Description and reality
A new look at how we know what we know
Copyright 2016 Graham Berrisford. A chapter in “the book” at https://bit.ly/2yXGImr. Last updated 07/07/2021 15:58
The first three parts of this book are largely about the models of reality that we consider to be systems. For example, cybernetics involves modelling the storage and transmission of information, in memories and messages, to describe and direct the state of things.
This final part discusses the acquisition, communication and verification of information from a more psycho-biological perspective. It discusses how we know what we know, and describe things by typifying them in memories and messages. This first chapter of the final part discusses modelling as the basis for thinking. It also details the semantics of the epistemological triangle used in other chapters.
The universe existed for nine billennia before the earth was formed. At that time, there was no knowledge, description or classification of things on earth. The earth rolled on for a while longer without life on it. Eventually, animals evolved to perceive and describe things in the world, because doing that helped them survive and thrive.
Here, thinking is not one particular biological process, but rather any process by which an organism can recognize something in its environment, observe or envisage some phenomenon, create or use a model of some phenomenon.
To recognize (to know) an object, we must have a model of it, which we can correlate (well enough) with what we observe. A model is itself a phenomenon, one that can be correlated with what it represents, near enough for some purpose.
A model need not be complete or perfect. It need only be complete and accurate enough to be helpful. In fact, it is usually very incomplete and somewhat inaccurate. To describe me as 6 foot tall reveals very little about me, but is accurate enough for some narrow purposes.
Throughout this book, the terms model and description are used interchangeably. Moreover, the terms are used at every level of granularity - from a large and complex specification of a steam engine - to a small and singular concept, property, feature, attribute or characteristic of an individual (like 6 foot tall).
Wholeism (considering every conceivable aspect or element of a real-world entity) is impossible. Even a grain of sand is beyond our full comprehension. All we can understand are models we abstract from phenomena we observe and envisage.
Thinkers <observe & envisage> Aspects of reality
Thinkers <create and use> Models
Models <represent> Aspects of reality
That is not to say (as some relativists or perspectivists do) that “perception is reality”, or that all models of a real-world entity or situation are equally true or useful. How closely a model corresponds to reality, should be testable and verifiable/falsifiable. Still, two different models may both be useful in different contexts. For example, light may be modelled as waves or particles (photons).
“The question touches on a very sensitive topic for those who follow science. The answer will quickly boil down to frustratingly phrased questions like "what is real" because what you ask is sufficiently tricky.
In the world of philosophy, science is categorized as part of "empiricism." Empiricism is the philosophy of what we can know using our senses (aka "empirical observations"). Empiricism is a subdiscipline of epistemology, the study of what we can know. Epistemology is separate from ontology, the study of what is "real," so from a philosophical perspective science does not actually make statements about what is real and what is not.
When you look at the evidence for the "photon," what is provided is a large body of empirical observations done by scientists. In each case, we find that if we model the world as though photons exist, the results of the experiment are consistent with the predictions of that model. It doesn't say that photons exist, and it doesn't say that they don't exist. It merely says that a model which declares that photons exist is effective at predicting the results of past experiments.
Science has shown that its models have an astonishingly good track record of predicting results. I would argue its results are far better than any other system out there, although to make such a claim we would have to sit down and agree on a metric for comparing systems first. If you need to predict what is going to happen in a system, modeling the system in a way that includes photons is more than likely to give you a solid answer. Science prizes itself on its ability to make predictions no other system can make, and make good on those predictions.
Science has been so effective at this that we start to get lazy with our terminology. We start to say "photons exist" or "photons are real." At the philosophical level, this is called abduction. Abduction is in the same category as deduction and induction; it's assuming that the most likely hypothesis is true. We assume that, without any better hypothesis, photons must simply exist. This is not part of the scientific method; it's not an empirical claim. However, the models are so bloody good at predicting the future that the more lengthy phrase "the universe is well modeled with the assumption that photons are real" is just not worth the extra breath it took.
