Activity systems thinking – and EA

Copyright 2020 Graham Berrisford. Now a chapter in “the book” at Last updated 22/01/2021 18:01


For a preface outlining the motivations for this work, click here.




Meadows’ view of activity systems. 2

Ackoff’s view of activity systems. 4

On EA as activity systems thinking. 5

On EA as cybernetics. 7

On EA as business system planning. 10

On EA as "design thinking". 11


Continuous change v. discrete change. 12

System state change v. system mutation. 13

Discrete natural mutation. 13

Discrete designed mutation. 14

Two kinds of agility. 15

Meta system thinking. 17

On tackling wicked problems. 20




In 1956, Kenneth Boulding wrote on what was becoming known as management science. In his article, he presented general system theory as the skeleton of science. He asked what are the elements of a social system? Are they actors, or the roles actors play in activities?


This question has hung over social system discussions since the nineteenth century.

"The first decision is what to treat as the basic elements of the social system. The sociological tradition suggests two alternatives: either persons or actions." Seidl 2001


We may put people at the centre of our thinking: e.g. study the people in a card school. Or put actions at the centre of our thinking: e.g.  study the game of poker. The example features three systems thinking concepts.


Abstract activity system

the rules of poker

Physical activity system

a poker game in progress

Social entity

a card school


A system's actors may be only part-time, and perform activities in other systems. E.g.  most card school members both play poker and pay taxes.


In this work, we distinguish two kinds of systems thinking.

Social entity thinking is about a network of actors, who perform activities.

Activity systems thinking is about a network of regular activities, performed by actors.


It has been said that enterprise architecture regards an enterprise as a "system of systems". We say EA sees a business as a social entity. The social entity may realize any number of (possibly conflicting) activity systems. EA strives to extend, improve and coordinate those activity systems.


The advice in short is:


Is your interest what actors do to

Then use

meet aims however they choose?

Social entity thinking


do both above and below?

Social entity thinking

Activity systems thinking

play roles in regular processes?


Activity systems thinking


The previous chapter looked at social entity thinking with regard to EA. This chapter looks at activity systems thinking with regard to EA.


Activity systems thinking is useful whenever we seek to:

·       understand how some outcome arises from some regular behavior.

·       predict how some outcome will arise from some regular behavior.

·       design a system to behave in a regular way that produces a desired outcome.

·       intervene in a situation to change some regular pattern of behavior.

Meadows’ view of activity systems

The table below quotes Meadow’s questions and answers on system dynamics. It compares them with the concepts of cybernetics and more general activity system thinking.



General activity system thinking (after Ashby and Weiner’s cybernetics)

System Dynamics (after Forrester) quoted from Meadows’ “Thinking in Systems: a Primer”

What characterizes an activity system?

A system is characterized by a pattern of interrelated activities.

An activity is a behavior (or occurrent) that changes or makes something.

An actor is a structure (or continuant) that plays a role in performing activities.

“A system is a set of things [elements] people, cells, molecules, or whatever

interconnected in such a way that they produce their own pattern of behavior over time.”

“The system may be buffeted, constricted, triggered, or driven by outside forces. But the system’s response to these forces is characteristic of itself. The behavior of a system cannot be known just by knowing the elements of which the system is made.”

What are the aims or purposes of a system?

Actors, which occupy space, may be visible or tangible.

Activities, which run over time, are harder to see.

Aims are even harder to see, and may be perceived or expressed as motivations or as outcomes.

“The word function is generally used for a nonhuman system, the word purpose for a human one, but the distinction is not absolute, since so many systems have both human and nonhuman elements.”

“If information-based relationships are hard to see, functions or purposes are even harder. A system’s function or purpose is not necessarily spoken, written, or expressed explicitly, except through the operation of the system. The best way to deduce the system’s purpose is to watch for a while to see how the system behaves. Purposes are deduced from behavior, not from rhetoric or stated goals.”

Is every composite entity an activity system?

No, a system isn’t just any old collection of things or actors. It is a collection of actors organized to perform the system’s characteristic activities.

“A system isn’t just any old collection of things. A system is an interconnected set of elements that is coherently organized in a way that achieves something.”

Is there anything that is not an activity system?

Yes, a passive structure (Linnean classification of species, the Dewey decimal system, the periodic table in chemistry). And a collection of actors that do not interact in a recognizable pattern of activities.

“Yes—a conglomeration [of elements] without any particular interconnections or function.”

How to know you are looking at a system? And not just some stuff that exists or happens?

a)      Are the activities regular and repeatable?

b)     Are the actors’ roles in those activities regular and repeatable?

c)      Do actors interact to produce effects (state changes and outputs) they cannot produce on their own?

“How to know whether you are looking at a system or just a bunch of stuff:

a)      Can you identify parts? . . . and

b)     Do the parts affect each other? . . . and

c)      Do the parts together produce an effect that is different from the effect of each part on its own? and perhaps

d)     Does the effect, the behavior over time, persist in a variety of circumstances?”

Which of actors, activities and aims are most important?

All are essential to what a system does, and interdependent. What matters most are usually its aims and the effects of activities. The actors, the most tangible and visible elements, are often the least important.

