Preface
Revised May 24, 2006
Foundations 2007
“Current computer science operates with models of
information networking, and databasing that were conceived in the mainframe era
and cannot serve the needs of a truly connected world.
Designing the Future of
Information, Harbor Research Inc
Section 2: Second School Principles
Section
3: The role of pragmatism
Section 1: Preliminary notes
In
this book, a specific information architecture is outlined from a mixture of
complex systems theory and knowledge representation theory. A strong attempt is made to make a principled
grounding in the natural sciences.
It
is with respect and appreciation that we identify an alternative to the
semantic web and artificial intelligence disciplines, as classically envisioned
by scholars like Jim Hendler [1]. We call the alternative the second school of
semantic science and contrast this with the first school of semantic
science. The first school does not make
itself available to the concerns of leading edge neuroscience, biology or the social
sciences, staying firmly anchored in scientific reductionism. This is a criticism, but we feel that the
criticism is valid, objective and has supporting evidence. However, we are more interested in revealing
the second school viewpoint.
The
second school uncovers an information paradigm that is human centric. It preserves many of the engineered aspects
of current information technology. The
second school recognizes that knowledge of how and why things occur is critical
to many modern activities. It
recognizes that the current intellectual activity does not focus often on the
how and why.
Statistical
knowledge helps to define some envelopes within which things normally
occur. The questions to who, what and
where are given answers. However a
discrete pathway, containing states and transition between states, is more
descriptive of phenomenon such as gene expression or cell signaling [2].
[3] Gene expression research is one area where
ontological mediated science is beginning to make transitions from artificial intelligence
to something unexpected [4]
[5]. The limitations of statistical approaches
and the limitations to artificial intelligence are different, but in the
viewpoint that I am describing they both take paths away from understanding the
how and the why.
The
second school suggests that social expression is as complex as gene and cell
expression. All of these expressions of
nature are part of the study within the principles of the second school. Properly understanding the second school’s
viewpoint on complexity requires recognition about the limitations to the first
school. Once this recognition is fully
in place, then the second school builds on the existing technology through
simplification and through the introduction of a specific set of principles.
The
second school principles model natural process is being “sometimes”
under-determined, with respect to deterministic causes. The natural processes such as decisions made
by living systems seem to not only follow Newtonian laws but some other set of
laws. The model requires a theory of
stratification and emergence since we feel that the beginning of emergence is
where a high degree of non-deterministic reality can manifest. This issue of emergence is common in the
chapters to follow.
A
fundamental principle of the second school has to do with conjectured
non-algorithm aspects to these decision events. Sir Roger Penrose is not the only scholar who talks about
non-algorithmic reality. For some
reason, it is as if one has to start out showing that the assertion of
“everything is algorithmic” is a false assertion. Francis Crick certainly makes this assertion in “The Amazing
Hypothesis”. Accepting the second
school principle is consistent with the assertion that artificial intelligence
is a mythology and that this heavily funded academic discipline has produced a
lot of poor scholarship. This poor
scholarship has been seen to feed upon itself, producing a perceived
illegitimacy to many aspects of “science”, particularly information
science. In technology based on the
second school viewpoint, we place an emphasis on having a human in the
loop.
What
does it mean to have information science be centered on the human being and not
on the technology?
In
the second school viewpoint, the human is (to be) supported by process and
structural standards in developing information and encoding information into an
ontological model. [6] The ontological model is something that will
be a common focus to second school discussions. We feel that Hilbert mathematics, the advanced mathematics in
physics, has a limitation that is only fully appreciated when the second school
viewpoint is examined. In particular we
suggest that introspection about internal feeling of self, that has so far been
separated from mainstream science, can guide the development of new category
theory, and from this category theory some new formalism based on category
theory and not based on number. This
will not be easy, for many reasons.
Our
goal is to move the activity of science so that taxonomical, ontological and
mathematical models are more easily interfaced, and mathematics is not over
used [7]. The issues are not beyond easy
comprehension, and require a specific background in literatures. There are several ways to approach this
literature.
A
discussion of the history of bio-mathematics could be developed at this point,
but we defer this discussion. The key
point that would be extended is that classical mathematics does not have the
same level of success with biological functions as it does with engineered
structures. Acknowledging this key
point will be critical if science, and society, is to move information science
in a new direction. In the text to
follow, the reader will see that the new direction merges mysticism with
science by making corrections to both disciplines.
For
many scientists, the non-formalizability of common everyday activity is a
given. Certainly one cannot point to
formal systems with which everyday activity can be modeled. Yes, in specific cases, and within specific
boundaries one can model, with formal methods, everyday activity. But these models are not flexible, and they
certainly do not exhibit the types of emergent behavior, like the behavior of
thinking, that we experience with great familiarity. The spiritual expression seems remote to the current
materialism.
It
is clear and simple to state that formal models using first order logic do not
provide an adequate replacement for human awareness and reasoning. Web ontology language, for example, provides
something that works, kind of, in some situations; but these formal methods
have a number of specific problems. I
mention web ontology language [8]
even though many people have not heard of this. I will state here for clarity that my criticism of the web ontology
language work product is grounded in a detailed history. Again; however, this criticism is not the
focus I wish to make now.
