Stratified complexity and the origin of mental/social events
Dr. Paul S. Prueitt
CEO, OntologyStream Inc
Chantilly, Virginia
August 31, 2002
Also at: http://www.ontologystream.com/threeYearPlan/mentalEvent.htm
Section 1: On Mental and
Social Events and Reduction to Formal Logic
Section 2: A social system as
a living system
Section 3: Sensemaking with stratified formalism
Section 4: Formative Ontology software as ubiquitous Knowledge Management
tools
Section 5: EEG based Human
Information Interaction science
Introduction
The OSI/BCNGroup has identified a correspondence between basic research on mental events and formalism associated with
1)
Memory
acquisition and use
2)
Selective
attention
3)
Anticipatory
mechanisms
The formalism is developed from a general systems
theory that assumes that natural systems organize into stratified levels and
that knowledge technology should have computer processes designed to match this
formalism. This assumption is stated as the BCNGroup Conjecture about
Periodic Tables.
An innovative paradigm for mapping data into and
out-of an in-memory data structure, called Orbs, acts
as the primary interface between this formalism and a “synthetic”
perceptual/cognitive system.
Simplification of algorithmic techniques, expressed as
computer science, has allowed our group to design and build a knowledge technology with no third party
software and a minimal code base.
Having completed the basic software our primary goal
has become related to a proper study of social and cognitive science.
Our existing code base is served by a methodology
that the OntologyStream has developed from our work in knowledge management and
cognitive science. We anticipate making
In-memory Referential Information Bases (I-RIBs) algorithms available under
Open Source agreements so that other scientists can participate in this
study. A business model has
been developed to generate revenue for this community of scientists.
Section 1: On Mental and Social Events and Reduction to Formal Logic
In many cases, when one talks about something then
this something exists with a boundary condition, a location, and a
structure. But when one talks about a
“mental event” the boundary conditions, location and structure has not been
easy to define in a crisp fashion.
Precision is absent. We take
this absence of crispness to be an indication of stratification.
Talking about the nature of individual mental events
reminds us of the nature of a social system. Both types of phenomenon are part
of a class of systems that operate far from energy equilibrium. A glass of water at uniform room temperature
is an example of a system at equilibrium, roughly speaking. Living systems and
systems of living systems are more complex.
Knowledge representation must ultimately account for
both the natural properties of social systems and the natural properties of
mental events. This accounting requires
three things:
1)
A
more full appreciation of the nature of abstraction, and the differences between formalism
and natural systems.
2)
Stratification
theory as expressed in a tri-level { memory, attention, anticipation }
Knowledge Operating System. The tri-level
architecture produces stratification in our formalism.
3)
Openness
to new input from humans who are using human perception and cognition to make
sense of real time experience.
Abstractions, such as the counting numbers, do not
have boundary conditions or locations but using them one can have a high degree
of precision about some aspects of structure.
But, how abstractions come to exist and the nature of formal logics and
mathematics are two quite different problems.
Formal logic is developed from the process of mental abstraction.
Abstractions arise from the memory of experience, as
shaped by anticipations. Humanity has
struggled with a dream that all mental abstraction can be properly reduced to
formal logic. But this dream is a dream
shared non-uniformly within the world’s populations.
We have suggested that the existence of a desire to
have but one truth is due to the aggregation of (many) individual needs to find
an agreement with personal views within the larger social systems. Perhaps this is what each type of religious
and philosophical fundamentalism shares.
Formal systems are cultural artifacts similar to
natural language, except in a rather particular enforcement of notions of
rational consistency and deduction. The
relationship between natural language and formal logic has been hotly
contested. The BCNGroup scholars must
touch on this contest to explain why our technology breaks new ground.
We recognize that in a specific situation,
that formalism such as first-order predicate logic and stochastic forms
acquires a type of completeness and consistency. But moment-to-moment a knowledge technology must be able to find
this completeness and consistency from the ground up, that is to say the
computer processes must allow direct observation by living humans to shape the
axioms of the logic and the viewpoint expressed by these axioms.
Once the situated viewpoint is experienced (by the
human) then aspects of this viewpoint, the formative process that lead to the
viewpoint, the categorical invariance and chemistry related to the viewpoint and
various other aspects can be annotated and made part of a persistent framework based
knowledge system.