Unfortunately, this abductive step can get us in trouble. You mentioned you've heard of photons as being described as "probability clouds." This is because, as we peered deeper and deeper into the universe, we noticed that modeling things as photons stopped fully describing what we saw. Strange experiments like the double-sit experiment started to suggest that we can't just model light as a bunch of individual photons. Experiments like that one were designed intentionally to push at corner cases where the old models simply broke down. We then replaced them with new models which are much better at predicting results.
Of course these new models need to line up with all of the old empirical evidence gathered before this new model was formed. As we step away from the difficult corner-cases proposed by the double-slit experiment, we find that there was a strong connection between some of the probability distributions which arose from the quantum mechanics wave functions and the "photons" that were assumed in the old model.
This, to me, is one of the brilliant parts of science. By making these connections between the models, we can say "as long as you stay away from these particular corner cases, you can get away with modeling light as photons, because the errors you pick up are small." Think about how amazing that is. We're comfortable enough with our models to say "even if you aren't using the most advanced model science has to offer, we can still safely use it and put limits on the errors."
Regardless, in quantum mechanics, at its deepest level, there are no "photons." There are waveforms which are continuous through space and time. However, in many, many cases, the behavior of these waveforms is discrete enough that we can capture part of the waveform and say "this part is a photon." But it's really our decision to call it a photon. Empirically speaking, it's a good decision. In well over 99% of cases, its a good enough decision to make predictions, and that's what we want from science.
So do photons exist? Nobody knows. Our best models of the subatomic world, those of quantum mechanics, all show behavior that's right in the place where photons should be. We have no particular reason to assume they don't exist. Whether that is enough for you is really a question of personal preference and philosophy.”
A type is a concept, is a model. Induction is a process by which a type may be abstracted from observation of resemblances between phenomena. Induction is an important ingredient of what we call intelligence. That fact that intelligence involves induction (also deduction and abduction) is important here; how the processes work is not important.
A mathematical type might largely and accurately characterize an individual mathematical entity (say an individual of the “triangle” type). By contrast, a natural language type may be a very incomplete and only somewhat accurate characterization of a physical, biological or social entity (say an individual of the “tennis club” type).
As the philosopher A J Ayer pointed out, a description (if not pinned to one individual in time and space) typifies what it represents. A description of one apple applies to every other apple that matches the same description. Even if we see only one member of a set (one apple, or one universe), we can, from its description, envisage more members. And even if we describe only one mythical unicorn, we can envisage many of them.
So, a description is model, is a general type or intensional definition. It sets out descriptive properties shared by each member of a set of near-enough similar things. Not only can one type describe (characterize, represent, typify) many individuals, but also, one individual can be described by (embody, exemplify, exhibit, instantiate) many types.
The chapter on typification explores the topic further.
“A biological approach to human knowledge naturally gives emphasis to the pragmatist view that theories [models of reality] function as instruments of survival.” Stanford Encyclopedia of Philosophy
Descartes is famously said to have started his philosophy from the premise “I think therefore I am”. Psycho-biologists presume more. They presume that animals, which occupy space and live for a period of time, can perceive phenomena, remember them and communicate about them. Implicit in those premises is the idea that there was no model of reality before life.
The Darwinian view of thinking is that both memories and messages are biological phenomenon that evolved to help animals survive. Remembering and sharing models helps animals to understand, predict and manipulate things in reality, and so, improves their chance of reproducing.
The good regulator theorem (Ashby and Conant) declares that a good regulator is a model of what it regulates. Here, a regulator is an animal or a machine that either has, or has access to, a model of the target that it regulates.
Regulators <monitor and regulate > Targets
Regulators <have and use> Models
Models <represent> Targets
To function and respond to changes in its environment, an animal must “know” what it going on in its world. It needs a model of entities and events its environment if it is to find food and mates, and avoid enemies. The more intelligent the animal, the richer its model. (Similarly, a business needs to know the state of things it seeks to monitor or direct.)