We
can state the obvious. In everyday
activities, human being and communities localize (real) ontology
situationally. This localization occurs
through real physical processes that are supporting human awareness and
cognition. The second school brings a
simple representational technology to mirror these everyday activities. We do not call this “ontology” but rather we
call it ontological modeling. The
second school has revealed a technology standard [9]
that is based on the notion that symbols reference concepts and that our human
concepts about ontology might be referenced using symbols that emerge as part
of human use patterns. I will try to be
clearer about this as we move further into second school thought. An ontological model is a representation of
reality via a set of concepts.
In
theory ontological models [10]
may be used by members of a community to encode new pieces of structured
information into bit patterns that are compact, easily manipulated and can be
visualized on a computer screen. Like
words into grammar, the elements of an ontology use symbols to annotate concept
indicators. The representation can be
in the form of the W3C standard, a standard that uses a triple having the form
< subject, verb, predicate>,
but
the W3C standard is only one form of ontological modeling [11].
The
other well-known standard is the topic map standard. [12] I will not discuss topic maps here, simply
because the history to far to complicated and the basis for not allowing topic
maps to be the leading standard are too difficult to examine. The position I take is that the markets
choose incorrectly. If we were looking
for the best tools for modeling complex expression such as gene, cell or social
expression we need topic maps and the “n”-ary representation presented in the
notational paper, Prueitt 2002. [13]
The
specific problems with the W3C standard can be listed, starting with the standard’s
assertion that class structure be defined precisely. This assertion forces symbol systems to fix formal semantics and
in this way to create formal ontology that is not the best model of natural
ontology. Categories of processes
cannot be modeled with this assertion in play.
The list would also include the assertion that the resource description
framework, on which the W3C standard depends, is sufficient to represent any
type of human knowledge. The list would
include the W3C’s dedicated professional support of a community of “knowledge
engineers” who feel that they can encode everyone else’s knowledge but do not
understand anything about biology or psychology and often nothing about the
foundations of mathematics. Finally,
the list would include the assertion that web ontology can infer new
information in a way that is general and similar to the cognitive awareness of
humans. This last assertion is the
assertion made by the artificial intelligence discipline.
In
the chapters that follow we make the case that there will not ever be machines
made of abstraction that are self-aware.
We make this case by looking at the nature of abstraction, and the
biological processes that are necessary for thinking.
In
work being developed from second school design elements, the manipulation of
ontological models will be done via visual elements that correspond to the
invariant aspects of experience as realized by humans. How one judges knowledge about “invariant
aspects of experience” is by following accepted scientific peer review,
modified by new methodology being synthesized from spiritual practice.
We
hold that those who have intimate knowledge of a specific phenomenon will best
develop ontological models about that specific phenomenon. For example biological scientists should not
have to learn the rules of computer coding to develop and use information
observed in biological laboratory experiments.
The second school technology allows direct development of symbol systems
by simplifying both the underlying data structure and the operating
environments. [14] The optimal
underlying data structure is an “n”-ary,
< r, a(1), a(2), . . . a(n) >
encoded
using hash tables. In our proposed
“.vir” subnet of the his underlying structure is managed without any software
dependencies using the very basic computer processes. The operating environment uses a type of category management
based on stratification (to be discussed later in the book).
These
models can be explicit, and how these explicit models are represented is also
addressed a bit later in this chapter.
Semantic extraction technology will be discussed first. We have to first clear the air from the
misuse of words by the knowledge engineering and artificial intelligence
communities. The issue of misleading
concepts has to be examined. I have
used the term “polemic” to mean a mythology that is specifically designed to
not encourage further examination or analysis.
Semantic extraction is misnamed.
The second school claims that this misnaming creates a polemical
structure that improperly elevates the notions of artificial intelligence while
disallowing a critical examination of what the full nature of “meaning”
is. What these well-known techniques
and algorithms do is to discover structural patterns, mostly based on
co-occurrence of words or phrases. The
“meaning” is then imposed using some type of taxonomy or web-ontology. But the “extraction” of meaning using these
techniques is incomplete and sometimes incorrect. “Semantic extraction” would be better called “structure
extraction”. The reason why the phase
“semantic extraction” is used is that the hyped up buzz phrase “semantic
extraction” has been rewarded with many grants and contracts. It is that simple.
In
fact, an entire class of technologies measures the presence of concepts using
subject matter indicators, such as the presence of word, stem or word phrase
co-occurrence patterns. Some leading
researchers regard semantic extraction as a generalized n-gram analysis [15]
and we agree that most of the best semantic extraction software is some type of
n-gram measurement with a set of heuristics defined over co-occurrence within
the windows of an n-gram [16]. N-grams can be generalized to most of the
methodologies developed in the extensive image understanding literatures. These literatures are well reviewed by Tapas
Kanungo [17].
Some
obvious technical comments are needed.
The n-gram does not have to be a linear contiguous measurement of
co-occurrence. As one of many
generalizations of n-grams, some structure/function relationship might be used
in the process of indicating information that may become knowledge if perceived
by humans. The fulfillment of a
function desired, or anticipated, by the environment is the same as the pure
concept of “semantics”. Meaning and
function are tied together by artificially imposing one community’s sense of
meaning. Thus the values of the
software vendors, which are very narrowly focused on their making money, over
ride the natural functions that communities in crisis need to access. Transactions are placed under the control of
that one community in a way that is not, up to now, transparent.
