Heuristics: from Artificial Intelligence to Practical Wisdom

Artificial Intelligence is is still not capable to React to unknown Situations.

AI-systems need Humans to adapt.

Human Heuristics are much smarter than AI-systems.

Wisdom is Accumulated Reflection on Experience.

For a Quick Start push this link or start with the Introduction.

0. Introduction

This blog is a follow-up on my blog about Heuristics.

In this blog I analyze the human Cognitive Architecture and its relationship with the architecture of our fellow-Organisms including the bacteria and viruses.

bacterium
Bacterium

Bounded Rationality

I also relate this architecture to the architecture of Artificial Inteligence, a result of Bounded Rationality , a theory that was taken over by Ecological Rationality also called Heuristics.

Heuristics are the “positive” part of the Human Bias, a problem we encounter when we are put under Stress and have not enough Time to Reflect.

Organisms

Organisms are Moving Memories exploring their environment to satisfy their need for the Energy (chemicals) they need to Survive.

To stay in de right place they move to the place their mover (“flagellum”) sends them to.

In between sensor and mover they are a short term Memory that is created out of the molecules they take in at the front.

We are auto-poietic auto-catalytic systems Born in the Sea.

The difference between the bacterium and us is called Higher Order Embedding meaning the same principle is applied to Itself, creating a Hierarchy.

This hierarchy is really a very complicated Network that contains many sub-structures we have Named using a principle that was a in Tune with the Spirit of the Time.

Narratives and Myths

This resulted in a lot of Narratives that turned into Myths that are summarized by Joseph Campbell i(a follower of Carl Jung) in the Hero Quest that starts with a Call to Adventure.

Currently many Succesfull movies are based on the Hero’squest with Star Wars and Avatar as an example.

The hero’s Quest is an old Cyclic model that was projected on the Zodiac.

We are now moving from Pisces to Aquarius meaning from Dualism to Holism.

The very longterm Precession cycle is moving Against the Clock of the yearly Sun Cycle meaning the we are now at the Start of New long term cycle meaning a new Call to Adventure for Mankind.
1192px-Zodiac_(PSF).png
The cycle of the Zodiac.

What are Heuristics?

When I was re-searching for ny blog about Heuristics I found the thesis of Sheldon J. Chow, called “Heuristics, Concepts, and Cognitive Architecture: Toward Understanding How The Mind Works”

Heuristics are cognitive procedures that satisfice, and that require little cognitive resources for their recruitment and execution; they operate by exploiting informational structures, called  Perceptual Symbol Systems (PSS).

Exploiting Perceptual Symbol Systems

Concepts are highly organized collections of linguistically coded perceptual information.

Our Cognition looks lke a Video-recorder connected to our real Emotions and Bodily-processes that are retrieved when certain concepts (Anchors) or Situations are recalled.

Concepts are Simulators that embody Perceptual symbols that are collections of neurons that are activated in the perceptual centres of the Brain.

Perceptual states arise in the Sensory-Motor system.A subset of the state is extracted by selective Attention and stored in long-term memory.

This perceptual memory can function as a symbol entering into symbol manipulation.Collections of the perceptual symbols comprise our conceptual representations.

The structure of a perceptual symbol corresponds (at least somewhat) to the perceptual state that produced it.
A Conceptual Model of a Perceptual Symbol System being a part of the PDF below.
Figure 3. (A) An example of establishing an initial frame for car after processing a first instance. (B) The frame’s evolution after processing a second instance. (C) Constructing a simulation of the second instance from the frame in panel B. In Panels A and B, lines with arrows represent excitatory connections, and lines with circles represent inhibitory connections.
An example of establishing an initial frame for car after processing a first instance. (B) The frame’s evolution after processing a second instance. (C) Constructing a simulation of the second instance…

Relevance Theory And the Frame Problem

The Thesis contains a reference to two for me interesting and highly interrelated Philosophical Questions called the Frame and the Relevance problem which is about determining what is and is not relevant in our cognitive processing or What do we Know when we Sense a Situation.

Wisdom

Another result of my search was a very impressive You Tube playlist “John Vervaeke on Wisdom and the Meaning Crisis” that contained the same two subjects and a link to the concept of Wisdom being the result of Insight.

