NEGOTIATING REALITY: USING LANGUAGE AND INFLUENCE DIAGRAMS
TO ARTICULATE KNOWLEDGE
Julia M. Di Stefano, Ph.D.
New Hampshire College
Manchester, NH 03104
ABSTRACT
This paper analyzes some of the recent literature on language and
information processing, focusing on graphic representations which model the
interactions between those transmitting and those receiving messages.
Having examined four models concerning interpersonal communication and
information processing, I conclude that today's most promising research on
dyadic communication is that based on the model of cybernetic control
systems. Most useful are the models which 1. recognize the need for the
speaker and listener to commit themselves to continue the dialog until they
arrive at consensus and 2. also recognize that the recursive interactions
between the two individuals are based on the principle of feedback, which,
in the words of Norbert Wiener, "is the property of being able to adjust
future conduct by past performance"(Wiener, 1954,33).
"A fool sees not the same tree a wise man sees." William Blake (1790)
For much more than 200 years poets and philosophers have been struggling
with the mysteries of the human mind, imagination, and perception.
Increasingly, researchers in artificial intelligence (Al), in their attempt
"to design computer tools suited to human use and human purposes" (Winograd
and Flores, 1986,8) are studying what happens when two people use language.
What happens when the fool tries to communicate to the wise man about the
tree the fool sees? And how can the wise man communicate about the tree he
sees? We now know that Korzybski was correct in recognizing that "The map
is not the territory," and that each individual carries in his/her own head
maps or mental models of reality formed by that individual's own life
experiences. In other words, the word or symbol is not the reality it
represents, and the words represent different interpretations of reality to
each individual. Small wonder then that our lives, professional and
personal, are fraught with miscommunication.
I. LAYERED MODEL OF INTERPERSONAL COMMUNICATION
Recent studies of human communication have challenged the Shannon-Weaver
sequential linear "transmission" model as inadequate for representing
interpersonal communication. The authors of one of the most recent and
insightful models, the Targowski-Bowman Communication Model: Layer Based
Links (1988) observe that the Shannon-Weaver model is seriously limited as
it addresses only the transmission of physical signals, and ignores
interpretation and meaning (Bowman and Targowski, 1987 ,24).
Copyright @ 1990 by Julia M. Di Stefano
293
294 System Dynamics '90
Targowski and Bowman do address the question of meaning, or more
specifically, what happens between the senders and receivers of "the
message" when people use language. They argue that:
"When a sender and receiver communicate, they obviously do so in a cultural
context that contains both shared and private meanings, and these meanings
influence the net result of the communication between them. Shared meanings
include both cultural meanings and those established by the immediate
environment in which the communication occurs. These meanings will differ
in saliency, with some being (or seeming) more important than others. This
results in a layering of meanings"(1988,9).
One purpose of the Targowski-Bowman model is to address not only syntax and
semantics, but the very important element of "the accumulation of
information"(1987,33) which includes the experience of the individual and
what semiiotician Umberto Eco calls "the encyclopedia"(1988,43) or the
cultural and historical experience of a society. The authors of the model
rightly observe that:
"Because the semantic reactions in the minds of sender and receiver depend
more on the information each brings to the communication process than on the
message communicated, even the best message properly delivered in a given
situation may be misinterpreted"(1987,33).
They argue, furthermore, that their "layer-based links" model shows how
"multiple levels of information are exchanged among sender, receiver, and
environment"(1988,10). See figure 1 below.
N Pee N ee Yyy Li)
S r . LEE 7 L
Retrieval
‘une
Won,
Figure 1
The Targowski/
Boman Communi-
cation Model
System Dynamics '90 295
The links start with level 1, the physical medium used for the exchange of
messages, such as paper or telephone or electronics, and rise up the chain
to level 10, the effect of the short term and long term memory and of the
communications skills -~ transmitting and receiving -- of the sender and the
receiver. The links also include the attitudes of the members of the dyad
to the physical medium of communication (level 2); the attitudes of the
communicators to themselves and to each other (level 3); the effects of the
time and place of the session (level 5); and the difference in status
between sender and receiver (level 6). It seems to me that as we progress up
these levels, we get further away from what can be objective described,
(i.e. paper or telephone, or the relationship between supervisor and
subordinate), and further into the black box of the mind engaged in the
communication process. For example, level 7 concerns the "degree to which
sender and receiver correctly interpret each other's verbal and non-verbal
symbol systems"; level 8 addresses "the expected and observed reactions to a
particular message"; and level 9 is concerned with the extent to which
sender and receiver share the same values. At level 10 we are definitely
into the black boxes, the minds, "the storage/retrieval links" of the sender
and receiver. At this highest level, the whole of the individual's long and
short term memory, including his/her mastery of communication skills,
influences the ways in which the messages are sent and received
(Targowski-Bowman, 1988, 11-19).