Service
Oriented Architecture [18]
has had the promise that lines of business will be facilities by a network of
routers that control to flow of information generated by requests for
services. This beautiful concept was
usurped by the US Federal CIO Council’s work in 1993 – the present on allowing
special interests to hard wire all procurement of services based on the nature
of the software programs. Similar fears
regarding the electrical process are facing us as we go into the 2006
Congressional elections. The solution
to this structural problems between the citizen and the government is
transparency and the type of personal knowledge operating systems that my group
has designed, and partially implemented.
The
independent observation of social discourse [19]
opens up the possibility of a neutral measurement of functions asked for and
received. For example, the grammar in
natural language might be used in the computing of knowledge
representations. The notion of “a
passage” might be extended to include theories of discourse where the
boundaries of passages are irregular [20]
[21]
[22]. Passages are then seen to be the expression
of the various elements of human expression.
More is to be said on a full generalization of n-gram measurement
later.
The
point is that, for many of the semantic extraction technologies, n-gram
analysis occurs after there is a measurement process. Consistent with second school viewpoint, the measurement process
should have human involvement as part of the real time situated activity
leading to situational models of complex phenomenon. The measurement of individual expression involves a full spectrum
of emotional, intellectual, cultural and spiritual realities. This is not what is occurring in artificial
intelligence or knowledge engineering disciplines. The second school viewpoint holds that without a non-controlled
real time involvement in measurement the phenomenon involved in normal response
behavior will not be captured. So with
academic and government supported research and development being somehow
pre-occupied; what might be done?
This
issue of what is to be done occupied me for two decades. The experiences I was fortunate to be
involved in gave me the background to understand a specific set of issues, but
the pre-occupation of the institutions providing academic jobs and
capitalization seemed to be overwhelming.
I personally and professional failed over and over. Over the years the broad outlines of a
solution emerged. At first the outline
of this solution was limited to what might be called the intellectual dimension
to human expression. Beginning in 2005
I began to see that human expression remains poorly modeled if the emotional,
economic, cultural and spiritual dimensions are not considered. This new understanding has been very
difficult to integrate with the hard realism of science for the sake of
commercialism and war fighting.
A
very simple to use “knowledge operating system” is needed. This system must be processor independent
and occupying less than 100k of code.
The system has to have an internal interface to humans, and an external
interface into the Internet. It must reside
on any device that has a computer chip inside.
The
first designs were focused on the use of sensors to bring information to the
human, provide a repository of subject indicators, and provide structural
information about common response patterns.
These designs were developed for application to fighting wars,
protecting commodity transport and creating service architectures for business
transactions. Clearly these
applications were skewing the underlying technology.
These
comments point out that the technology that has been developed by government,
business and the academy has not taken into account systems theory. What is
known about design and functionality of natural systems? Lines of business defined as XML based web
services between the federal government and commercial business establish
empires. These programs have created
advantages for a narrow type of economic transaction. No transparency on this service provision process is allowed
simply due to the practice being so widespread as to be “business as
usual”. What is dis-advantaged are
human values related to family, the individual owning of homes, and the
economic prosperity of individuals.
The
concept of a utility function may help my inadequate description of the nature
of the problem space we current face.
This concept is applied to models of processes that are largely
reducible to computer simulation. The
discipline of genetic algorithms is an academic study of the evolution of
computer simulations. [23]One
sees the utility function as an essential part of these simulations. But more broadly we see the utility function
as being the non-Newtonian guide to the evolution of natural systems. The point to my discussion about technology
being shaped by commercialism and scientific reductionism is that this
technology itself has become a utility function over our cultural and
individual expressions.
Let
me give a specific example. Dr Richard
Ballard has for several decades worked on a non-serializable knowledge coding
system. Several of his systems have
been developed and deployed under government contract. These systems are designed to help manage
government contracts in the area of national defense. The work solves certain types of problems but at core asserts a
set of social and cultural values that are skewed towards economic transactions
supporting war efforts. Ballard’s
contributions are only very partially published since his work has been part of
a proprietary process, to which he has shown dedication. Personal communications between he and
several members of our community have helped in the development of views about
how “semantic web” knowledge systems will work in the future. As my group and other groups work to achieve
success in our professional efforts, the utility function created by war
efforts has been shaping our work product.
Ballard’s
work is not the only contribution that is beyond n-gram technologies. Another member of the second school is Tom
Adi [24]. As does Ballard, Adi works with the notion
that sets of semantic primes exists and are composed into subject matter
indicators (during the process of generating language in normal everyday speech
and writing). The work of both Ballard
and Adi pointed me more clearly to the notion of generative data encoding. There is a generative progress involved in
creating systems that “do something” in the real world. The question had become, “what have our
technology systems been designed to do?”
Tom
Adi’s work was always focused on understanding the cultural and personal
aspects of human expression and thus has deep spiritual roots. Richard Ballard’s work has focused on
providing ownership over the intellectual property produces by a narrow range
of social expression. This work leaves
out the spiritual values and replaces these with a competitive reality where
ownership is essential.
A
range of consequences develops due to the patterns of economic
reinforcement. The drive to assist the
individual and communities move toward sustainable and resilient
ecosystems. The drive to achieve
spiritual qualities has to compete with a fierce measurement of utility driven
by commercialism. These facts are the
facts of our lives in 2007. These facts
are everywhere evident. Which drives
will win out and become the dominate social reality of the twenty first
century?
By
examining my own work during the period 1991 – 2004, one can see how effective
the utility functions have been.