“To ask whether elements, interconnections, or purposes are most important is to ask an unsystemic question. All are essential. All interact. All have their roles. But the least obvious part of the system, its function or purpose, is often the most crucial determinant of the system’s behavior. Interconnections are also critically important.”

What does it mean to change a system?

Changing actors usually has the least effect on a system (change every player on a football team, it is still a football team).

But if a systems’ activities change, then it mutates into a new system generation, or a different system.

Changing a desired aim usually implies changing the activities, which sometimes implies changing the actors.

“Changing relationships usually changes system behavior. The elements, the parts of systems we are most likely to notice, are often (not always) least important in defining the unique characteristics of the system.”

And in chapter 1.

“Changing elements usually has the least effect on the system. If you change all the players on a football team, it is still recognizably a football team. A system generally goes on being itself, changing only slowly if at all, even with complete substitutions of its elements —as long as its interconnections and purposes remain intact. If the interconnections change, the system may be greatly altered. It may even become unrecognizable. Changes in function or purpose also can be drastic.”

Which is to say that changing the purpose of a system is to change the system itself


Despite the high degree of correspondence in the table above, there is some ambiguity in Meadow’s comments, explored in chapter 5.

Ackoff’s view of activity systems

Russell Ackoff was known as a brilliant observer of humans, their institutions and their failings. He was also known as a systems thinker. In “Towards a System of Systems Concepts” (1971) he started some distance from human society. He set out more than thirty general ideas about systems. He started with eleven terms and concepts that clearly derive from more general system theory and cybernetics.

1.     System: a set of two or more interrelated objects (parts or elements).

2.     Abstract system: a system in which the elements are concepts. 

3.     Concrete system: a system that has two or more objects.

4.     System state: the values of the system’s properties (state variables) at a particular time. 

5.     System environment: those elements and their properties (outside the system) that can change the state of the system, or be changed by the system. 

6.     System environment state: the values of the environment’s properties at a particular time. 

7.     A closed system: one that has no environment. 

8.     System/environment event: a change to the system's state variable values.

9.     Static (one state) system: a system to which no events occur (its state does not change).

10.  Dynamic (multi state) system: a system to which events occur (its state changes).

11.  Homeostatic system: a static system whose elements and environment are dynamic.


Ackoff’s system has an internal state. The state of people, processes, materials and machines of interest changes over time. Their current state can be remembered in the values of state variables and communicated in messages.


Ackoff’s system is bounded, and can be closed or open. An open system interacts with its environment by consuming inputs and producing outputs. The outputs change the state of things in the environment of the system.


Ackoff characterized a system as a whole in which two or more parts interact. And he characterized parts as elements that interact to produce the behavior of the whole.


Contrary to what some interpret Ackoff as having said: You can remove parts from a whole (e.g. motor car) without affecting its ability to meet its main aim (e.g. to carry people from A to B). Because some parts interact to that end, but other parts (e.g. airbag, safety belt, carpet, arm rests, radio) are designed to meet other aims.

On EA as activity systems thinking

"Managers do not solve problems, they manage messes". Ackoff.  The typical business "organization" is less-than-fully organized; it is a mess of aims and activities.


EA sees a business as a social entity that may realize any number of (possibly conflicting) activity systems, and strives to coordinate them. EA is about activity systems that can be modelled, and models that can be realized. EA cannot define the whole of a business as one activity system. It can identify where selected parts act holistically in regular or repeated ways. And identify many such activity systems - nested, overlapping, competing. It strives to optimize how systems are best coordinated to the benefit of the whole.


A business activity system may reasonably be defined thus. It is orderly; it is dynamic; and characterized by regular activities. It is realized by a set of actors and resources that is coherently organized and interconnected so as to perform required activities.


Actors are active structures (people, computers, other machines) that occupy some space and play roles in activities.  Activities are regular behaviors, performed over time. That advance the state of things: people, processes, materials and machines of interest. The state can be remembered in the values of state variables and communicated in messages.


EA is about systems that can be modelled, and models that can be realized as systems. As Groucho Marx might have said: "An enterprise that is simple enough to understand without a model of it is one that doesn't need an enterprise architect." EA models feature the concepts below.


1 Aims or goals

EA starts from the aims of system sponsors and other stakeholders. It identifies the desired outcomes that currently or should emerge from business activity.


There are accidentally evolved-systems, like the solar system, a tree and a beehive. But EA is about purposefully-designed systems. These are ordered or organized by designers to produce outcomes that meet some given or aims or goals.

·        Goal/objective: target aim for activity, a desired effect of outputs being used by external actors.

·        Outcome = an actual effect of outputs being used by external actors.


2 Product or services

EA identifies the aims of sponsors and other stakeholders. And identifies the products and services that currently or should emerge from business activity.

·        Product = an information or material structure output from a system

·        Service = the external view of an activity or process that produces outputs or state changes (results) of use to some external actor(s).


3 Activities and processes

EA sees a business system as a holistic network of activities. Activities are sequenced in processes (or value streams).

·        Process (cf. Value Stream): activities sequenced to complete a service.


You may define a system in two ways:

·        externally (from the outside) in terms of input-to-output transformations made

·        internally in terms of activities performed and state variables maintained.