The egyptian goddess of Wisdom was called Maat . She represents the Concept of Harmony.

The Feather of Maat is a Measure that has a value of almost Nothing.

The Heart is the Pump of the Up (Brain) and the Down Circulation of the Body.

Both parts are in Balance when their Sum is Zero.

The Ape represents the Human.
Two four-fold cycles connected at the Connection point (green) of both Cycles.

Left a comparable model where the parts (now called tatva’s) are called chakra’s (meaning “wheels“).

Every part of this model is able to turn into a new cycle until a new “emergent” level is reached.

Reading List

0. Introduction

In this blog I compare the human Cognitive Architecture with the general Architecture of Organisms and detect many things that Explain our behavior.

What are Heuristics?

I found a thesis about Heuristics and the concept of Perceptual Symbol Systems being hierarchical semantic models linked into a Semantic Network.

The Semantic network not only contains perceptual symbols but also Emotions, Patterns, and Intuitions, being Wisdom Gathered in many generations and transmitted by the epi-genes of the mother and the father.

In the thesis I also found the Frame -Problem and Relevance theory. Both are a result of the big problem AI-systems have to look like humans.

1 Interpersonal Theory : Humans are dialogical beings.

They comunicate internally and externally with many voices .

They expect that the sender is providing meaning but their expectation is related to their Point of View.

People that have only one Point of View are disconnected because every View is Independent (Making an angle of 90 degrees) from all the others.

Normal communication takes two complemenatary views into account.

2. What is the Architecture of the Human Mind,

What is the link between AI and the Brain.

what is Innate ? What architecture do we share with the organisms?

3. Relevance Theory:

What is important in a conversation? It has to have just enough information, with the right quality in the right manner,

4. The Frame Problem: What do humans really know? What is Intelligence and therefore Artificial Intelligence?

5. general Semantics and other AI Architectures.

: How to prevent that you are manipulated by language? Alfred Korzibski deloped a theory called General Semantics that can be used to create AI that is able to Adapt.

A description of many Other AI-Architectures shows that they are all based on Bounded Rationality.

6. From Hierarchy to Network:, What is the principle behind life itself?

About Autocatalytic and Autopoietic systems.

What are Anticipatory Systems? , To be an anticipatory system it has to contain a model of the outside world.

Life is an expanding system in which all te possibilties of a very basic simple model are created until the basic model innovates itself.

Gategory Theory: Is it possible to define theories as Functions?

What are Meta-Meta-systems? Is it possible to expand a System in all directions?

7. Meta-heuristics The Thesis of Seldon J. Show is about Reasoning about Heuristics.

Meta-heuristics.has become a new field of science,

It looks a lot like a Merger of Psychology with Operations Research (Mathematics).

8. From Biophysics to BioMathematics.

9. Rodney Cotterill: The Material World.

10. Conclusion: How to design a modern Expert System:

1. Interpersonal Theory:

Paths of Change is a Scale Free (fractal) theory of Change developed by Will McWhinney.

Relevance theory has a lot in common with Interpersonal theory and the theories of Bahktin about the Utterance or the internal and external Dialogue we take part in.

Interpersonal theory assumes that humans expect meaningful information from other humans and from the humans that are inside. We are always in a dialogue.

It also assumes that a mental illness can be explained by a connection-problem. It is all about the Tension between Agency and Communion.
Interpersonal Circumplex. The interpersonal circumplex is defined by two orthogonal axes. In recent years, it has become conventional to identify the vertical and horizontal axes with the broad constructs of agency and communion.
Human interaction can be modelled with two orthogonal (90 degree angles)variables that were also called yin (communion) and yang Agency) or o en 1 or just yin and not yin. The four 90 degree angles support a 4 dimensional Universe of Discourse that contains rotating rotations being Spirals.
Humans learn when an Expectation is violated. This happens when an External or Internal Event takes place.

The event is evaluated by the Emotions and (when there is enough time (and interest and energy) Reflected upon by the Imagination and Abstracted (moved Up) into a Model (frame, scape)that is again Instructed and or Tested.