Starting in 2001, differential ontology [25]
was directed at the production of knowledge representation from semantic
extraction techniques. The extraction
techniques develop co-occurrence patterns and these patterns are (in my
architecture) presented to the individual for manipulation. A support structure for development of
ontological models focused on the nature of individual human memory and
anticipation. But the underlying
motivation for the systems I was designing was to control commodity transport
worldwide. [26]
In
theory the focus of ontological model development becomes individually
centric. The individual is supported in
acquiring information that is relevant to real time situations. The general principles were that individuals
should be empowered to take control of personally defined information spaces. There were issues of course. The nature of human expression is shaped by
the means with which the expressions occur.
The powerful consumerism shaped by the American success is reinforced as
these models are used to control commodity transport. The utility function distorted my work product and limited what I
was able to achieve. Was my experience
unique to me? I do not think so.
Many
people see that consumerism is wasteful and has created specific classes of
imbalances. The principles that
motivated the development of my model for US Custom’s control over worldwide [27]
commodity transactions could be used to reduce waste in these transactions and
to decrease the imbalances to or cultural, economic and environmental
systems. But few in government were
interested.
Over
the past year, 2006, I have been more deeply involved in meditative practices
in several retreat resorts in Taos and Santa Fe, New Mexico. The full integration of the human soul has
been the subject of meditation practices for centuries. In retrospective, the approach I had been
taken regarding the use of categorization seemed incomplete. Being human centric should mean more than an
increased ability of the individual to be a successful consumer and war
fighter.
The
core issue seems to be how balances can be made between the various energy
centers of the human being. The
integrated self, made whole by meditative contemplation seems possible only
when the more aggressive drives subside.
On the other hand, the paradox remains as to what to do when there is
such an imbalance between those who have economic power and those who do
not. We will return to these issues
often in the “Foundations 2007” [28].
Semiotics
is a discipline that refers to a system of signs, such as natural
languages. In differential ontology,
signs may develop from an interaction between socially oriented knowledge
management and normal everyday activities.
This type of interaction depends on the development of instruments of
communication like normal language, but in addition to the normal mechanisms of
natural language there is a computational form. This form becomes part of a
symbol system reified by use as a communications tool involving more than one
person. The symbol system is managed
through the human interaction. The
individual symbols are each formal topics in the topic map.
In
chapter one we will talk more about the cognitive neuroscience that inspired
the full concept of differential ontology, and the “stratified architecture”
that makes it work. At present I can only suggest that in the near future,
humans will develop sign systems based on ontological models in a way similar
to how geometry and arithmetic was developed historically.
We
have come a long ways, in spite of obstacles. The first school is sidetracked
by the assumption that classical logic is sufficient in computing the
consequences of knowledge experiences in real situations. In part, they (and “we”) were lead into this
mistake by the success that Hilbert mathematics has had in physics and
astronomy.
The
primary difference between ontological models and mathematics is that with
ontological models the abstraction is more situationally focused. We may remember that physics and astronomy
deals strongly with “universals”.
However, living systems seem to have aspects that are not accounted for
by the universals of physics and astronomy.
This is interesting, and if correct this perception tells us about
science and also about life itself!
Imagine that!
Because
incompleteness, inconsistency and uncertain information are often important
properties of real biological processes, the focus on phenomenon is
complex. For example, decisions are
often made with uncertainty or informational incompleteness. In Chapter one we will look closer at the
notion of “coherence” and see why this misuse was almost unchallenged.
We
understand now where the open questions are located. The second school refers to most current “semantic extraction” as
a syntagmatic extraction, since a structural pattern is “all” that is found. The certainty of formalism like geometry and
arithmetic is found in the measurement of these structural patterns. There is no ambiguity in the results of this
measurement. But the measurement process itself can be flawed. The measurement
of structure, such as the concentrations of protein expressions at a specific
time by a specific biological system, can be incomplete or mismeasured.
Beyond
measurement issues are issues related to emergence of function. The measurement of the function of structure
is where uncertainty, not related to incomplete knowledge, occurs. With differential ontology the measurement
presents to human visual inspection those ontology signs and symbols that evoke
meaningful mental experiences. The
human is then allows, in real time, to make the categorical abstractions based
on human intuitions and cognitive abilities.
In
this way, the natural process of human induction is applied directly to the
output of computational processes mediated by structured information, taxonomy
and formal ontology. By leaving out the
formal inferencing, we present for human inspection sign systems that are
measuring precise patterns. The origin
of knowledge is then within the control of the human who uses the system. [29]
A
human experiences meaning. The
computational “meaning” extraction processes is not “from a human” but rather
from the organization of words in text.
We measure a transaction space that is not directly the space where
human thought is being developed and shared.
We hold that when we misuse language, use the word “semantic” when we
really should use the term “structure”, then we diminish the quality of the
technology that is produced. We make
mistakes because the misuse of language directs us into these mistakes.
Section 2: Second School Principles
The
second school finds bypasses to specific hard problems that have blocked
success in transitioning information technology to individual control. These bypasses produce a knowledge operating
system useful to individuals or communities.
We have been designing our knowledge operating system since 1998. The original design of KOS was made partially
public. [30] In making these public disclosures we were
acting in a way that is consistent with core principles, in particular the core
principle that all software IP will eventually be set aside. This
principle is based on a set of “optimality proofs” [31]
suggesting that computational technology must eventually be optimal and not
patentable. The current, 2006, details
of the technology designs are layered with a social philosophy being on top. [32] At the bottom is a process methodology for
starting with something fully specified.