Checkland's "soft systems methodology" features business activity modelling. A business is modelled as a set of regular activities that transform inputs into outputs. The SIPOC view below encapsulates the processes of such a system and connects it to actors in its environment.



Activity System






This SIPOC diagram is only a simple overview. It doesn’t show feedback loops between a business system and actors in its environment. The outputs of a business system (say invoices) can influence its future inputs (say, payments).

·        Input = an information or material product consumed by a system, supplied by some external actor(s)

·        Output = an information or material product provided by a system to some external actor(s)

·        External actor = a player of customer, supplier, user or observer roles.


Nesting of systems

The boundary of a system depends on the perspective of the observer(s). E.g. a boiler transforming water into steam may be seen as a system, or as a subsystem of a steam engine. Having encapsulated an activity system, you may divide it into subsystems that interact. Then, each subsystem can be decomposed, and so on, until you reach what you regard as atomic actors and activities.


System decomposition is a simple idea; it is important and useful today. However, there are many other ways of looking at systems, and systems thinkers use many terms ambiguously.


4 Logical active structures

EA is largely not about physical structures (buildings, hardware, vehicles and human bodies). But it does assign activities to logical structures - functions and roles.


Some misleadingly liken activity system architects to building architects (think of a windmill). True, both shape what is to be made and govern those who make it. But you can visualize a building and instantly recognize it in a picture or photograph. Whereas, you cannot visualize a dynamic activity system (like a poker game) in a comparable way.


However, EA does define logical structures to which activities can be assigned.

·        Function (cf. Capability): activities grouped for understanding and assignment.

·        Role: activities grouped for assignment to one or more actors.


5 Passive data structures

EA is about systems in which human and computer actors are “active structures”. These actors create and use passive structures, notably, data/information structures.


Generally speaking, systems consume flows of energy, matter, forces or data. In EA, the focus is on the last of those, on data/information flows. Since businesses operations depend on business information, a business needs information systems to

·       monitor and direct business actors and activities (people, processes and machines).

·       maintain state variables that represent the state of business operations (actors, activities and resources).

·       are updated by inputs submitted by business actors.

·       produce outputs that inform and direct the actors and activities in business operations.


General speaking, actors remember information in memories, and communicate information in messages. The memories and messages contain data structures. Actors create and find information, in those messages and memories, by encoding meaning in symbols and decoding meanings from symbols. Successful communication requires the sending and receiving actors share a common language, and so share the meanings of symbols in messages and memories.


EA does not model ad hoc or spontaneous communications. It models only data structures that appear in regular messages and memories.

·        Data Flow or Data Store: information encoded in a message or memory.


In practice, the words “data” and “information” are used interchangeably. It is assumed that receives will find the information in a data structure that its sender intended. So, the data and the information are in 1-to-1 correspondence.


For more data, information and language read the section on cybernetics below.


6 The social entity that implements a system

Logical functions and roles must be mapped to physical organization units and actors. So, EA does require some social entity thinking about:

·        Organization units: how activities and/or actors grouped for management.

·        Actors: the individual that play roles in the systems of interest.

On EA as cybernetics

Ashby’s “Introduction to Cybernetics” (1956) contains many ideas applied by EA in modelling business activity systems.


Ashby 1: Systems are related by information feedback loops

One system can sense and direct changes in another system, or its environment, by means of an input-output feedback loop.


EA features business actors that consume and produce information (descriptions, decisions and directions) in data flows or messages.


Ashby 2: Abstract systems represent physical systems

Ashby noted the term “system” is ambiguous; is used in two ways describable here as:

·       System as entity = a whole real-world thing regardless of any particular observer.

·       System as pattern = a set of regular or repeatable behaviors that advance some variables.


Ashby said it wrong to point to a named entity and call it "a system". An entity can only be seen as a physical system when, where and in so far as it realizes what Ackoff called an abstract system.


The triangular concept graph below distinguishes physical systems from entities that realize them. And distinguishes logical/abstract system descriptions from physical systems. Read it left to right thus: Systems thinkers <create and use> abstract systems <represent> physical systems.


Activity systems thinking

Game playing

Abstract systems

<create and use>          <represent>

Systems thinkers <observe and envisage> Physical systems

Game rules

<create and use>          <represent>

Gamers   <observe and envisage>   Games

<are realized by>

Physical entities

<are realized by>



(BTW, the four rows of the "enterprise continuum" in TOGAF suggest modelling at three levels of abstraction by idealization.)


Ashby 3: The relationship between abstract and physical systems is many-to-many

Many physical entities (instances) can realise one abstract activity system (a type). E.g. Countless card schools can play the game of poker.


And one physical entity may act in many physical systems. E.g. One card school may play poker, or play whist, or share a pizza.


In EA, the relationship between activity systems and social entities is many to many. One generic activity system model can be realized by many businesses. And one business can realize several activity systems, from generic to organization specific.


(BTW, the four columns of the "enterprise continuum" in TOGAF suggest modelling at three levels of abstraction by generalization from organization-specific.)


Ashby 4: Most systems of interest can be modelled using discrete dynamics

Ashby’s cybernetics is based on simple presumptions about how we describe reality. To describe the continuous space of the universe, we differentiate discrete things - entities. To describe the flow of time universe, we differentiate discrete changes – events.