Section 2: A social system as a living system
The OSI/BCNGroup team is composed from a virtual,
distributed, community of scientists.
We are developing computer theory and applications based on a theory of
stratified complexity and category theory.
We are using the core notion of synthetic perception by a categorical abstraction
engine embedded as a Knowledge Operating System
(KOS). The inclusion of framework theory
in the design principles has recently (August 2002) allowed OntologyStream Inc
(OSI) to simplify the design of the knowledge base components of the KOS. The development kit for this system is
completed and can be demonstrated. We have argued in the previous section that
the major next step primarily involves social and cognitive science.
We argue that a social system is a living system
under the definition of living as provided by Maturana and Varela (Tree of
Knowledge, 1989). Consistent with
Maturana and Varela, we recognize that the "components" of a social
system are created by the social system and are not to be mistaken to be humans
who “occupy” the role within the social system in some way. The individual and the role have a “cross
scale” interaction that has been (up to now) very difficult to model in
non-stratified systems.
Figure 2: A generic model of
emergence as applied
to the understanding of emergent
thought
A new science of Human-centric Information Production
(HIP) demonstrates a theory with formal categorical notation supporting the
view that social roles are strongly independent of the individual. Seen in terms of
But by “strongly” independent we mean that the
dependency will vary, perhaps even appear and disappear, not that there is a
complete independence. The study of
this variation is a subset of the more general study of cross scale
interaction, where a great many unsolved problems remain.
By
using Maturana and Varela's notion of autopoiesis, one picks up a class of
specification that all autopoietic systems have. In particular one may address the concept that social systems
create their own components. The
component so created is marginally independent of the individuals who might
come to assume the role.
A
particular social system has a set of ecological affordances that are expressed
as an individual human assumes a specific role. This concept, of autopoiesis, is then used in developing a formal
separation of the notion of an ecological affordance for the social system and
the notion of a person who happens to be fulfilling a social role.
By
allowing the distinction that we have made (supported by Kugler's and Rosen's
distinctions on complexity) one can see that the human who is playing a role is
like the biological cell whose autopoiesis is separate from the autopoiesis of
the multi-cellular organization that happens to have the cell present, as
discussed in Maturana and Varela (page 75 – 80, “Tree of Knowledge”).
What
our discussion points out is that when complexity is treated in a
non-stratified fashion, there are paradoxes and ambiguity. "Being separate from" is a
stratified concept with two different systems sharing in physical phenomenon.
The two systems “exist” at time scales governing interactions within two quite
different ecosystems.
To treat the human who is playing a role in a
social system as a cause of the role is to completely miss an empirical
observation. The observation is that if
this person where not playing this role than some other person (different from
"this" person) would have been "brought into" the social
system in order to fulfill exactly this role. The role is little affected by
the specific personal qualities that might be seen as the differences between
people who might fulfill the exact same role.
The recent BCNGroup discussions have
pointed to the importance of emergence in developing stratified formalism. It seems clear to us that cross scale interactions
play various roles in ANY type of physical emergence. And emergence is then seen as critical to situated knowledge
representation.
But it is also clear that the emergence
of mental and social events has not yet been well enough understood to provide
definitive guidance on how our new stratified formalism should be refined. The voting procedure
has been shown (as early as 1997) to provide a decision support process that
takes into account inconsistent, incomplete and/or uncertain information. Like so many other computer aided decision
systems, it gives an answer. But the
measurement of the quality of the answer has not been done.
The Mill’s logic and QAT
is claimed by several groups of Russian logicians to have solved certain
problem related to the prediction of functional characteristics (of biological
agents) from the qualitative structural analysis of the periodic table of
atoms. But the work is not peer
reviewed in the West, and has been obscured in Russia by social/political
events.
Section 3: Sensemaking with
stratified formalism
To model “cross scale interactions” we
extend, and unify, the notions of linguistic functional load AND the Maturana
and Varela notion of structural coupling.
We note that the following have similar
function / substance considerations:
1) Interpreted
meaning of language.
2) The occurrence of
language in mental experience.
3) The meaning of
concepts that are formative to experience.