That technology infrastructure is then expected to evolve in a natural
way. The knowledge sharing foundation concept was
first developed (2003) as a suggestion supporting the US intelligence agency
needs to develop information about event structure. [33]
Previous to this, a small group of scientists had talked, since 1991, about the
need for a curriculum, K – 12 and college, to support an advancement of
cultural understanding of the complexity of natural science. [34]
So one of the layers has an educational
grounding.
The
knowledge operating system can be as small as 17K and be independent of the usual
software platforms. The KOS develops a
referential base having a specific well know and standard format (essentially
topic maps) and this referential base can be large. [35]
When
a particular knowledge operating system has been provided structural
information, a measurement process and the visual interface to a human
perceptional system, then we create the basis for knowledge-based anticipation
of “what is next”. At this point six
classical Greek interrogatives; who what, where, when, how and why, are used to
produces both new human knowledge and to allow a greater sharing of this
knowledge. The technical means to
achieve this interrogative based information definition uses what I began in
2003 to call “general framework theory”.
[36]
We
believe that the nature of the structural information and the nature of the
transaction space for encapsulated digital objects will be accepted by the many
virtual communities such as Second Life [37]
. Structural information encoded as
generative digital objects provides control mechanisms for the individual.
The
key element of this potential acceptance is the second school principle that
separates push and pull advertising.
But
not all of the residents of Second Life are happy about the commercial gloss
that is starting to spread throughout the world. They argue that the appeal of these fantasy realms was that they
offer an escape from the uniformity of a globalized society. [38]
The
separation of push and pull information has the potential to transform many
kinds of social activity, from on line shopping to group collaboration within
virtual communities.
The
separation is made possible using several of the second school principles. In order to talk about these principles, we
need to introduce the technology movement called “SOA”. Service oriented architecture (SOA) began to
be a buzz phase in 2003 and 2004. Since
that time great effort has been to develop standards that reinforce the existing
centralized control over service definition.
Individual business powers saw that a repository of service access
points would change the status quo if services were merely competitive based on
consumer oriented outcome measures. We
observed that advertising is what keeps firmly rooted service providers in
market positions. We also observed that
the industry powers have a collaborative synergy that protects the concept of
centralized authority.
I
was living in Northern Virginia and attending government-sponsored meeting as
the interest in SOA began to ignite. It
was clear to me and to others that big business was working hard to not lose
the existing strong monopoly on government funding for services. However, SOA is a concept whose time had
come. I studied the emerging standards
and followed the discussions.
An
example of the business orientation of SOA is expressed in a recent advertising
for a web seminar on SOA
The
goal of a service-oriented architecture (SOA) is increased IT adaptability,
reduced cost of application development and maintenance, and better aligned IT
professionals and business users. But the ultimate benefit of an SOA is better
information. And better information benefits business users. Done right, SOA can help business users shift their focus
from merely running the business to maximizing the performance of their
business. Not only will they be able to lower infrastructure costs, they can
optimize their organization’s key assets – information, customers, and brand.
Literally
hundreds of IT companies are in competition to “deliver” SOA
transformations. However, it is
conjectured, that there are real deficits in how the concept of a service is
realized in any of the current leading SOA vendors.
The
“.vir” standards start out with the general systems theory notion of a
transaction space. The standard
recognizes that the transaction space has a manifestation in natural reality
and that any model of the transactions spaces is limited not only in the
initial form but as a consequence of the normal evolution of natural
systems. I am regarding the term “transaction”
to be appropriate in describing the exchanges between living systems in complex
ecosystems. [39]
The
strategy we have developed is to lay down a best of kind MUD (multiple user
domain) based on a long study of the 1988 DARPA MUD engine as realized in the
Palace virtual community system and the Manor virtual community system, as well
as the new system designs we see in Second Life. This work has been studied by several members of my close
associates, in particular Nathan Einwechter in Canada and Amnon Meyers in California. As of November 2006 we are awaiting
capitalization of OntologyStream Inc. [40]
VirtualTaos
has also been designed to use a specific selection of the SOA standards,
starting with the SOA Reference Model [41] A number of models describe how repositories
of information are to be created and how programs or individuals should best
interact with these repositories. Perhaps
the reader would reflect on what he or she observes about the competitive
marketplace of today. De-centralization
of service selection would change the utility function, creating service
information that was rated on objective community, ie consumer, measurement. Advertising per se would be reduced in
importance and would change to allow a measure of the truthfulness of the information. Oh my!
The
evolution of the concept of service-oriented architecture initially was focused
on bringing order to a de-centralized flow of goods and services. The notion was, and still is, that the many representations
of service potential should compete on a just in time delivery of these
representation in response to service request.
The service request drives the entire system, even if not
decentralized. The orchestration of a
selection is, in theory, totally managed automatically with no human
interaction in the real time.