EA is about digitizable business activity systems. It involves analysing and designing human and computer activity systems. The systems are modelled in terms of discrete entities and events, actors and activities. The events trigger actors to perform activities that advance the system’s state. The state variables have discrete (rather than continuous) values.


Ashby 5: An activity system is rule-bound

3/1 “We are concerned in this book with those aspects of systems that.. follow regular and reproducible courses.” (Ashby 1956)


EA is about regular behaviors of systems in which the range of possible activities is defined.


Ashby 6: An activity system advances state variables

“A variable is a measurable quantity that has a value.” “The state of the system is the set of values that the variables have.” “A system is any set of variables which he [observer] selects from those available on the real machine.” (Ashby 1956)


EA is about systems that maintain data recording the state of entities and events that the business must monitor and direct in its environment.



Activity System






Actors <perform> Activities

Inputs + States <determine> Activities

Activities <consume> Inputs

Activities <change> States

Activities <produce> Outputs



Ashby 7: Adaptation by mutation differs from adaptation by state change

EA is about systems that change in two ways.


The second part of this chapter discusses design for change..


Ashby 8: Self-organization implies reorganization by a higher process

EA is a "higher process" for organizing the systems of an enterprise. The second part of this chapter discusses "self-organization" and meta systems thinking.


Ashby 9: The law of requisite variety

"The larger the variety of actions available to a control system, the larger the variety of perturbations it is able to compensate".  Ashby 1956


EA is about systems that must remember enough about entities and events (inside and outside a business) to direct them as need be.


Ashby 10: Coding is ubiquitous in thought and communication

Input and output messages contain data structures that encode the state of entities and events. The data may be recorded and maintained in the memory of a system for future use.


In a system of communicating actors, meaning is found in

·       the intention of an actor when encoding a data structure, with reference to a language.

·       the interpretation of an actor when decoding a data structure, with reference to a language.


In human society, the meaning of a word like "policy" is ambiguous. And it may be changed by one actor regardless of how other actors interpret it.  So, to succeed in communicating, actors must share the same code or language. And where mistakes cannot be allowed, as in science and business, the meaning of data is defined in domain specific languages.


In enterprise architecture?

EA is about activity systems in which business actors create and use data; they remember information in persistent memories. They exchange information in messages.


There is no meaning in a data structure on its own. Meaning appears only in moments when actors create and use data structures or symbols, with reference to a language in which that data is mapped to a meaning. So again, the meaning of data in memories and messages may be defined in some kind of meta data or domain specific language.


Architects can acquire or define a controlled vocabulary in some kind of data dictionary or canonical data model. For example, "policy number" as the identifier of a particular "policy", which has an agreed set of attributes/variables. This "meta data" defines the meanings created/used and by actors when encoding or decoding any variable, field, attributes called “policy number”.


An activity system is a bounded domain in which actors perform activities to meet aims.  Where one system depends on another, they exchange messages or data flows. EA may resolve discrepancies between languages used in different bounded contexts by standardization. Or else, by inserting transformers or adapters which translate data structures in transit between senders and receivers.


For more on Ashby's core ideas, read this chapter.

On EA as business system planning


The relationship to business planning

Is there anything "higher" than EA? Yes. Business directors are responsible for business planning.  They respond to business drivers by declaring strategic directions and top-level goals/objectives.


Business planning may involve predicting demand and directing changes to any of the following.


Enterprise architects may both stimulate and contribute to business planning (above). But their primary responsibility is business system planning (below).


The need to address a mess of systems

The boundary of a system is determined by observers with an interest in its outcomes. You can draw a boundary around a steam engine with some hope of describing its internal actors and activities - reasonably comprehensively. You cannot draw a boundary around an enterprise with the same hope.


In reality, a business is more a mess of systems than a system of systems. You can find countless different activity systems in it.


EA is a never-ending struggle to improve, synchronize and extend business systems. To make them more efficient and effective to the benefit to the wider business, and its stakeholders.


The need for methodical design

A general activity system design method generally proceeds along these lines:

  1. Aims to be met
  2. Activities required to meet 1
  3. Actors required to perform 2.


To paraphrase Meadows: the actors are the most concrete and tangible elements. The activities are harder to see, and the aims are even harder to see. A general business activity system design method proceeds along the lines below.

  1. Aims, goals, objectives, requirements of stakeholders
  2. Product and services of use to actors in meeting 1.
  3. Activities and processes that sequence them to complete 2
  4. Functions, capabilities, roles required for the performance of 3
  5. Data flows and data stores that hold and convey information needed by 4.
  6. The social entity (organization of actors) that implements the system, performs roles and manages resources.


The sequence is flexible. The choice between robot or human actors might lead you to modify the roles or the activities. But in general, you can't or don't want to acquire actors with no regard to their roles. Or define roles with no regard to activities and processes they are responsible for. Or define processes and services with no regard to their goals.


See the references for illustration and elaboration of the design process above.


The need for change control

EA grew out of looking to extend or optimize the performance of an enterprise. By taking a cross-organizational and strategic perspective. By standardizing and integrating systems. So, EA shapes and steer changes to business activity systems.

The actors playing roles a system cannot continually change how they individually interpret messages and respond to them. That would undermine the concept of an activity system.