The question is not whether an animal (or a business) has a model of its environment; it is how complete and accurate is the model? To which the answers might be both “very incomplete and somewhat inaccurate” and “remarkably, complete and accurate enough”. Demonstrably, animals remember and share knowledge, like where food can be found. But a model need not be complete or perfect. It need only represent things well enough. A cat remembers a mouse’s features well enough to spot and catch mice. A honey bee remembers the location of some pollen well enough it can direct other bees to that location.
A model must correlate in some way to the described phenomenon.
· An iconic model, like a statue or photograph, mimics features of the described phenomenon, and is recognisable using the basic senses.
· An indicative model, like the smoke of a fire, points to effects produced by a described phenomenon.
· A symbolic model encodes some features of the described phenomenon using a code that can recognised by an animal or machine that knows that code.
Our main interest is in symbolic models, and by default, you can read model as symbolic model.
This section shows how thinkers in different domains of knowledge describe things.
Cartographers create maps (models) of territories (described things). The map is not the territory, but the two must be correlated well enough to help map users find things.
Mappers <observe & envisage> Territories
Mappers <create and use> Maps
Maps <represent> Territories
Hmm… Is the territory real? Or merely another a mental model we form of the territory as we travel through it? It doesn’t matter to us, as travellers; it only matters that we find the map useful when navigating through a territory.
Building architects define the structures of buildings (described things) in architectural drawings (models).
Architects <observe and envisage> Buildings
Architects <create and use> Drawings
Drawings <represent> Buildings
Whether the drawings of a building are best called architectural, or blueprints, or detailed designs, is a subjective decision. Other chapters address the topic of system architecture in some depth.
Have you heard that Einstein didn’t shine at school? The truth is, his insights were deeply informed by extensive study of mathematics, which he mastered at a young age.
"Max Talmud says that after he had given the 12-year-old Einstein a geometry textbook, after a short time he “had worked through the whole book. He thereupon devoted himself to higher mathematics... Soon the flight of his mathematical genius was so high I could not follow." His passion for geometry and algebra led the 12-year-old to become convinced nature could be understood as a "mathematical structure". Einstein started teaching himself calculus at 12, and as a 14-year-old he says he had mastered integral and differential calculus". Wikipedia
Physicists <observe and envisage> Nature
Physicists <create and use> Mathematical structures
Mathematical structures <represent> Nature
Throughout his life. Einstein distinguished his mathematical models from the reality of the universe, and declared that it is impossible to directly comprehend the reality.
“There are several kinds of theory in physics. Most of them are constructive…When we say that we understand a group of natural phenomena, we mean that we have found a constructive theory which embraces them.” Albert Einstein, The Times (28 Nov 1919)
“As far as the laws of mathematics refer to reality, they are not certain; and as far as they are certain, they do not refer to reality.” Albert Einstein, Sidelights on Relativity (1920), 28.
“Physical concepts are free creations of the human mind, and are not, however it may seem, uniquely determined by the external world. In our endeavour to understand reality we are somewhat like a man trying to understand the mechanism of a closed watch. He sees the face and the moving hands, even hears its ticking, but he has no way of opening the case. If he is ingenious he may form some picture of a mechanism which could be responsible for all the things he observes, but he may never be quite sure his picture is the only one which could explain his observations. He will never be able to compare his picture with the real mechanism and he cannot even imagine the possibility or the meaning of such a comparison.”
Albert Einstein, The Evolution of Physics (1938), 33
Mathematicians define a type (like “point”, “line”, “triangle” and “even number”) to describe the properties of a thing that instantiates that type.
“a number divisible by two”.
A type is a pattern for (one member of) a collection or set. An instance is an exhibition or embodiment of a type in a described individual – a set member.
Mathematicians <observe and envisage> Set members
Mathematicians <create and use> Types
Types <are instantiated in> Set members
Two mathematicians had a famous debate about the meaning of descriptive types or axiomatic assertions. The debate might be distilled as follows.
