Joel Muzard, Ph.D.
The rapid development of networking and the growing acceptance of internet communication as a mode of group collaboration has brought the communication problem into focus. We define sometimes communication as the conveyance of information from one person to another. Before information can be successfully transmitted it must comply with three constraints: Unambiguous content, adequate form and a common carrier. The information content to be transmitted exists as cognitive structures, (ideas, thoughts, a variety of conscious processes) in the mind of the transmitter.
Sharp definitions of these structures
permit translation into a form suitable for transmission on a
chosen carrier. The natural, and therefore usually, chosen form
of human communication is language: Spoken language, for the
most economical carrier, air, and written language, for the most
economical memory, paper. The computer screen, a hybrid between
the two, is by now a significant third alternative. The translation
in the given medium uses modulation according to some rules that
enables the receiver to decode the message and re-assemble the
original structure in his own mind. In spite of the natural endowment
for engaging in this process, the participants in information
exchange experience a fundamental handicap: the written or verbal
communication is serial. The transmitter has to disassemble his
cognitive structure into its constituents, arrange them in serial
order imposed by the code (grammar and syntax) and send the completed
sentence one component at a time.
The receiver then must reverse
the process to regain the original structure. Problems are caused
by the narrow bandwidth of the serial channel, accrued probabilities
of noise distortion over time and the nuisance factor of skill,
i.e. how gifted or talented the transmitter is to convey meaning
in spoken or written language. As a result reception is punctuated
with errors, misunderstandings, fuzziness and white spots, which
trigger the annoying interpellations "huh?", "what
do you mean?", "sorry, I didn't get that."
We propose a parallel form of communication, using computer assisted diagrams or schema. The cognitive structure, or idea, of the sender is translated into a pictorial format - a bubble - and conveyed in one piece using bitmapped graphics. This way, the message is instantly available to a parallel examination by the receiver. The message becomes a graphic object that can be manipulated in an external mode, it can be scanned repeatedly using the large bandwidth of the visual channel. Using this parallel communication (visual and spoken) , collective intelligence may emerge in a group.
Take for example the following
definition of knowledge pasted from an on-line dictionary server:
knowl.edge \'na:l-ij\ n ME knowledge,
fr. knowlechen to acknowledge,
irreg. fr. knowenX obs 1: COGNIZANCE
2a1: the fact or condition of knowing something with familiarity
gained thr ough experience or association 2a2: acquaintance with
or understanding of a science, art, or technique 2b1: the fact
or condition of being aware of something 2b2: the range of one's
information or understanding 2c: the fact or condition of apprehending
truth or fact : COGNITION 2d: the fact or condition of having
information or of being learned archaic 3: SEXUAL INTERCOURSE
4a: the sum of what is known : the body of truth, information,
and principles acquired by mankind archaic 4b: a branch of learning
Think for a moment. From this
text, what is your understanding of 'knowledge'? Compare it now
with the following cognitive graph built with the K-language
implemented in Ideaprocessor:
You "see" the difference
between the understanding derived from the text-language serial
mode and the graphic-language parallel mode. Assuming that the
graph corresponds tightly to the internal structure of the multi-faceted
concept of 'knowledge' it is possible, with the graphic language,
to explore the meaning of knowledge in a "comprehensive"
form. The cognitive structure has become an object that can be
transposed directly from the original representation of the sender
to an identical representation by the receiver. And both know
they are sharing the same graphical object, building together
a common representation.
The graphic language:
The graphic language is composed of a simple ontology of knowledge: A classification of things and links. Each concept has a type. The Class of things include for example: Concepts or class, objects, properties, values, agents, task-actions, states and location.
Each link has also a type. Classes of links include: associative, distinctive, temporal, spatial, logical, and data-flow relationships. There is an iconic code for each type of element using contour, colour and alphabetic symbols. Thus things and links are colour coded with the name and type displayed for identification. With this grammar the sender can express his cognitive structure in a graphical form, allowing (even forcing) the explicit expression of his idea.
We can distinguish three steps
in the construction of a graph. First comes the collection of
the basic elements that come to mind in raw form, a loose assembly
of abstractions such as "things" connected with "case-of"
links. In a second step, we look for a more formal organization.
Now we are grouping things using associative links to build a
model of the domain, or a flow of processes with temporal links,
or hierarchies of tasks based on logical dependencies and so
on. We can merge the domain, the dynamics and the tasks in an
inference model or a state model. This formalized structure can
now progress to an operational structure that will take into
account the real constraints and the real values of its properties.
The whole process is guided by the given question or some other
communication goal from the very beginning.
The construction of a graph,
the instantiation of an internal representation, is a modeling
process. A representation can have five roles (Davis, Shrobe,
& Szolovits, 1993): it is a surrogate, it is a set of ontological
commitments, it is a fragmentary theory of intelligent reasoning,
it is a medium for efficient computation and it is a medium of
human expression. We can ask: how well does this representation
function as a medium of expression? How general is it? How precise?