E.g. Think about game of poker. To change the roles or rules of the system is to design a new system, or system generation. To implement such a change requires the agreement of stakeholders and the understanding of actors who play the roles.


So, EA presumes business activity systems will be stable for a generation. They will be changed in discrete and testable steps, from one generation to the next. In agile software development, the generation is short; in EA, the generation is long. Either way, changes to designed activity systems must be made under change control – else the concept of the activity system evaporates.

On EA as "design thinking"

Generally, designers order elements of an entity to produce desired outcomes. More specifically, enterprise and business architects order elements of a business to produce desired outcomes. Above all they order those regular business roles and processes that create and use business data)


The most well-known architecture framework is called TOGAF. It is neither a prescriptive process, nor entirely consistent and coherent. It is a flawed but highly flexible assembly of processes and products people adapt as they choose for change initiatives that need big-up-front design and planning.


Herbert Simon defined "design thinking" in “The Sciences of the Artificial” (1969).  TOGAF clearly embodies Simon’s "design thinking" principles. His core ideas being that designers


TOGAF embodies those and other design thinking ideas below.


In the 1980's Chris Argyris promoted double-loop learning, teaching people to think about their assumptions and beliefs, as a tool for system design. While TOGAF does not explicitly promote double-loop learning, it is crammed with directions to revisit assumptions. At every phase, architects check what is being proposed against, business drivers and goals, and the concerns and requirements of stakeholders. And this is done alongside continual requirements change management.



System change can be classified in three ways:

·       continuous or discrete

·       state change or mutation

·       natural/accidental or designed/planned.


Recognizing change as a three-dimensional phenomenon helps us to think more clearly about what it means to model change and design for it.





State change











Continuous change v. discrete change

The universe is an ever-unfolding process of continual change. A social entity continuously and naturally mutates. Its members change, and it responds to environmental changes that were not predicted or anticipated.


By definition, an activity system is an island of regularity in the universe. A system that continually changes its nature would be a contradiction in terms. If there is no stable pattern, no regularity, no repetition, then there is no system to describe. A system cannot possibly be designed to continually mutate into infinite different systems.


Ashby and Maturana, separately, rejected continual mutation as undermining the concept of a system. However, continuous change can be simulated by dividing changes into steps frequent and small enough to appear continuous.

System state change v. system mutation

When the environment of a system changes in a way that threatens its survival, the system may adapt or be adapted in two ways. Ashby wrote.

“5/7. 'adaptation' is commonly used in two senses which refer to different processes.”


State change

There are four varieties of state change

1.     Continuous natural state change (e.g. the growth of a crystal in liquid)

2.     Continuous designed state change (e.g. analogue light dimmer switch)

3.     Discrete natural state change (e.g. asleep to awake, or day to night)

4.     Discrete designed state change (e.g. light on to light off)


“A system generally goes on being itself… even with complete substitutions of its actors as long as its interactions and purposes remain intact.” Meadows

Meadows’ system changes state when, by regular and repeated interactions, the stocks grow and shrink. But the stocks and the flows that relate them don’t change.


System mutation

There are four varieties of mutation,

1.     Continuous natural mutation (e.g. maturation of child into adult)

2.     Continuous designed mutation (impossible)

3.     Discrete natural mutation (e.g. generational change, parent to child)

4.     Discrete designed mutation (e.g. system version 1 to version 2)


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

If you change the stocks or the flows in a causal loop diagram, then you define a new system. It may be seen as the next generation of the same system, or a different system altogether.


Primarily, EA is about the design of activity systems that change state in discrete steps and are changed under change control. The rest of this chapter is about discrete step mutation – natural and designed.

Discrete natural mutation

This is organic generational change, as from parent to child. A biological species mutates naturally, not continually, but via discrete birth and death events.


When one generation of a system is replaced by the next, there is usually some kind of handover. And given limited resources, each generation must die to make space for the next.


·       "Genetic mutations arise by chance.

·       They may or may not equip the organism with better means for surviving in its environment.

·       But if a gene variant improves adaptation to the environment (for example, by allowing an organism to make better use of an available nutrient, or to escape predators more effectively—such as through stronger legs or disguising coloration), the organisms carrying that gene are more likely to survive and reproduce than those without it.

·       Over time, their descendants will tend to increase, changing the average characteristics of the population.

·       Although the genetic variation on which natural selection works is based on random or chance elements, natural selection itself produces "adaptive" change—the very opposite of chance."


Biology doesn't design for the future. Most (99.9%) of the species that evolved are now extinct. And evolution would be impossible if organisms didn’t die.


Nature does not "design processes that foster adaptability and robustness for a range of scenarios that could come to pass."  Rather, biological evolution replaces older entities by new ones, the fittest of which survive. Nature kills off "designs" that don’t work well enough to be reproduced. It replaces them by whatever new "designs" turn out to better fit today's environment.


The adaptability of a species does not lie in the regular processes of an organism. It lies in the higher process or meta system of evolution by natural selection. This kind of evolution discards almost every new feature it creates.


The analogy in the business world would be the most brutal of capitalist systems. One in which every business fails, to be replaced by start ups. And only a small number of the start ups' random innovations survive for long in a changing world.