As
the SOA architectures have become functional we see that the effort to hard
wire competitive advantage extended beyond the IT vendor to selected
collaborations with non IT service providers, further extending control over
services to the IT vendors. Nowhere is
this more obvious that in the education sector. Rather than get off into this discussion here, we simply give the
URL where the most advanced deployments of service oriented architecture into education
might be judged to be occurring. [42]
The
anticipatory architecture, suggested in grant proposals by myself to US
intelligence agencies in 2001 – 2004, uses syntagmatic expansion and contraction
mechanisms to tease out the response patterns that are anticipatory of
behavioral state transitions made by individual, economic and social
systems. The notion of anticipation has
not been developed in the SOA architectures, in spite of a number of standards
describing such things as “service discovery” and “service fulfillment orchestration”.
The
anticipatory technology’s principle technical innovation has to do with a
separation of the substance of information into two types of ingredients. One
type of informational substance is related to phenomena observed to be involved
in human memory [43]. The other
informational substance are those mechanisms, largely functional between the
frontal lobes and the limbic system [44],
involved in what Karl Pribram called “executive” control over mental event
formation [45].
The
second school’s formalization of this separation is inspired by a study of
cognitive neuroscience and related scientific domains like immunological
theory. The study of cognitive
neuroscience and immunological theory was part of my “early career” (1985 –
1995). Knowledge Foundations [46]
provides a look at this background. In Foundations, we address the technology
and architecture inspired by this early work.
The technology and architecture is informed by applied semiotics as
practiced in a particular Soviet era school of cybernetics [47]. In this “applied semiotics” school,
information primitives were derived from a set of invariances measured across
multiple instances. A system of symbols
arises that serve as control elements to computer assisted decision support
systems. This scientific work was
largely disrupted due to the collapse of the Soviet political and economic
systems in the late 1980s.
Many
core concepts are re-emerging as Soviet era science settles in other parts of
the world and finds lines of research, native to those other communities, which
are consistent with these core concepts.
If one looks carefully it is possible to see that in all parts of the
world general systems theory type work has been published for many decades.
One
could study human memory research to fully appreciate the theory underlying our
use of informational primitives and top down templates. However, one might also appeal to private
everyday experience. In common
experience we see invariants in our concepts of texture and color or emotional
responses. These invariants are part of
the experience we have as living human being.
The notion of an invariant across many instances seems to have always
been in my mind, but certainly my exposure starting in 1994 to the work of the
Soviet era cyberneticians [48]
intensified my focus on invariance as a means to identify “semantic
primitives”.
Cognitive
science proposes that the invariance that is experienced by a human mind is
aggregated into a physical memory store.
Direct experience subsets part, but not all, of these invariants in
response to perception of specific things.
In the biological processes supporting human awareness, the subsetting
process is constrained by anticipatory responses and images of achievement and
action [49]
.
How
does anticipatory technology compare with what we humans are familiar with in
everyday life?
Everyone
has direct experience with awareness, and experiences that come from the many
acts of communication with other humans.
In natural language, as in all other mental phenomenon that the human
mind shapes, abstractions are used to refer to the invariance found in real
experience. Examples of a system of
abstractions are the concept of a counting number and the concepts about
color. The concept of the counting
numbers is a particularity nice example of an abstract upper ontology. The upper abstract concepts should have the
property that they are unchanged by use.
This independence from situational use is one property of arithmetic
that gives us commerce and the engineering sciences.
Given
a specific biological entity, a set of top down anticipatory templates will
have been developed over a period of time.
These templates model how events are connected together. Part of the mechanisms will be genetic and
part will be specific to experience.
How these mechanisms work is highly complex and is not yet fully
understood as a matter of objective science.
The
first ontological descriptions of behavioral patterns will be
simplifications. The distance that has separated
IT from the core behavioral sciences is a profound problem. We are not given funding support to bridge
this gap, for complex reasons. It is
simply beyond my personal ability to support this assertion in a scholarly
fashion, and I apologize for this.
We
could look at biologically motivated models of neural function. The ontological template is certainly
reminiscent of the top down templates that Stephen Grossberg developed,
starting in the late 1960s, as part of his mathematical theory of human
perception [50]. The top down template in Grossberg’s
architecture has been used in many types of applications and has been subject
of thousands of research articles. The
basic concept is that a basin of attraction of a continuum dynamical system
develops through iterative adjustments to a mathematical model, in Hilbert
space mathematics; resulting in what is called “reinforcement learning”. Parallels to reinforcement learning as
studied by cognitive neuroscience have been reviewed by hundreds of major
research articles. This research
literature is vast, and we will allow the reader to look into this or not,
depending on the reader’s level of technical background in Hilbert type
mathematics and in the relevant sciences.
Ontology
templates have a different nature. They
are the explicit side of a dualism between explicit and implicit
representations of subject indicators.
Figure 1: Fundamental diagram for
differential ontologystream
By
“implicit ontology” we mean an attractor neural network system of one to the
variations of latent semantic indexing or rule based semantic extraction
systems. By “explicit ontology” we mean
a bag of ordered triples
{ < a , r, b > }
where
a and b are locations and r is a relational type, organized into a graph
structure and perhaps accompanied by logic.
Our
templates have explicit forms and these forms can be part of additional
technology such as colored Petri nets [51]. In my PhD thesis (1988), I developed several
chapters on what I called homology between discrete dynamical systems and
neural network models of the type that Grossberg invented. Later, in 2000, I made the obvious
connection between discrete formalism state transitions and the state
transitions seen in neural network models.