While the authors offer the disclaimer that this is "only the first
step"(24) of an in-depth analysis of dyadic communication, there are some
serious deficiencies of the layer-based link model. First, the model is
hierarchical ~- layered -- and implies a sequential order of events and
cognitive processes, while in fact the interpersonal communication process
is both recursive and simultaneous, with both speaker and listener using
simultaneous processing for the information flowing up and down the links
and along the paths mapped by Targowski and Bowman. Furthermore, 'links'
carries connotations of a chain, and with it associations of rigidity which
are inappropriate for the fluid and volatile process the model describes.
The authors admit that communication is dynamic and that a message is not
conveyed the way a "football is passed from a quarterback to a receiver."
Yet even though they appear to recognize the need for feedback loops in
communication models "so that senders can repackage and redeliver any
‘footballs' of information that go astray..."(1988,15) the diagram of their
layer-based link model appears static and rigid compared to the fluidity of
feedback loop influence diagrams as used by system dynamicists.
A second problem with the Targowski-Bowman model is its over-emphasis on the
cognitive process, as though we understand clearly exactly how the human
mind processes information. It may well be that non-cognitive elements,
those processes that operate beyond our levels of conscious awareness, are
more important than the modelers allow. Linguists, psychologists, and
semioticians are not entirely certain about how human memory works.
296 System Dynamics '90
In a witty fictitious dialog between a computer and a professor of
cognitive sciences, the computer says of its masters:
"They are very uncertain about what they have inside them....That is the
reason why they set me up. They know what I have inside me and when I speak
in a way that they understand they presume that they have the same software
inside them" (Eco,1988,59).
In other words, it is largely by the efforts of researchers in AI attempting
to make computers "think" that we are gaining insights into human
information processing. However, the layered model of Targowski-Bowman, and
the knowledge acquisition and memory access models described in the next
section rely rather too heavily on unproven assumptions about how human
beings really process knowledge. The most promising model, as I shall show
in the last section, shifts the focus from what goes on in the individual's
mind to the individual's attempt to negotiate reality with another
individual; or, in other words, to ascertain whether s/he sees the same
tree the other sees. But next we shall examine some models of intrapersonal
communication.
II. KNOWLEDGE ACQUISITION AND MEMORY ACCESS MODELS
In contrast to the theoretical approach of Targowski and Bowman, the
methodology of information processing in advertising, which builds on models
of knowledge: acquisition and memory access, brings together researchers
from various fields, including cognitive psychology and linguistics. These
two models, unlike the interpersonal model discussed earlier, are
intrapersonal; they speak in terms of "the cognitive processes that extract
the conceptual content for a text or a discourse" (Lehnert, 1988, 155). In
effect, then, this line of research is a kind of "applied epistemology"(157)
or, in other words an attempt to ascertain how we know what we know so that
we can design intelligent computers.
Lehnert's memory access model, building on the work of Tulving,
distinguishes between semantic vs. episodic memory. While semantic memory
deals with generic terms, like dog, episodic memory tells about a specific
dog, like Spud on television commercials. But it is not always easy to
distinguish where one kind of memory ends and another begins when we process
language. To illustrate the difficulty of deciding whether a proposition is
processed as a universal, generic or an episodic truth, Lehnert asks how we
would process the statement "Do penguins have skin?" and then diagrams a
hierarchical structure to manage the generalizations in the deductive
reasoning process. See figure 2 on the next page.
System Dynamics '90 297
Does a penguin have skin?
Figure 2
Semantic memory vs.
Episodic memory
According to Lehnert, figure 2 "shows a fanciful picture of semantic penguin
memory vs. episodic chicken memory to illustrate the difference more
concretely. The general point is simple: given any ‘semantic proposition,
it is possible to imagine an episodic structure that could encode the
information under consideration....The controversy in artificial
intelligence between semantic and episodic memory goes far beyond natural
language processing applications. Proponents of production-based expert
systems are banking on semantic knowledge, while people who argue for case-
based reasoning are committing themselves to episodic memory structures"
(Lehnert 170-171).