Discrete designed mutation

This is also generational change, but the term design implies some motivation or intent. Which in turn implies there is a designer who makes a conscious effort to invent or change something, and there is a design or plan to change a system from version 1 to version 2.


Prioritizing design effort

Macro-scale or transformational innovation doesn’t have a good track record. It has been reported that 70% of business transformations fail. Instead, some recommend continuous improvement, which is a principle of Kaizen.


EA doesn't have to be about whole-scale transformation. It can be about stimulating, prioritizing, coordinating and optimizing smaller scale incremental innovations. Obviously, you should match your approach to your situation.


Is everything OK?

Suppose your services/products appeal to customers, and you're in profit. Then, to embark on redesigning your whole business is needlessly costly and risky. Better to prioritize things needing attention and fix them incrementally; see the next section.


Is radical transformation necessary?

Suppose your services/products are not selling, and you're losing money. Then you must consider substantially revising or largely replacing services/products. And then redesign and transform your enterprise to provide those different services/products.


We cannot cling too hard to the idea that old businesses should be transformed to meet new demands. The obstacle to evolution is usually what some call "sunk costs". That is, the investment in people, processes, technologies, buildings and other resources that are hard to discard or change.


The history of civilization tells us that old businesses are replaced by fitter competitors and/or start ups. Sooner or later, most businesses are taken over by new management or replaced by other businesses.


Else, look to continuous improvement

EA is about planning discrete step changes. If the steps are frequent and small enough they can appear continuous.


Google is said to have 2 billion lines of code. The remarkable thing is not the difficulty we have designing complex software systems; it is that we succeed at all. We do it by incrementally complexifying what starts out as a relatively simple system. We replace each generation of the software system by the next. (Biological species also adapt to environmental changes in this way.)


Ackoff is often quoted as saying "Improving a part does not necessarily improve the whole". Nevertheless, improving a part on its own is a reasonable way to improve the performance of the whole. Attending to the most costly part, and removing the largest bottleneck (one at a time) are recommended practices for improving a system. Such incremental development is a feature of biological evolution and agile system development.


So, you may incrementally redesign the relevant parts of your enterprise as you go along.

“Instead of obsessing over spreadsheets, he said, executives should walk factory floors or interact with more customers. Innovation often doesn’t come through one breakthrough idea but a relentless focus on continuous improvement, he said.” Elon Musk


The only time to put continual improvement and fixing things on hold is just before and during large-scale transformation. Between transformations, you should fix whatever parts of the whole need attention, and practice continuous improvement.

Two kinds of agility

The two kinds of agility may be distinguished


Agile system mutation

Agile system development implies a designed system mutates frequently. Each mutation produces a new system generation, slightly different from the last. Among the best-known principles for agile system design are:


In enterprise application design, price must be paid for short-term development-release cycles. Applying the KISS and YAGNI principles inevitably leads to the cost of database and software refactoring. In higher level business architecture, the cost might be in the redesign of physical machines or other resources, or the retraining of people.


Agile activity systems

An agile activity system is one that can handle changes in its environment, without having to mutate.  To design a system in anticipation of future changes tends to make it more complex than is needed initially. It has to allow for a wider range of possible activities than actors perform to begin with.


So, contrary to the agile development principles above, the principles are:


An enterprise may fail because it cannot, or does not, adapt to changing circumstances, inside or outside the business. We don't need any system theory to tell us that systems must adapt or be adapted We do however need to unscramble the many system terms and concepts discussed in this connection, such as agility, antifragility, robustness and resilience. What can EA do to prepare a business handle to future when it arrives?


Design for antifragility

Antifragile implies robust and resilient, but those words are used variously. This paper defines robustness thus. “A core property of robust systems is given by the invariance of their function against the removal of some of their structural components.” That maths may be fine; the applicability to business and other social systems is unclear.


The paper makes some questionable assumptions. It seems assumed the structural components are active (actors) rather than passive material or information resources. And assumed that one actor plays one role, and vice-versa, so removing one removes both.


Given one role played by many actors, we can remove actors without qualitatively changing the system behavior. But given one role is responsible for activities in several business processes, removing the role may disable all of them.


However, some processes may be completable without some ideally-expected activities. And in practice, some processes are more central to business success than others. This “centrality” may not be evident from a model of system structure and behavior.


Businesses conventionally mitigate the risk of losing a vital component by redundancy and back up. The term robust has other meanings; it commonly means adaptive in the face of a changing environment. The term resilient is also used with reference to changes in environmental conditions.


A convenient way to distinguish the two terms is this. A robust system handles disruptive or unwelcome events and conditions (think immunity to infection). A resilient system mutates to handle new events and conditions (think evolution).


To design a robust or resilient system usually implies a design that

·       is generic to the point of vacuity, or

·       features redundant components and processes and/or back up versions

·       more complex and resource-intensive than it needs to be right now.


We might look to "big up front" design of systems, and to the "complexify" principle above. E.g. The architects of a logistics business might design over-sized and resourced "hubs" with highly mechanized storage and retrieval of items.


E.g. Software architects might design a broad and rich database structure that will be stable, or perhaps configurable. The idea being the database need not be restructured during the following process of incremental and iterative agile software development. It might work, but note the challenges of design for antifragility include these. Can you

·       correctly predict what kind of changes or shocks your system must deal with?