This lead to the concept that one might be able to have both a discrete
formalism, like a Petri net or like an ontology defined set of symbols with
state transitions defined explicitly, and a continuum formalism like a neural
network. Several technical means were
developed to explore how to extract the discrete formalism involved in subject
matter indication using the SAIC owned patented Latent Semantic Index
technology [52].
My
work in 2005 has focused on using human reification of the results of several
commercial semantic extraction systems, including those sold by Convera,
AeroText, MITi, and Applied Technical Systems.
This work is discussed in the context of a Global Information Framework
based on what I called the differential ontology framework (DOF) [53].
Ontological
models extend the modeling function of Hilbert mathematics to formalism that
serves to unify social construction about objects of inquiry into an objective
set of concept representations. But,
unlike Hilbert mathematics, there is no demand that universal truth is established
through the formalism itself.
Situational truth is established only via the usefulness that these
constructions find as human communities use ontology to serve the community or
individual purposes. Truth that extends
beyond single instances is available within the context of reified standards,
but ontological modeling should always acknowledge the possibility that new
categories will arise.
There
is a separate issue related to the nature of logical coherence. Logical coherence has always been loosely
defined based on a sense of sufficiency in reasoning, i.e., what is considered
rational. Rationality, in spite of its
value, breaks down at a number of places.
First is the problem that other minds do not always agree. This problem can sometimes lead to extreme
problems. The fact is; however, that
the underlying physics of coherence is what gives the sense of rationality a
firm basis. The brain requires phase
coherence in the electromagnetic spectrum in order to support perception and
cognition. [54] The fact is also that in physical coherence
we have phenomenon such as C-tuning forks and D-tuning forks. A D-tuning fork does not make a C tuning
fork ring.
The
core difference between the first school and the second school is that the
first school assumes that Hilbert mathematics is the ultimate representation of
truth, and that the imposition of first order predicate logic brings the
ultimate truth of Hilbert mathematical reality to the ontology of social
processes. At core this is the
“ontological commitment” made by Cyc Corporation [55],
and the history of this corporation is one key to understanding the limitations
of the Semantic Web concept produced by the W3C standards body. The second school suggests that natural
science shows this first school’s ontological commitment to be a profound
error. The second school also suggests
that when one gives up this error, one is able to define new technology that is
both human centric and easy to use.
Differential
ontology framework does in fact set up an anticipatory technology.
In
our architecture, an encoding of templates and representation of invariance is
achieved via a type of category theory and set theoretical operations called
convolutions. Our formative category
theory is used in the formation of behavioral atoms and periodic tables related
to specific objects of investigation.
The elementary operations that we defined using set theory is performed
on a hash table like, key-less hash table, ontology persistence construction
called an Orb (Ontology referential base).
Referential ontology is regarded as a set of concepts, and can be
translated into standard RDF (Resource Descriptive Framework) [56].
We
conjecture that in the near future, ontological model based templates will be
designed, by scientists or domain experts, to model behavior expressed in the “.vir”
subnet of the Internet. This is fully
consistent with a new OASIS standard specification called Business Centric
Methodology. [57] In fact,
the BCM standard is more general that what is needed to provide services to
business processes. A derivative
standard is being outlined with the name “Process Centric Service Methodology”. The two key innovations in the BCM standard
is the notion of service blueprints, and provide a formative framework and
human choice points that allow humans to choose between blueprints. This new standard and related standards is
discussed more fully in chapter one.
In
any natural situation, the function of emerging composites is formative within
a specific template or between templates or categories of templates [58]. Opening up the interpretation of structure
creates something for the human to do and the human will do this very well. How well future ontological models are
defined depends on human individuals knowing that reality has situational
features. Having a handle on these
situational features is vital if one is to understand as much as one can about
something like the intentionality of living systems [59].
The
fact that categories drift and new categories emerge, unexpectedly, is the key
to future ontology specifications.
Section 3: The role of
pragmatism
Perhaps
nowhere is the difference between the first and second schools more apparent
than on the issue of pragmatism. The
second school holds that meaning has a pragmatic axis. This axis exists only in
a real situation.
Oddly
enough, human awareness also exists only in the present moment. Natural situations occur only in real
time. Abstraction is separated from
this pragmatic axis and sits in a time independent fashion. Meaning is often indicated by an abstraction
but it takes a perceptual experience to realize the full meaning. The abstraction in an ontological model,
like text on a page in a book, evokes mental experience. In normal social discourse, this evocation
occurs only in a non-abstract situation; i.e., in a pragmatic axis to reality
(in the moment). The first school has
been busy developing a world where the ontological specification occurs by
computer science professionals. Up to a point this work is important and
useful, but the true limitations to standardizing our representation of natural
processes must be acknowledged.
The
distinction between what is experienced and what is abstracted into natural
language or formal constructions is the key to a Human-centric Information
Production (HIP) paradigm. HIP is not
free. The use of categorical abstractions
and anticipatory templates by users requires an educational background that is
different from that which most college graduates receive. We have to then take up the question of our
educational system. We have asked
questions about what should be in a K-12 curriculum designed to make knowledge
operating systems as familiar as natural language is today.
The
second school follows the path of scientific realism. While we acknowledge the utility of scientific reductionism, we
also acknowledge the limitations that are created when reductionism is
practiced as if a religion. We suggest
that there are good reasons why science and mathematics is ignored and rejected
by the majority of students. Science
and mathematics have become a pathway to a narrow profession, not necessarily
to an increase in personal knowledge.