The Learning cycle moves with and against the Clock and also moves through the diagonal between Belief and and the Prove of Belief and Dependence and in-dependence.
This picture describes the model of Nico Frijda about the Emotions: the readiness (or unreadiness) to enter into contact or interaction with some object, and the mode of that contact or interaction. [ … ] Different motivational states or states of action readiness exist … Emotions are tendencies to establish, maintain, or disrupt a relationship with the environment.
Reflect: Finding a pattern in the many actions the senses detect.

2. Our Cognitive Architecture is shared with the Organisms

This part is about the architecture we share with all of the Organisms.

It is mainly based on earlier work of Rodney Cotterill and Will McWhinney, both diseased.

Peter Carruthers: Innate Human Cognition. What is the Architecture of the Mind?
Modular architecture of Peter Carruthers

The Self-Assembling Brain

“our genome contains the information required to create our brain. That information, however, is not a blueprint that describes the brain, but an algorithm that develops it with time and energy. In the biological brain, growth, organization, and learning happen in tandem. At each new stage of development, our brain gains new learning capabilities (common sense, logic, language, problem-solving, planning, math). And as we grow older, our capacity to learn changes”

Every Organism is a step in a very long term process we call evolution in which the outside-world is internalized and externalized in a new outside-world. The genome reproduces a body that is able to reproduce a body but also a system that is able to reproduce and adapt itself.

The Basic Architecture: We are Moving Memories

The picture shows the most simple structure of an organism being a Motor and a Receptor connected with a Short term Memory.

The bacterium is Exploring its environment by changing the rotation of its flagellum with and against the clock.

The short term memory is an autocatalytic, chemical reaction that determinates not only the Rythm of the rotation but also the Direction of the rotation of the rotor.

Rodney Cotterill’s neuro-physiological work supports the motor-environment-mind schema of complex feedback loops. The complex of feedback loops enables one to “Know what one knows.”:

The ability to know that one knows is referred to by psychologists as first-order embedding.

Higher embedding, such as that exemplified by knowing that one knows that one knows, merely depends on the ability to hold things in separate patches of neuronal activity in working memory.

This manifests itself in the creature’s intelligence, which also dictates its ability to consolidate existing schemata into new schema.” is a unique feature of mammalian brains.

Intelligence and reflection are explained as due to functioning of the web of feedback loops, both internal and external, that evolved in higher animals“.

3. Relevance Theory

The word Relevant comes from Latin relevare “to lessen, lighten, levis: light.

Relevance theory is a framework for understanding the interpretation of utterances.

It aims to explain the well recognised fact that communicators usually convey much more information with their  utterances and their Body than what is contained in their literal sense. 

Our Cognitive Architecture Supports Collaboration:

Daniel Kahneman and Amos Tversky used “cognitive biases,” to show that humans make irrational choices.

One of these biasses is called the conformation bias. Humans love to hear a confirmation of what they believe.

Lately a new “science” called Evolutionary Psychology” is trying to find out what “functions” helped the human to survive.

Historically, evolutionary pressure has resulted in cognitive systems that recognise potentially relevant stimuli and try to draw relevant conclusions.

Dan sperber and Hugo mercier argued that this type of bias, being a part of communication, is supporting collaboration.

To Attract the Attention the “uttering” has to be Relevant, making things easy, simple, light.

They also looked at the amount of Energy it takes to communicate.

Relevance theory started with the research of Robert Grice. The Gricean maximsQuantity (Just Enough), Quality (True), Relation (Relevant)), and Manner (the Right Way) describe the principles observed by people in pursuit of effective communication.

Argumentation ?Theory tries to show the evolutionary benifit of many cognitive biasses.

The Perceptual Symbol Systems are so rich in the type of connections that they generate free Associations that keep a group in the right direction making it possible to Nudge a group by giving them only one one choice into the desired direction. This is comparable with a vector-field.

Conclusions Precede Arguments:

Argumentation theory and Rhetorics. We use arguments to convince others but also to convince ourselves.

We are able to use many different cognitive tools (modules) at the same time without knowing we do that.

We use generalizations and are not aware of the differences between the generalisations.

We love Dopamine and therefore love to act in situations were dopomine is produced (“gambling).

There are two systems active the Controlled (Dopamine) and the Control system (Observer).