How well does it function as a medium of communication? "A
representation is the language in which we communicate",
and "the goal of knowledge representation is to describe
and understand the richness of the world" (Davis, et al.,
We have placed emphasis on the
communication aspect of this graphical language. But it is also
related to the other roles of the representation function. The
analytical effort required in graph construction leads to a crystallization
of the internal idea (surrogate role) as well as to explicative
schemes why a specific representation was chosen (theory of reasoning).
External feedback from communication partners assures that the
graph serves efficient computation and, naturally, unambiguous
representation makes a commitment to believes and values.
The classes of
things and links
Where do the classes of things
and the classes of links come from? Research on brain damage
victims (Bishop, 1993) tells us that a person may have semantic
difficulties with a verb, but when the same word is a noun there
is no problem. This finding suggests that the brain sorts words
according to some grammatical categories. Starting from this
idea, we have selected basic elements that are abstractions of
The process of understanding
entails linking elements together in a fashion congruent to perceived
reality. In this view understanding is a meta-process relative
to the internal representation of the world, it is dependent
on knowledge, derived from information and therefore subject
to distortion and error as all inductive processes are. The discipline
of graphic explicity should lead towards consistency, completeness
Several researches have worked
on the categories of links (Brachman, 1979; Brachman & Schmolze,
1985; Cox, 1993; Dieng & Trousse, 1988; Gaines, 1991; Knuth,
Hannáck, & Radó, 1988; Wielinga, Schreiber,
& Breuker, 1991). The work of Otman (Otman, 1990; Otman,
1992) is particularly relevant. We have extended his classification
of links in order to cover distinctions among specific cognitive
Assistance to the
The user is helped in his efforts
to build a diagram by the simplicity of the classes and the proposal
of links between two types of things that make sense. A knowledge
base has been constructed to support this feature.
Further assistance is provided
by the graphical interface that allows continued visual inspection
while the graph is under construction. The graphical interface
also offers iconic tools that facilitate the cognitive analysis
by displaying types of things and types of links.
The facilitation of communication with a graphical language, the K-Language, implemented in a computer assisted idea processor has been described.
The K-language allows the creation of webs of meaning, semantic webs, a more powerful way of communication. It facilitates systemic work. And it allows knowledge work, interpretation and the social construction of meaning. A powerful way of making sense, of collective intelligence.
See IWorkshop , webIDEApro and IdeaProcessor for
the implementation of the K-language
The Knowledge-Cafe is a way of experience the K-Lnguage.
This is a work in progress from our research-lab at Applied-intelligence-Atelier. Contact us for more info.
Bishop, J. E. (1993). Word Processing:
Stroke Patients Yield Clues to Brain's Ability to Create Language.
The Wall Street Journal, CCXXII(72), A1-A14.
Brachman, J. R. (1979). On the
Epistemological Status of Semantic Networks. In V. N. Findler
(Eds.), Associative Networks, Representation and Use of Knowledge
by Computers (pp. 3-50). New York, USA: Academic Press inc.
Brachman, J. R., & Schmolze,
J. G. (1985, An Overview of the KL-ONE Knowledge Representation
System. Cognitive Science, p. 171-216.
Cox, L. A. J. (1993). Knowledge
Acquisition for Model Building. International Journal of Intelligent
Systems, 8, 91-103.
Davis, R., Shrobe, H., &
Szolovits, P. (1993). What is a Knowledge Representation? AI
Magazine, 14(1), 17-33.
Dieng, R., & Trousse, B.
(1988). 3DKAT, a dependency-driven dynamic-knowledge acquisition
tool. In 3 International Symposium on Knowledge Engineering,
(pp. 85-93). Madrid
Gaines, B. R. (1991). An Interactive
Visual Language for Term Subsumption Languages. In IJCAI-91 12th
International Joint Conference on Artificial Intelligence, 2
(pp. 817-823). Sydney, Australia: Morgan Kaufmann Publishers
Knuth, E., Hannáck, L.,
& Radó, P. (1988). A taxonomy of conceptual foundations.
Data and Knowledge (DS-2), 205-219.
Otman, G. (1990). Les réseaux
sémantiques en terminologie: l'exemple de l'IA. In Second
International Congress on Terminology and Knowledge Engineering,
. Trêves (RFA)
Otman, G. (1992). Cogniticiens,
ne négligez pas la terminologie. Collecte et formalisation
des connaissances expertes par le cogniticien. In DEXA 92, .
Wielinga, B. J., Schreiber, J.
A., & Breuker, J. A. (1991). KADS: A Modelling Approach to
Knowledge Engineering No. KADS-II/T1.1/PP/UvA/008/2.0). University
page design by: Roberto Muzard
iWorkshop, webIDEApro , IdeaProcessor, Statex, and Applied intelligence Atelier are trademarks
Copyright © Applied
intelligence Atelier inc, 1996 -2009