·       correctly predict how likely or often each kind of event will happen - over the period you judge your system will have to last?

·       afford whatever redundancy, complexity or resource consumption that design for anti-fragility requires?


Beware that you may guess wrongly about the events you need to anticipate. Designing to anticipate a disruption in the market may put you out of business before the disruption arrives. And a robust or resilient capability may be lost if not exercised now and then.


These concerns put pressure on business managers to do the minimum they can get away with – which is a challenge for EA.

Meta system thinking

What defines or changes a system is usually considered to be at a “higher” level. It doesn’t have to be a system itself, but it might be. It could be a social entity, or a process, or a “meta system”.


Self-organization in cybernetics

Krippendorff wrote:

“Although second-order cybernetics (Foerster et al. 1974) was not known at this time, Ashby included himself as experimenter or designer of systems he was investigating.”


Heinz von Foerster was an Austrian American scientist who combined physics with philosophy. He is widely credited as the originator of “second-order cybernetics” in the 1970s. It is less about regular business operations, and more about how those operations evolve. The basic idea was to include the system describer in the system that is described. And allow the system to be self-organizing and creatively self-improving.


However, this conflates actors with roles (remember Boulding’s article in 1956). It undermines the  concept of an activity system, in which actors are defined by their roles. It conflates two roles, for example: the role played by a gamer in defining the rules of poker, and rhe role played by the same gamer in a card school, playing games.


Game playing

The rules of poker

<create and use>          <represent>

Gamers   <observe and envisage>   Poker games

<are realized by>

Card schools


In classical cybernetics

Ashby and Maturana rejected self-improvement as undermining the concept of a system. Ashby wrote "the term "self-organization" perpetuates a fundamentally confused way of thinking about how a system may evolve". And “No machine can be self-organizing in this sense.” He said that re-organize a system, it must be coupled to another system.


Ashby proposed that the rules of a regular system can be modified by a higher process or meta system.  His principle is that a system (call it S) cannot reorganize itself. But it can be reorganized by “higher” process or system (call it M). M can observe and change the behavior of S. When circumstances demand it, M changes the roles and rules of S. By doing this, M creates a new generation of S.


Looked at this way term “self-organisation” seems inappropriate and unhelpful. The process of system mutation might better be called “rule-setting reorganization”. And it happens in discrete steps.


Meta system thinking

Ashby’s idea was that a system can be re-organized by a higher process or meta system. Meta systems thinking adds one more idea, that one actor may play two different roles:


·       a role as a tennis player in tennis matches

·       a role as a law maker in the Lawn Tennis Association.


The higher social entity, process or meta system is a level up from the base level system. E.g. the process of biological reproduction is a level up from the lives of organisms. And the Lawn Tennis Association is a level up from the progress of tennis matches around the world.


(We could speculate about multiple levels of meta system: M, MM, MMM…).


Meta systems thinking reconciles activity system theory with self-organization. And it goes some way to reconcile second order and classical cybernetics. It presumes that a system evolves by inter-generational steps.  In the case of human activity systems, it implies some kind of change control.


Applying the idea

It turns out that meta system thinking can be applied in a variety of domains.

·        Homeostatic machines

·        Biology

·        Sociology

In biology the higher process is discrete-step evolution by natural selection. In sociology, the higher social entity, process or meta system is discrete-step evolution by design. Typically, a committee or governing body determines the rules of a collective or other social entity. It designs changes to the roles of actors, and directs actors in the social entity to follow them. It also has some power to ensure compliance to the rules.

Applying meta system thinking to a homeostat

Ashby built a homeostat to illustrate inter-generational reorganization. A “higher” machine detects when environment variables move outside the range safe for the lower machine to function. The higher machine.

1.     Takes as input the rules applied by the homeostat to its variables

2.     changes those rules at random

3.     triggers the homeostat to realise the new rules.


What if the higher-level system detects this doesn’t improve matters? Then it can change the rules again, until the lower system ether dies or works better.


What if the lower-level system is not an individual, but a species or population of actors or agents? If the higher-level system is like biological evolution, it will change the rules of each individual differently. And the leave it to the environment to favor individuals that work better.


Or else, if the higher-level system is like a human government, then it can change the rules for all. And then monitor the effectiveness of those rule changes.

Applying meta system thinking in biology

“[Consider] the process of evolution through natural selection.

The function-rule (survival of the fittest) is fixed”. (Ashby’s Design for a Brain 1/9)


The rules of organic living are encoded in an organism's DNA. The rules of the organisms in a species are changed from one generation to the next by the fertilisation of an egg. The process of sexual reproduction embodies the “survival of the fittest” rule.

1.     male and female individuals mate

2.     their DNA mixes to form new DNA (think of it as an abstract system description)

3.     the new abstract system is realized by a new individual

4.     the environment favors individuals that make best use of the available resources.

Applying meta system thinking in sociology

Can we stretch ideas about mechanical, biological and psychological machines to the level of sociology? At this level, people act with intent; they envision future changes, make decisions and act accordingly. People are consciousness of the future; they have the ability to invent new system roles and rules.