Conclusions come first, arguments later.

We always pick the last we hear and find the arguments to justify our decision.

Falsification avoids bias.

We are social animals and our brain has to support cooperation.

We could benefit from deceiving but we have to keep our Status.

Gossip supports our status.

We have to Sell ourselves.

Conformation Bias helps to do that.

Humans are Curious Toolmakers .

Coding vs Decoding. Humans are Curious. We are tool makers improving our tools. the older the more knowledge and problems to retrieve out of memory. Attention is given to something that fits what we want know and is relevant. Time and Energy are always Limited.

Our Communication is mainly Implicit

Maximilisation of relevance.Balance between Energy and Cognitive-Contextual effects. Assesment is looking for relevance.

The difference between the foreground and the background of comunication or the difference between Explicit and Implicit, figurative language such as hyperbolemetaphor and irony.

4. The Frame Problem:

The frame problem is about What a “system” is able to see in a Situation and what changes what (“an Event“) and what is Changed and is Not Changed.

The frame problem was created when IT-technology moved into the area of Artificial Intelligence.

At that time every “problem” was solved by designing a new Language like LISP and PROLOG.

Logic (“rules“) put into a program was leading the way and logic was guided by Set-theory.

Although Chomsky’s Language theory was already there nobody realized that human Language is very diferent from a Progamming Language that is Context-indepentedent.

The Artifical Intelligence of the Computer does not understand what was out there and tried to catagorize the outside by also define the not-categories not realizing that they were not bounded (infinite).

Until now the AI-computer looks like a savant a highly specialized human that is in an expert in one area but incapable in the Infinite area’s outside its speciality.

It is not able to define what is relevant and does not see the difference between implicit and the explicit conversations. It is also not able to interpret the many illogical signs we send with our Body, Facial Expressions and Manner until Geroge Lakoff and his group of scientitst detected that we the human communication is embodied.

5. General Semantics and other AI-architectures

This chapter contains many AI-architectures even a model based on General Semantics at the end.

As you will see all the models are different from the model that is explaining heuristics in the Introduction.

The main reason is that the models are based on the theories of bounded rationality that existed before the time of Herbert Simon and Daniel Kahneman, both winners of the Nobel Prize in Economics , desribed in my blog about Heuristics.

5.1 General Semantics

Alfred Korzibski was forced to fight in the first world war. He did not understand why humans act like cruel animals. After moving to the Us and marying he started to research human Communication and developed the theory of General Semantics.

General semantics is concerned with how events translate to perceptions, how they are further modified by the names and labels we apply to them, and how we might gain a measure of control over our own responses, cognitive, emotional, and behavioral.

General semantics had a lot of influence on cognitive-therapy including NLP.

The model of Korzybski looks a lot like the model of Interpersonal theory hat describes Mental Illness.
The model of Interpersonal based on the model of Paths of Change. Return to the Start of this blog.
General semantics explained by Korzibski himself.
Korzibski designed a model of Model of the Nervous-System

5.2 Artifical Intelligence Architectures:

When you press the link above you will find an impressive website about everything you want to know about AI-architectures. Below you find a few examples I have taken out of this website.

5.3 The LIDA Cognitive Cycle Architecture 

5.4 CYC : understanding common Sense:

CYC

5.5 AIXI Optimizing Reward:

AIXI is a AI-theory developed by Marcus Hutter. IXI is a reinforcement learning agent. It maximizes the expected total rewards received from the environment. Intuitively, it simultaneously considers every computable hypothesis (or environment). In each time step, it looks at every possible program and evaluates how many rewards that program generates depending on the next action taken.

5.6 CALO The Perfect PAL:

the DArpa CALO-project ( Cognitive Assistant that Learns and Organizes ).

5.7 ACT-R A new AI Program Language:

ACT-R is a cognitive architecture: a theory for simulating and understanding human cognition. Researchers working on ACT-R strive to understand how people organize.

5.8 AI Architecture based on General Semantics:

6. From Hierarchy to a Network of Hierarchies

6.1 Organisms are not Machines

In the history of the Cognitive Architectures of the Organisms two things happened:

(1) the processes of life became auto-catalytic meaning that a chemical process became a repeating cycle (auto) that was sustained and became independent of other processes.