Frege believed mathematics is carried out at the level of thoughts about real things, rather than models of them. He believed models are imperfect representations of our thoughts. We “know” what a geometric entity (like “point”, “line” and “triangle”) is without reference to a model of it, or how it relates to other entities. And every detail of it may be relevant to answering questions about it.
Hilbert said that, even if we do know other details of an entity (say, its color), they are entirely irrelevant to understanding geometry, which merely defines relations between entities. Geometric models are holistic in the sense that they define how geometric entities are related, rather than what those entities are or contain; they can be anything (large or small) that is relatable as described. Mathematics is carried out at the level of models, regardless of what the entities are in reality.
According to the Stanford Encyclopedia of Philosophy, Hilbert is now regarded as the winner of the debate. Geometry does not address the whole of a thing (Frege); rather, it addresses only those features of it that are describable by geometry (Hilbert).
Thinking creates and uses models that represent phenomena that thinkers (be they honey bees or humans) encounter. A model is a description that typifies what it models.
If the way of thinking we call system thinking is more than thinking, then what we call a system must be more than any old model we create or use. Whatever the system model is used for, it is impossible to duck the question: What characteristics differentiate a system model from any old model?
Ashby’s cybernetics is about modelling the behavior of a real machine as a system.
Observers <observe & envisage> Real machines
Observers <create and use> Abstract systems
Abstract systems <represent> Real machines
A real machine can be only rightly be called a system in so far as it is correlated with an abstract system. The remainder of the machine is either beyond our ken or not described as a system.
Social systems thinkers model the behavior of a social entity as a system.
Observers <observe & envisage> Social entities
Observers <create and use> Abstract systems
Abstract systems <represent> Social entities
A social entity can be only rightly be called a system in so far as it is correlated with an abstract system. The remainder of the social entity is either beyond our ken or not described as a system.
Some definitions of system (inter-related parts, or interacting actors) are so general they are vacuous. Some definitions (complex non-linear relationships or loops) are ambiguous, unclear or very restrictive. Better definitions are found in system dynamics, cybernetics and soft systems methodology. These three approaches all view an entity as changing state and/or transforming inputs into outputs, in rule-bound ways. Even simple entities, modelled thus, may exhibit “non-linear” lines of behavior and "adapt" to changing conditions.
Enterprise architecture does not address the whole of an enterprise; rather, it addresses selected business operations. It is concerned with business activity systems in which business roles and processes create and use business data.
Architects <observe and envisage> Business operations
Architects <create and use> Abstract systems
Abstract systems <represent> Business operations
Architecting is a process that encodes, decodes and compares models of baseline and target systems. At design time, general entity types like customer, event types like order and payment, and process types like billing are defined. These types are related to each other in logical data structures and process structures. E.g. order value = order amount * unit price. At run time, physical systems consume inputs that carry information about entities and events, record those entities and events with attribute values, and use remembered information to decide how to respond to events. In all these ways, an enterprise can be seen as imitating what comes naturally to animals.
Thinking involves observing and envisaging phenomena. It involves conceptualizing phenomena - creating, using and comparing models of phenomena in terms of discrete entities and events, actors and activities. It embraces identifying (by induction, deduction or abduction) rules that govern activities, including the laws of physics, chemistry, biology, psychology and economics
Epistemology is about what we know of reality, through observation, modelling, testing, reasoning and learning from others. The constructivist epistemologist distinguishes models (constructed) from described things (observed or envisaged).
We have no direct knowledge of things in reality. We know them only via models of them that we construct (in the mind, in speech, writing, mathematics, whatever), which are (ultimately) associated with our sensations of the world.
We (thinkers) can only understand physical reality in so far as it is correlated with a model we have access to – which we can correlate well enough with real-world phenomena for practical use.