From the linguistic notions of functional
load, the linguistic scholars see a double articulation of two levels of
organization,
http://www.ontologystream.com/aSLIP/files/functionalLoad.htm
http://www.ontologystream.com/aSLIP/files/stratification.htm
In
particular if anyone has John Lyons "Introduction to Theoretical
Linguistics" Cambridge University
Press ..1968.. you might have read the following:
"Linguists sometimes talk of the
'double articulation' (or ‘double structure’) of language; and this phrase is
frequently understood mistakenly, to refer to the correlation of the two planes
of expression and context. What is
meant is that the units on the ‘lower' level of phonology (the sounds of a
language) have no function other than that of combining with one another to
form the ‘higher' units of grammar (words).
It is by virtue of the double structure of the expression-plane that
languages are able to represent economically many thousands of different
words. For each word may be represented
by a different combination of a relatively small set of sounds, just as each of
the infinitely large set of natural numbers is distinguished in the normal
decimal notation by a different combination of the ten basic digits." page 54.
Using
our new work on frameworks, one is
able to extract patterns of similarity, which can be refined by the user over
time, using OntologyStream software.
The frameworks seek to elicit knowledge from humans by asking questions
based on the pre-structured cross product of enumerated dimensions of discourse. Our cA/eC tools then seek to find patterns
of invariance in the conceptual structure of the answers. This conceptual structure can be trended
over time and further annotated. There
is very little technology that is required to support this trend analysis. It is just a question of focus. The patterns form cA/eC bring simplicity
from seeming chaos. This software provides
a powerful tool, a kind of brain-extending prosthesis that seems capable of
addressing the problem of articulating the patterns of emerging and morphing
social form. The software however is
built to reflect new formal theory and scientific work that is still
incomplete.
And
yes with the existing OSI browser software,
we now have this ability in simple cases.
We can demonstrate cA using this very simple and yet powerful software,
the new categorical formalism, and solid experimental science. We cannot under estimate the importance of
doing the cognitive and social science.
We
also have a formal framework for extending this existing software into the tri-level
machine intelligence architecture where we anticipate that a triple
articulation of various systems will build a knowledge base directed at
understanding events when viewing only the middle of the three levels. Modification to the sub-structure and the
ultrastructure will be by automated process or by a second order control system
managed by specialist. The user will
interact with the middle layer of the knowledge base.
As
can be seen in the current software (Figure 5), the tri-level substructure is a
stratum that that has a rich population of event atoms. These event atoms are categories where
variation within the category is or can be ignored when looking at
eventChemistry. Individual variation of
the event atoms, of course, occurs. But
the individual variation is lost during the assembly of middle layer events
within categories.
Complexity
theory is often, like quantum mechanics, very puzzling. For example, one can regard event atoms as
particles that are the components of social systems or mental events. This picture of the relationship between
atoms and various types of compounds is not entirely correct even in the case
of physical chemistry. But the use of
the OSI-KOS will not demand all users understand stratified formalism.
Figure 3: The senseMaking
architecture proposed as
a National Cyber Defense Knowledge Base
We
now recast the ideas from Maturana and Varela.
We posit that there is a substrate of human language use in a community,
and by that we mean utterances and written text. From this substrate, there is an emergent set of fluid, complex,
interlocking systems at the social level.
Human
information interaction events occur within a community of practice. The events can be modeled by various means
for producing machine ontology but any such ontology can be decomposed into a
set of ordered triples
{ < a, r, b >
}
where a and b
are locations in the event space and r is a relational variable. This decomposition then allows the Recomposition
of part of this ontology within a situated context. Within our approach we advocate using Zachman and/or Sowa-Ballard
framework to provide and reveal information regularity to the data structure
preserving the decomposed ontology (a sub-structure of cA atoms).
Formative ontology, from a tri-level
architecture, is a construct has not as yet found an economic means to be
expressed as information technology systems.
But a class of knowledge operating systems will soon exist in the
marketplace. Some of these will be
based on our software. OntologyStream’s
business model
assumes that economic models will shift from the economic control of
information tools to the value that is derived from having these tools freely
available for common tasks but we realize that is a disruptive technology.