(2) the cycles got connected and created bigger cycles without breaking the katalytic mode until one of the cyclic cycles was able to look like an independent entity.

A cycle that reproduces itself is called an Autopoiesis, a Self-reproduction.

Below you can watch 4 video’s that give examples of autopoietic auto-catalytic systems and the big change in thinking they accomplished.

One of the major Scientists was Robert Rosen: See chapter 5.2.

He tried to explain Anticipatory Systems and found out that it is impossible to explain them with our current modelling language that supposes that Nature is a Machine made by a (Divine) Architect (Pthah) that got the Specifications from Ma,at (responsible for Balance, Harmony) and gave them to Khnum, the divine potter, to Make the animals (including the Humans) out of Clay.

statue of egyptian god ptah
Ptah, the Divine Architect.
Observe that Pt Ah is giving the Breath of Life (Ankh)into the Mouth at the beginning of Spinal Column, the Djed-Pillar.

5.3: Dynamic Networks.

5.5: Rosen also tried to create a new type of Mathematics called Biomathematics using a new meta-mathematical theory called Category Theory. In Category Theory theories are functions called functors.

5.6: is What are Meta-models?

Fritjof Capra talks about the Web of Life.
Autopoietic Systems are Self-Reproducing.
Auto catalytic systems explain the Origin of life
Mihai Nadin is one of the still living collaborators of Robert Rosen.

There are Reactive and Anticipatory systems. The last category has to predict the Future.

To predict the future a system has to Contain the Future. we call an Expectation .

A good example of an anticipatory system is a Pregnant woman.

Humans learn from an Expectation Failure sometimes called a Mistake.

Collaboration and learning are active in their own space (the physical context and the social context) and join in mental space (the internal state of a person).

6.2 Robert Rosen: Anticipatory Sytems

Robert Rosen made a distinction between Simple and Complex Systems.

In Complex Systems the Whole is more than the Sum of its Parts meaning that there are many Relations between the Parts that are not only Causal but have another Meaning.

In Simple systems the nodes are Nouns and the links are Verbs creating a Narrative that explains its Behavior.

Simple systems are Mechanical systems.

They can be explained by the Four Causes of Aristototle that always begin with asking How (Processs, Verb) an Agent (in general a Human Craftsman (Agent, I, he, ) has assembled Material parts (Noun) ending with the Why being its Purpose.

Living System have no purpose because they are not made by an mechanical process controlled by a craftsman although many people believe this happened when the autocatalytic proces took place.

Systems that can anticipate, con­tain internal predictive models of themselves and/or their environments,

They are part of the class of Third Order Cybernetic Models.

Robert Rosen, Natural systems are explained by mapping them on a Machine-Model mapping (Encoding, Decoding) them on a Formal Model where the Causal relation turns into an Inference.
Pattern formation in Nature.Te model contains two processes called aan Activator and and an Inhibitor.
Gierer-Meinhardt-model is one of the important types of pattern formation and morphogenesis observed in development.

A citation of chapter 1 shows that the same pattern of Acitivation and Inhibition is happening in the body : ” The peripheral reflexes can act independently of the brain, but they can also be coordinated and sequenced by it. The main point to be grasped here is that the brain’s primary purpose lies in the need for inhibiting peripheral reflexes, when particularly sophisticated sequences of muscular movements have to be executed“.

6.3 Dynamic Networks

Dynamic Modularity. ?How Life expanded from a Bacterium to a Human.
The Evolution of the Organism can be seen as an expanding structure in time that is controlled by a Linear Bodyplan that starts with the “Head” and ends with the “Tail” in which the genes act as Switches that start a deeper level of expansion.
Brain of a Fruitfly is divided in functional area’s.

6.4 The Human Brain is an Adaptive Dymamic Network:

More about the development of the brain push here.

6.5 Category Theory: Systems are Functions, Metaphors are Functions:

How to combine category theory and network theory.

6.6 Kent Palmer: Meta-Systems

About meta-systems

7. Meta-Heuristics

8. From BioPhysics to BioMathematics.

9. Rodney Cotteril : The Material World: