but these are not mutually exclusive starting points – and the end product is indeed a re-engineered human activity system.


Checkland said he designed the Soft Systems Methodology as a “learning system”.

OK, but all traditional system design methodologies include activities intended to help designers learn about the context.

And it is important not to confuse the meta system with the system.

Defining a systems thinking approach as a learning system is very different from defining the target system as a learning system.

Class 3: “Critical” systems thinking approaches

This section is incidental in the context of this paper, and less thoroughly researched.


Midgley (2000) listed “critical” systems thinking approaches.

1983 Critical system heuristics (Ulrich)

1990 System of systems methodologies (Jackson)

1990 Liberating systems theory (Flood)

1991 Interpretive systemology (Fuenmayor)

1991 Total systems intervention (Flood and Jackson)

2000 Systemic intervention (Midgley).


I gather “critical theory” emerged from the Frankfurt school of sociology, which came to prominence in the 1930s.

Largely a Hegelian invention, the word “critical” implied a dialectic, a logical investigation or discussion of the truth of propositions.


A logical investigation?

Ulrich (1983) may well have been thinking of critical theory when he added the word “critical” to “systems thinking”.

He defined three heuristics of the approach.


·         Making sense of the situation: understanding assumptions and appreciating the bigger picture

·         Unfolding multiple perspectives: promoting mutual understanding.

·         Promoting reflective practice: analysing situations – and changing them.


Yet all systems engineering methods start with understanding the context and motivations.

They stress the importance of stakeholder management and viewpoints.

Even an EA framework like TOGAF recommends defining views and value propositions for each stakeholder group.

And it recommends analysing the current situation and considering changes.

So what is new?


A system transformation framework?

Gerald Midgley (2000) presented critical system thinking approaches as a post 1980 development of hard/soft systems thinking approaches.

The terms “total” and “systemic” (in the book titles he listed) imply these approaches lead to a root and branch transformation.

If not a revolution, the result is a major generational change from a current state of being to a new state of being.


Yet an EA framework like TOGAF is a transformation framework.

1.      forming a model of a current system (N): analysing, discussing and criticising it, envisaging changes

2.      creating a model of a target system (N+1): discussing, reviewing and agreeing it

3.      planning the change from system N to system N+1.

4.      working to change system N to system N+1.


Of course, formalising a systems thinking approach involves documenting mental models for discussion and agreement with others.

The approach should recommend the kinds of model you can or should document.

And it should recommend techniques such as stakeholder management and risk management.

You can recognise all the ideas above in both hard and soft systems thinking approaches.

So what is new?


System encapsulation?

One reader has suggested encapsulation of systems is a feature of critical systems thinking.

Again, agreeing the boundary of a system (expanding it, contracting it, or shifting it) is a feature of classic system theory.

So what is new?


An advanced approach?

Naturally, gurus present their preferred approach as the latest development in a historical progression.

And some modern-day consultants refer to their preferred approach as systems thinking.

That doesn’t mean their approach represents a historical evolution of classical system theory, or applies its core ideas.

Read Marxism and System Theory for a challenges to the notion of inexorable progression.


Gerald Midgely’s reference dates run suspiciously neatly in sequence from class 1 through class 2 to class 3.

Is it meaningful or useful to regard class 3 as an evolution of class 2, and that as an evolution of class 1?

It isn’t obvious that critical systems thinking brings many new methods, essential to the task, that could not be seen in earlier approaches.

I stand to be corrected on that.


A unified approach?

It is proposed that critical systems thinking unifies different systems approaches

And advises managers how best to use them.

But this is to gloss over a deep schism in systems thinking.


Many social systems thinkers adopt the second order cybernetics perspective of systems.

Checkland observed the distinction between hard and soft system approaches is slippery.

By contrast, the distinction between classical and second order cybernetics is schismatic.

The real schism: first and second order cybernetics

The hard/soft system distinction is questionable, for reasons explained above.

Even Checkland observed the distinction between hard and soft system approaches is slippery.

If the distinction is worth drawing at all, it is not as strong as another distinction in systems thinking.

The primary schism in systems thinking is between first and second order cybernetics.
And the trouble with second order cybernetics is that it can eviscerate the concept of the system.


“Second-order cybernetics, is the recursive application of cybernetics to itself.

It was developed between approximately 1968 and 1975 by Margaret Mead, Heinz von Foerster and others [including Bateson].

Von Foerster referred to it as the cybernetics of "observing systems" whereas first order cybernetics is that of "observed systems".

It is closely allied to radical constructivism, which was developed around the same time by Ernst von Glasersfeld.

Its concerns include epistemology, ethics, autonomy, self-consistency, self-referentiality, and self-organizing capabilities of complex systems.

It has been characterised as cybernetics where "circularity is taken seriously".” Wikipedia 24/02/2018


Observation: Second order cybernetics treats systems actors as system thinkers.

But the idea that actors continually modify the roles and rules of a system they work in undermines system theory.

How to maintain the integrity of the system concept?

How to extend classical system theory to embrace second order cybernetics?


1.      Distinguish a social entity from the many social systems it can realise.

2.      Separate meta systems (in which actors are system thinkers) from operational systems (in which actors are workers).

3.      Allow actors in a social entity to switch between roles in systems and meta systems.

4.      Allow actors to make incremental (generation-by-generation) rather than continual changes to system roles and rules.


Thus, it is the social entity (not the social system) that has self-organizing dynamics.

Read System Stability and Change for more on separating the meta system from the system.


Process and target-oriented management

The table below contrasts two management styles which correspond to this schism in cybernetics.



Process-oriented management

Target-oriented management

Processes are

Behaviors are

People are given



tightly constrained

processes to follow

call centre people given scripts


loosely constrained

targets only

door-to-door sales people given targets


Classical cybernetics

Second order cybernetics.


Hard and soft systems thinking approaches usually involve modelling roles and processes.

And intervening to change some of those roles and processes.


By contrast, in an extreme version of target-oriented management, you ask a group of people to meet some goals by doing whatever they choose.

You rely on the individuals’ abilities to interpret your direction and choose behaviors that lead to the goals you set.

If the nature and nurture-given abilities of the actors are up to it, they may succeed better than you expected.


Most people management involves something of both process and target-oriented approaches..

The question here is not which is better or worse, more or less advanced, it is whether it target-oriented management is well-called “systems thinking”.

There is a social entity, but is there a social system?


In classical system theory, a group of actors only form a system where, when and in so far as they agree some roles and rules and follow them for a while.

If the actors do agree some roles and rules, the system they describe is probably a simple one.

Whatever else they do – that is ad hoc, irregular, unrepeated, disorderly - is not systematic or part of that system.

Footnotes: Afterthoughts

These afterthoughts are only that – not essential to this paper.

Every enterprise employs actors who interact, directly or indirectly.

The enterprise may well be called a social entity, but is it also a “system”?

The answer depends on what kind of systems thinker you are.


System thinking or situation thinking?

System thinking can be seen as a kit bag of ideas and techniques for “situation thinking”.

It may lead to the description, testing and implementation of a new of changed system, or it might not.


From the 1970s onwards, systems thinkers have claimed institutions are in crisis, and something must be done.

Surely, institutions will always have problems, and system thinking will never provide a final answer to those problems.

It will always be necessary to intervene now and then.

Where an intervention involves describing regular behaviors, realizing them and testing the outcomes, it is an application of classical system theory.

But consultants may make interventions of others kinds – with little or nothing by way of process definition or testing of outcomes.

It isn’t always clear that the situation addressed is a “system” in any sense beyond being a named entity - a named organization or named group of actors.


If every problem or situation is called a “system”, then the word system tells us little or nothing.

A problem/situation might be a system, or it might not; a solution might be a system, or it might not.


More on the schism

Some modern system thinking is about defining purposes, goals or targets and motivating people.

You may assemble some people and motivate them to work towards given goals, paying little or no attention to how they do the work.

That group is a certainly a social entity, and perhaps a successful one, but it is not a system in the sense Ashby would recognise.


The fact is: the “system” in second-order cybernetics is radically different from the “system” in classical cybernetics.

There is a schismatic distinction between.

·         A social system - in which actors realise roles and rules - describable in accord with classical cybernetics.

·         A social entity - in which actors choose their own behaviors to reach agreed goals - as in second order cybernetics.


In classical cybernetics, a system is describable as a set of roles, rules and regular behaviors.

If you change the roles and rules, then you change any concrete system that realises that system description.


In second order cybernetics, a system is a social entity - a named entity or organization – in which actors are engaged.

The actors may change the roles and rules of the entity they are engaged in (perhaps even change its aims).

As long the system name remains the same, people speak of it as same system, though its roles, rules and behaviors may be very different.


Social entity thinking tends departs from general system theory in one or more of the ways listed below.

General system theory

Not general system theory



General to all domains of knowledge

Specific to situations in which humans interact

About roles, rules and regular behaviors

About individual actors (purposeful people)

About systems at the base level of interest

About meta systems that define and change roles and rules

About describing testable systems

About solving any problem in any consensual way


Promoting a “participative democracy”


These differences are further explored in related papers at

The suggestion here is that social systems thinkers and enterprise architects can benefit from gaining a deeper understanding of general system theory, and respecting it more than they do.

System theory is good to know, good for the soul, and practically useful in all kinds of thinking about systems.


Of course, seeing a business as a social entity is important; and is a primary responsibility of business managers.

Management consultants continually generate approaches to identifying problems in social entities and solving them.

The question here is whether classifying all these approaches as varieties of "systems thinking” has a useful meaning.

If every problem or situation is a system, if every entity we name or point to is a system, then the term “system” is meaningless.


This table maps the 3-way system thinking classification we started with to the 2-way schism discussed in this paper

3-way classification


4-way classification

2-way classification

Class 1 hard systems thinking

Systems in which actors perform regular behaviors

General Systems Thinking

General System Theory and

Classical cybernetics

Class 2 soft systems thinking

Systems in which human actors perform regular behaviors

Social System Thinking

Named organizations in which humans determine their behaviors

Social Entity Thinking

Second order cybernetics and

Management Consulting

Class 3 Critical system thinking

Problematic situations

Situation Thinking


On science and scientism

Classical cybernetics is scientific in so far as its deals with behaviors that are regular, or deterministic, or reproducible.

Actual (empirical) performances of behaviors can be tested for conformance to abstract (theoretical) descriptions of those behaviors.

As, for example, the actual orbits of planets are tested for conformance to astronomers’ descriptions of those orbits.


Evidence-based medicine is scientific; a medicine man (or shaman) is scientistic.

Some social systems thinking ideas are scientistic, meaning there is little or no evidence to verify or falsify them.

It easy to find problems in human organizations, more difficult to propose viable changes, and more difficult again to make them.

And then sometimes even more difficult again to prove whether the change was for the better or not.


A new system classification?

Here is the latest of several system classifications I have toyed with.

I don’t mean to present this as a scale of complexity, or a progression of any kind.

System kind

Actors relationship to scripts


Strong or Involuntary

Actors/parts use no script


Solar system


Cuckoo Clock

Actors/parts rigidly follow scripts


Biological organism


Software system

Weak or Voluntary

Actors/parts interpret scripts


Orchestra, Church

Actors/parts script their own roles


Marriage, Small business


Other reading

For a different background the one presented above try “Introducing Systems Approaches



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..