While it is true that the science and mathematics professions product
economic value within our social system, it is also true that the economic engines
that are being created are damaging our environment and perhaps the social
structure. It need not be this
way. We must look at nature and see it
for what it is and not require that it fit within our expectations. The second school is a bridge to the
knowledge science. We predict that the
science of knowledge systems is to be built based on scientific realism.
We
envision that the use of anticipatory technology will result in the formation
of a world wide anticipatory web of activity.
Different from the mainstream notions of a “semantic web”, the
anticipatory web will arise from the activity of many individual humans using
computers to provide structural information about aspects of real time,
experienced, reality through a measurement of relevant data. The core issue is the separation of a
reasoning component from a visualization and computational component.
Tim
Berners Lee properly addresses this core issue in a 1998 paper [60]
on “The Semantic Web as a language of logic.
A
knowledge representation system must have the following properties:
1. It must have a reasonably compact syntax.
2. It must have a well defined semantics so that one can say
precisely what is being represented.
3. It must have sufficient expressive power to represent
human knowledge.
4. It must have an efficient, powerful, and understandable
reasoning mechanism
5. It must be usable to build large knowledge bases.
It has
proved difficult, however, to achieve the third and fourth properties
simultaneously.
The semantic web goal is to be a unifying system which will (like the web
for human communication) be as un-restraining as possible so that the
complexity of reality can be described. Therefore item 3 becomes essential.
This can be achieved by dropping 4 - or the parts of item 4 which conflict with
3, notably a single, efficient reasoning system. “
Tim Berners Lee’s remarks are consistent with a
school of thought that points out the limitations that have been seen in formal
systems. But his insight in to the limitations
of formal systems is not complete. He
does not question whether it is possible, under any circumstance, to have sufficient expressive power to represent human
knowledge. An extensive scholarly
literature exists about these limitations (see Chapter 2, Foundations).
Beginning in 2005, the BCNGroup (Behavioral
Computational Neuroscience Group) declared that Semantic Web technology falls
into two major schools of thought:
a)
The First School of Semantic Science stipulates that ontology supports common
sense reasoning with the imposition of constraint logics like OIL
(Ontology Inference Layer for RDF – Resource Description Framework).
b)
The Second School of Semantic Science stipulates that ontology enables
knowledge sharing, which can best occur with minimal dependency on
constraint logics and inferences based completely on algorithms.
One school holds onto the polemics of
artificial intelligence by acting as if computer inference is more desirable
than machine to human interfaces. This
school brings us software systems, like Protégé [61]
and Jena [62]. Very few professional computer scientists
can get these software systems to work, and the possibility that average
individuals will agree with the assertions of Protégé and Jena are zero. It is quite easy to point out how ridicules
these assertions sometimes are. As a
general rule confirmed by community experience, one gives up one’s human
centric design principles in order to develop a working software system. The first school also participates in an all
or none polemic where one is expected to completely accept the standards-based
assumptions. The first school advocates
tend to ignore those who do not agree.
Proponents of the first school often demonize those who will not buy
into the standard. This behavior is
reinforced by the control over the funding mechanisms that the first school has
enjoyed. The reader is reminded that
the author understands that his criticisms of the system have been hard to
justify. The situation is both complex
and filled with difficulties having cultural roots.
We
can make this very simple. The second
school rejects the notion, in principle, that one can substitute computational
inference for human reasoning and the human experience of meaning.
We
can be polite to the old ladies. This
rejection does not give up advanced computational algorithms such as those
found in what is currently called “semantic extraction”. Algorithms such as neural architecture
inspired pattern extraction and categorization, latent semantic indexing and
probabilistic latent semantic analysis, are extremely useful in developing
computational instrumentation that measures invariance and patterns of
invariance. The algorithms produce
subject matter indicators and these indicators are used in a methodology
developed within the second school. So
we build on what is correct about what we now have.
In
the second school paradigm, the individual will produce information in a
fashion that allows both a pre-structuring of response and communication channels
and the creative input in real time.
This human centric approach sets aside the unreasonable desire for an
efficient, powerful, and understandable reasoning mechanism. In making this choice we move away from an
often very difficult technical discussion about logic constraint language
standards and consistency checking. We
move toward an analysis about knowledge sharing that is often not at all about
the limitations of computer science and formal constructions. We are able to speak in a plain language
about the social and individual experiential aspects to knowledge sharing
within communities of practice. The
discussion shifts from dysfunctional IT to social theory, human factors and
knowledge management issues.
No
one can tell what social transformation will occur co-incident with the
appearance of low cost and easy to use to use anticipatory technology. Certainly the advent of the second science
of semantic science will transform the criterion that has controlled federal
and private funding of semantic web activity.
This transformation will mark the end of a period of time in which
science has been held back by religious type beliefs. Oddly, the religion of reductionism is what separates theories of
intelligent design from mainstream funded science. By respecting, but putting aside, the polemics that come from
this separation, we are allowed to regard human introspection as a proper
subject of scientific investigation.
We restate that the role of the individual is critical in the formation of constructions relied on by anticipatory web mechanisms. These constructions do share properties related to the origin of natural language and properties from the induction of logico-mathematical formalism. These properties suggest a way to overcome the specific types of limitations that computer science inherits from the foundations of mathematics. So one does not avoid the need to understand the history of logic, and the natural science related to mental events. In fact the sec