Section 4: Formative Ontology software
as ubiquitous Knowledge Management tools
First order predicate logic has a role in
modeling what appears to be a “set” of emergent information interaction
rules.
We hold that reliance only on first order
predicate logic (such as with the Cyc ontology) is not correct and has a glass
ceiling. The glass ceiling cannot be found
as long as the Artificial Intelligence
mythology is maintained as a belief system.
Furthermore, assuming that human
information interactions are completely reducible to a first order predicate
logic leads to false sense making,
as pointed out in Karl Weick’s book, “Sensemaking in Organizations.” The glass-ceiling effect and the false-sensemaking
effect open up specific vulnerabilities in any intelligence vetting system that
depends solely on first order logic.
Visual rendering of formative and
situated ontology, derived from the annotation of structural invariance in the
framework data, allows humans to do what first order predicate logics cannot
do. The difference comes from the fact
that a human is living, and that the human cognitive processes are enabled with
memory, attention and anticipatory mechanism supported with the substance of
physical reality. The first order logic
is a simple formalism that serves to encode only part of the causes of events
into rules. These create an illusion
that the entire spectrum of cause has been accounted for. In practice, rules are encoded by pre-specifying
a viewpoint that may or may not be the instrumental viewpoint that a
knowledgeable human would make in a real time setting.
In the notation < a, r, b >
the nodes a and b are thought of as locations in an
event topology that depends on the relationship operator r and on
other metrics that might be brought into play.
But it should be clear from anyone’s private introspection that event
topology is formed uniquely moment-to-moment.
It is also clear that event topology regularly develops due to social
anticipation and other constraints on information sharing.
So,
the heart of our interest is in a stratified knowledge management formalism and
software implementation for rendering uttered and written language as
categoricalAbstraction (cA) and eventChemistry (eC). In these systems, it is felt that the glass ceiling imposed by
first order predicate logics is not present because of a dependency that is
placed on human-in-the-loop responsibility.
The
formalism we have developed has the ability to express a middle layer between
the substrate of invariance in uttered language and the seemingly intangible,
but apparently very powerful social structures that emerge from language and
mental behavior. The formalism maps to
the frameworks that we illustrate in our recent work on framework theory. The frameworks provide an action-perception
cycle that drives action by humans and results in learning (changes in the data
structures) by the stratified ontology.
Our first problem was with the terms like “occupy”
and “membership”. But using stratified theory and new category formalism, we
see objects of reference that are actually not a single object. The KOS objects of reference (those data
structures that are persistent) do not occupy a location or context. Membership has been weakened to the point
that set theory is replaced by category theory. We have separated substructure, built by the cA browser, and top
down constraints, built using the eC browser.
The human then is allowed to guide the formation of any one of several
views of the physical or informational world.
This is what we mean by being “situated”.
By using this approach, we are allowed to shift focus
from one level of organization to another level of organization. The OntologyStream formalism and ontology
specifically lacks crispness so that human agility is allowed to act in a
formative fashion. The human provides
interpretation to information artifacts.
This is very natural.
The tri-level formalism, expressed in the minimal voting procedure,
under-constrains the formative process so that the human cognitive agility can
freely act based on the human’s real time emersion in personal tacit knowledge
derived from direct experience and situatedness. If ontology is fully
constrained with artificial intelligence algorithms and is fully specified with
first order predicate logics, then the agility of human perception is
lost.
In the alternative, a closed form computer aided
decision support system is helpful in 85% of the cases. But one cannot tell (using the system) when
the computer system is being misleading.
Perhaps
the most central question involved in understanding the private experience of
knowledge has to do with state transition between events and not the content of
events. From the experimental
literature one can posit that the role of natural language during cognition is
expressed in the state transitions of mental experience. Language acquisition research, in particular
second language acquisition research gives us an experimental window into
mental states and the transitions between states. In the next section, we will look at the OntologyStream R&D
in this area.
Section 5: EEG based Human Information Interaction science
There are 4 OSI browsers, and the new Framework
browser (as of September 2002), and a number of tutorials on how to use these
tools. They are not integrated into
software to be marketed at a high price.
We are scientists and have not figured out how to make a marketplace
work for us. Our software, a complete
development environment, is free to a few and can be made available for
projects where our team supports the deployment of the techniques. We stress that the software is very mature,
but that the social science and the supporting study of deployment has not yet
occurred.
Over time we hope to develop an Open Source licensing
agreement so that the KOS can be developed in much the way that
Linux and Linux applications are developed.
The problem that we are facing and trying to solve is not how to make
money or develop software, the problem is on how to develop the social science
sufficient to explain how to use the tri-level architecture within a community
of practice.
Using the OSI KOS as a foundation technology, we have
proposed an EEG based technology to study cognitive state transition, in real
time in the context of learning.
Several research groups have produced EEG systems that detect certain
categories of cognitive state transitions.
Some of the BCNGroup scientists (Kropotov, Kugler, Pribram) have been
involved in developing this research.
Prueitt has long felt that game theory might be altered to reflect the
type of axiomatics openness that both the minimal voting procedure and Russian
QAT has. We would like to develop this
R&D.
In the context of open game theory, we posit that
there may be an alphabet of fundamental "content-type" states, so
that one could always, almost always, say that a mental state transition was
from content-type a to content-type b via a
transition-type q. The
object of the game theory is to predict the expression of social/mental
structural coupling by providing cognitive clues, anticipating cognitive state
transition of specific type, and modeling the future states given treats and
vulnerabilities. This is a big task,
but many different groups have completed much of the preliminary work. It is a question of bringing some of the
best minds together as advisors to a small core group of technology developers.
Figure 4: The Ontologically Relative Stratification having different locations
In theory, one can see the content states a
and b as aggregations from a small set of substructural elements
- derived from the invariance across a number of content occurrences. So that
a = f(A) and b =
f(B)
where A and B are subsets (bags) from this small set
of substructural "content" elements.
The discovery of this small set of event atoms (see Figure 5) allows
eventChemistry (eC) to be expressed with the aid of human cognitive acuity.
Figure 5: Some of the event atoms from
a study of computer port access
In reference to Figure 4 we might find that
Sa = A and Sb
= B.
However, outside of the pragmatics of a specific
situation, one cannot “know” that A and B are to be necessarily causally
related in a real situation that has not yet occurred. To show that Sa = A and Sb
= B are causally related in a single specific situation is not often a small
matter for many reasons. The only way
(again by hypothesis) to measure A or B is indirectly by observation of the
behavior of content states a and b. But most often, in stratified systems, the
measurement itself induces change in A and B and f due to the cross scale
aspect of the measurement. So it is
absolutely necessary that human cognitive acuity be allowed to make sense of
formative eC.
In formative ontology the bags from substructure become relative to location and are subject to “top down” constraint that has built up over time in the reaction (production) chains at the various levels of the stratification. So, over any period of time, we have a small finite state machine as a model. Using the SLIP Browsers we built and have access to a small finite state machine that is open to the occurrence of new states depending on human perception of the categorical invariance in the data. So we are already far into the challenge of demonstrating in practice what we feel we see from formalism and theory.
Once the mental events are modeled by cA/eC formalism
and detectable by EEG, we will be in a position to present states from the computer
tri-level architecture to the human and measure whether
1)
the
state is known,
2)
the
state is unknown
3)
the
state is not recognized as being known or unknown
and then further examine mental state transitions
using EEG if there are subtypes of
1)
recognition
2)
puzzlement
3)
planning
These subtypes of transition may be sufficient to
begin the scientific study of human interactions with complex knowledge
sharing.
A few questions should be answered.
- What are the different kinds of exploratory or avoidance behavior?
- Can these behaviors be represented as concatenations of verbs used
in the KOS command structure?
The tri-level induced formative ontology results from
the aggregation of "memory" invariance under the constraint of top
down expectancy due to previous event analysis. So the formation of emergent representation of an event might be
seen to correspond to a sudden insight, to an uncertain decision, or other
mental event transition. Human
information Interaction science must be based on the measurement of these kinds
of events.
The apparatus of a machine perceiving in the
synthetic sense and the human perceiving the formative ontology from the SLIP
browsers provides a means to study learning behavior in real time
investigations of events. Prueitt’s Generalized Framework Theory provides a
means to mark-up the human orienting behaviors.
We ask for your support. www.bcngroup.org