Typically, attached to a human activity system is a meta system. There is some kind of committee or governing body, which determines the rules. It makes designs changes to the roles of actors, and directs actors in the social entity to follow them. It also has some power to ensure compliance to the rules.


And typically, people are not automatons who inexorably and helplessly play their roles. They are intelligent and creative; they change the rules of an activity system they play a role in. As free agents, they can

·       follow the rules of the system.

·       ignore or break the rules – which may be recognised in the system as an “exception”.

·       propose changing the rules of the system.


Of course, one actor can act in a one role (as a tennis player) in a system of regular activities (a tennis match), and in another role as a law maker, in a meta system (the Lawn Tennis Association).


Generally speaking, an actor can act in the lower system (S) as an actor in its regular operations, and the same actor can act in a higher system (M) as observer of the lower system (S).


How can M change a system from “bad” to “good”. A robotic M may iteratively make random changes to a system, favouring ones that lead an entity to behave better (in some pre-defined way). A human M can observe the system, understand it and invent changes that are likely to make it better.


E.g. In software design, an agile system development method is a meta system. “Stand up meeting” are times when developers monitor and modify the software development system in which they work. Akin to biological evolution, inter-generational changes to the software system are small; and designer and testers eliminate changes that don't work.  Unlike biological evolution, changed systems are fed with all the electricity they need, which can lead to wasteful use of resources.


In the theory of evolution by natural selection, can a social entity be treated as an organism? Might natural selection favor cooperation and oppose competition? Thinkers who addressed this include:

Applying meta system thinking to EA

In business, enterprise architecture (EA) is meta system thinking. The architects work more or less systematically to observe and redefine regular business operations.

On tackling wicked problems

A "wicked problem" is one with conflicting requirements and no ideal solution. How to define and "solve" wicked problems in the most messy of social entities or situations? The first step is not to model a particular system; it is to do some meta system thinking.


This video introduces General Morphological Analysis as a tool to scope a problem and define solution options. GMA is a kind of meta system (see above).


GMA prompts a panel of actors to think, question and consider options in a systematic way. The video suggests these actors are supported and enabled by a software tool. Typically, there are three two-day workshops with 6 or 7 carefully selected and heterogeneous experts. They define the critical dimensions of the problem (cf. the state variables of an activity system). In each dimension, they define a range of options (cf. values for the variable values).


The experts examine each pair of dimensions to decide if it is possible or impossible. Rejecting impossible permutations narrows the permutations to be considered. The many remaining permutations are akin to Ashby’s measure of a system's “variety”. Fixing the choice between options in one or more critical dimensions further narrows the possible option permutations.


This exhaustive and exhausting analysis of (potentially hundreds or thousands of options) can identify unexpectedly viable and non-viable solutions. Assessment of the most favored options is done in whatever empirical, logical, normative (social) ways can be applied.


Further, proceeding to a solution, you might be able to use casual network analysis. Or else Thomas Saaty’s analytic hierarchy process (AHP). And eventually, apply system theory to design a chosen business activity system.


The modelling done by enterprise and business architects can be seen as an application of activity systems thinking and cybernetics. For illustrations of ideas read this slide show. For how social entity thinking relates to EA, read this other chapter.


How EA applies activity systems thinking

EA sees a business as a social entity that may realize any number of (possibly conflicting) activity systems, and strives to coordinate them.

EA is about activity systems that can be modelled, and models that can be realized.

EA cannot define the whole of a business as one activity system.

It can identify where selected parts act holistically in regular or repeated ways.

And identify many such activity systems - nested, overlapping, competing.

It strives to optimize how systems are best coordinated to the benefit of the whole.

EA starts from the aims of system sponsors and other stakeholders, and identifies the desired outcomes, products and services, that currently or should emerge from business activity.

EA sees a business system as a holistic network of activities, which are sequenced in deterministic, probablistic and possibilistic processes (or value streams).

EA is largely not about physical structures (buildings, hardware, vehicles and human bodies) but it does assign activities to logical structures - functions and roles.

EA is about business roles and processes in which active structures - human and computer - create and use passive data structures to store and convey information.

EA requires also some social entity thinking, where logical functions and roles are mapped to physical organization units and actors.


How EA applies cybernetics

EA features business actors that consume and produce information in data flows or messages.

EA is about systems that can be modelled, and models that can be realized as systems.

In EA, the relationship between activity systems and social entities is many to many.

EA is about digitizable business activity systems.

EA is about the regular behaviors in which the range of possible activities is defined.

EA is about systems that maintain data that records the state of entities and events in interest.

EA is about systems that change by state change and by mutation under change control.

EA is a "higher social entity, process or meta system" for organizing the systems of an enterprise.

EA is about systems that remember enough about entities and events to direct them.

EA is about actors and systems that exchange of information. It may resolve discrepancies between languages used in different bounded contexts by standardization by inserting transformers or adapters.


On the terminology torture in EA

A challenge in teaching EA modelling is the terminology torture. Standards architects like TOGAF, ArchiMate, BizBOK and UML are terminology torture on their own - let alone in combination. Authors use ordinary words (like service, process, function, capability, actor, interface and process) with a variety of meanings. For the meanings of these terms and more try these related Linkedin articles. They include a slide show relating EA and BA concepts to general system theory.