A thinker is any animal or machine that can observe or envisage something, and create or use (encode or decode) a model of it. A phenomenon is any aspect or part of reality that a thinker can observe or envisage. A model is a representation (in mind, speech, writing or other form) that is correlatable with a phenomenon.
The three concepts are relatable thus:
<create and use> <represents>
Thinkers <observe and envisage> Phenomena
Note that each pair of concepts is related in a many-to-many way. One thinker can create several models of the same thing. Conversely, several thinkers can contribute to one recorded model of the same thing. It may well be that none of those thinkers knows whole model, which can be larger, more complex, consistent and informative than any human memory can be.
The view of model and reality above seems obvious to some and obscure to others. The semantics of the triangle are explored below.
Models <represent> phenomena
Models embrace every kind of model that an animal or machine can make of the world. It can be a mental or documented model, in a memory or message, in speech or writing. It can be a 2D picture or 3D model, in the brain or in stone. Here, we mostly discuss models expressed using words and other symbols.
Phenomena are everything in reality that can be observed or envisaged, every kind of entity and event, include including models and thinkers.
Two models of the same phenomena may be compatible or in conflict. Is light rightly described as waves or particles? Physicists do not say either model is the “true” model, they say only that each can be useful.
Thinkers <create and use> models
Thinkers are animals or machines that can encode and decode descriptive models of phenomena. Models are given meanings in the acts of encoding or decoding.
We speak of the thinker in the triangle as an animal or a machine. Strictly, it that actor’s cognitive processes – its ability to observe and envisage, and create and use models at the apex of the triangle.
In short, to create a symbolic model is to encode a model that represents some feature(s) of a phenomenon. To use the model is to decode it, then use it to respond to or manipulate whatever is described.
Thinkers <observe and envisage> phenomena
To observe or envisage things is to create and use models of them.
Both thinkers and models are also physical phenomena. In observing a concrete house, a thinker is creating and using a model of that house, which exists in a physical form, and can be translated from memory to message and back again. Similarly, in envisaging a purely fantastic thing, such as a flying elephant, a thinker is creating and using a model of it, in some physical form.
It is difficult for thinkers to observe the largest known prime number, because it is so long; still, it exists in the memory of a computer. And thinkers can envisage the next prime number (beyond today’s largest known prime) as a future instance of the abstract type the thinker has in mind or in writing. In short, to observe or envisage things is to create, use and compare models of them.
Thinking involves acquiring knowledge (through sensing and thinking) and capturing it in models of the world for later use.
The psycho-biological view of thinking is called “cognitive embodiment”, meaning that the brain is inseparable from the rest of the nervous system. Thinking is not done in the brain’s grey matter alone. Below, distilled from the next chapter, are some staging posts in the evolution of human thinking.
In molecular memory, organisms sense and respond to molecular structures.
In primitive animals, sense-response thinking is an end-to-end input-to-output process (cf. a value chain in business architecture). It begins in perception by a bodily sensor, where the state of things is first described by encoding a neural message; and ends with a direction to the body’s motors and organs, where a message is decoded and used in actions to change the state of things.
In higher animals, with memories, perception is a mix of observation (sensing) and envisaging (guessing) at what there is out in reality. Both observation and envisaging are processes that create (encode) and use (decode) models in memory.
Animals cannot not only inherit and remember knowledge, and learn from experience, but also communicate knowledge. Even very primitive animals signal mating intentions to each other. Other early social acts were related to marking territory, signalling danger and locating food. E.g. Cats spray scent to mark their territory and other cats smell that scent. By 100 million years ago, some animals had evolved to cooperate in groups by communicating models of things to their fellows.
Every remembered model of the past serves as a type that defines (one member of) a set that may contain more members in future. Envisaging the future involves creating and playing with models of possible futures. Consciousness enables an animal to compare models of past, present and possible future phenomena.
Speech and symbolic language
In humans, memories are translated into and out of verbal messages for communication. Ashby observed that in thought and communication “coding is ubiquitous”. The multiple translation steps involved in social communication are illustrated in the next chapter.
In writing, humans recorded ever more complex models in a persistent and shareable form.
The proposal here is thinking involves encoding, decoding and comparing models. Moreover, in humans at least, this is process is not entirely internal. I observe my own thinking process is partly externalised via writing. It involves a circular feedback loop in which I encode some half-baked thoughts in written words, then read back what has been written, test it for consistency and coherence, and correct or clarify it. Writing makes it possible to document a model of the world that is more complete, consistent and coherent than any I can hold in mind.
Moving data from the brain, to the hand, to the keyboard, to the screen, to the eyes and back to the brain, is a succession of processes that translate data encoded in one form to data encoded in another. In short, thinking involves observing and envisaging - creating and using models - encoding, decoding and comparing models, both internal and external.
Reasoning and science
By reasoning, humans refined how to form theories, predict outcomes and test them.
Reasoning (like other human abilities) evolved because it helps a human to survive and thrive. Consider for example this reasoning: "if we drive the buffalo over the edge of a cliff it will kill them, and provide us with the meat and leather we need." Reasoning helps us to predict how things will turn out, generalize from observations, abduce rules of behavior, form logically coherent descriptions of reality, form theories and test them, communicate with other humans, design technologies, etc. All of which gives one human being, and others in the same co-dependent social group, an advantage over other individuals and groups.
In machine learning, human-created machines abstract models from phenomena.
AI software can now abstract types/patterns from things it observes as sharing features.
Human and computer actors process data in different forms and ways. A human’s memory, being encapsulated in a brain, is inaccessible to others (unless translated into a message). By contrast, in software, memories and messages are merely varieties of digital data storage; both can read/written by many software components.
Note also that software can read a memory without creating one. By contrast, to read a human a memory is also to create a new one. (Which is one reason why human memory is less reliable, as in “false memory syndrome”.) Nevertheless, both human and computer actors can be reasonably be thought of as writing and reading – creating and using - memories and messages.
The proposal here is that to observe or envisage things is to create, use and compare models of them. Our position has been set out in this chapter, and can be represented at an overview level in this triangle
<create and use> <represents>
Thinkers <observe and envisage> Phenomena
There is an essential difference between our triangle and comparable triangles you can find in semiotics and philosophy. Most other triangles separate models into two kinds: internal mental models and external models in speech or writing. By contrast, our triangle separates thinking processes (observing, envisaging, remembering, recalling, writing and reading) from all the descriptive structures those processes create and use (in the mind, in speech, on the page, wherever). So, a model in the mind is at the apex, not the left.
Philosophy has moved over many centuries from a position in which models perfectly capture the real nature of things (Plato), to positions in which models are, at best, approximations to things. Some propose models are convenient fictions to organize sense data. Others propose all models are untethered from external reality, and equally justified. To decide where you sit in the mess of countless philosophical positions leads you into the endless morass of philosophical debate, which you can find in the Stanford Encyclopedia of Philosophy.
This book may be read as endorsing three philosophical positions. Instrumentalists say models are instruments of prediction. Pragmatists say models are concepts or artifacts used in producing scientific knowledge. Constructive empiricists say models are symbolic representations of empirical phenomena.
Some say these three positions are “anti-realist”, meaning they deny that models give a true model of reality. Here, it seems futile sophistry to deny any of the following
a) Reality does exist.
b) A model can be true empirically (enough to be useful) or logically (a consequence that follows from some axiomatic assertion)
c) We can share models and so share some knowledge of reality.
Our main interest is in symbolic models, which are encoded using symbols (or signs) and in how animals and machines create and use them to describe things. In our epistemological triangle, we could replace model by symbolic representation.
Chapters after this explore the evolution of thinking, the use of types to describe things, how we share knowledge and verify truth, and challenge alternative views of model and reality and philosophical positions.
EA is much about modelling human and computer activity systems. See part one of the book.