A similar model of information processing in advertising by Andrew Mitchell
is based on the premise that "semantic knowledge and episodic knowledge seem
to be linked" (Mitchell,1983,19). Mitchell views the knowledge acquisition
process as one in which information from the environment is interpreted and
then a representation of this knowledge is stored in long-term memory. See
figure 3 below.
“The Boy Carried the nN External Stimuli
le % Feature Analysis
Figure 3 { | ) sen
coat Letter Analysis Analysis
0
Lexical Anolysis
The Boy Carried the Cat Phrase Analysis
[@a\ _
Interpretive Analysis * Analysis
Text Analysis
298 System Dynamics '90
Mitchell's illustration of this process, shown in figure 3, represents
semantic knowledge as a "network of associations"(18). He emphasizes that
"the semantic interpretation of the phrase [the boy carried the cat] that is
stored in semantic memory is not the actual phrase." In other words, we
interpret or translate messages before storing them in knowledge structures.
Mitchell views knowledge acquisition as "a serial process with the
individual playing a passive role," although he does mention, but does not
elaborate upon other models that show simultaneous processing of information
at the sensory and semantic levels.
The focus of Mitchell's research is to assess information processing in
advertising. He posits that consumers "have packets of information about
different brands and about specific products...organized into a network of
associations" (Mitchell 20). See figure 4 on Ford Fiestas below.
Stiff
Suspension,
Few
Repairs
inexpensive
To Drive
Two
pie
Fig.44 A packet of information about Ford Fiestas.
According to Mitchell, this network of associations suggests that semantic
memory stores knowledge at different levels of generality; that there are
horizontal (Ford Fiesta-few repair) linkages; and there are vertical
linkages (Fun to Drive-Ford Fiesta) (20). Significantly, he fails to note
that key phrases in these links -- FORD FIESTA, FUN to drive, FEW repairs --
are alliterative. The copy writer for this campaign astutely observed that
alliteration, an ancient poetic technique, helps consumers to learn and to
link these positive associations with their product.
While the work of Lehnert and Mitchell and others sheds some light on human
information processing, many questions remain unanswered. There are,
moreover, some serious limitations to their models. For example, in her
quest for mathematical vigor, Lehnert focuses so closely on the trees--in
this case nominal compounds--that she may have lost sight of the forest,
i.e., the larger picture of how our mental models develop as a result of our
own individual and collective experience in a society. Mitchell's model
also is too limited in scope and in depth. He merely states what he
imagines is going on in the black box, conjecturing, not proving,
vertical and horizontal linkages.
System Dynamics '90 299
One point on which most researchers agree is that knowledge is always
interpreted, translated, or paraphrased before it is stored in semantic
memory. The “message-in-itself" does not go directly into storage because
in order for the message to be received it is filtered through the previous
experience of the listener/perceiver; this experience includes not only that
of the specific receiver of the message, but also, according to
semioticians, the whole "cultural repertoire of a given society, its
historical memory" (Santambrogio et al., 1988, 22).
Attempts to provide computers with modules of our "cultural repertoire" and
"historical memory" so they can process data intelligently are being
developed by Roger Schank and his team. They came up with Scripts, "a
sequence of conceptualizations with some variables in them," such as the
"Restaurant Script," which attempts to encode the sequence of events one
experiences when dining at a restaurant (Schank and Kass, 1988, 190). They
then developed a more refined system, called MOPs (Memory Organization
Packets), which allows scripts to be "modified dynamically" by using smaller
modular units of scripts, such as the waiting room scene, which might be
appropriate for several types of experience, such as a waiting room in a
doctor's office, a lawyer's office, or a personnel office (192).
Research using associative networks, such as Scripts and MOPs, is
interesting because it may provide some knowledge about how the human mind
processes information, but I believe the research models based on feedback
are the most promising for understanding how interpersonal communication
really works.
The most realistic models recognize language and cognition as social
phenomena. I agree with Terry Winograd that:
"Knowledge and understanding (in both the cognitive and linguistic senses)
do not result from formal operations on mental representations of an
objectively existing world. Rather, they arise from the individual's
committed participation in ... a socially shared background of concerns,
actions, and beliefs" (Winograd and Flores, 1986, 78).
III. CYBERNETIC MODELS BASED ON FEEDBACK
The most insightful research on cognition and human communication recognizes
the following:
@ knowledge is the result of interpretation
@ knowledge is neither subjective nor objective
@ language and cognition are fundamentally social
@ communication demands commitment
Most important, communication to establish consensus, to "negotiate
reality," is a recursive process based on feedback.
300 System Dynamics '90
Terry Winograd's diagram of a conversation for action, i.e., a conversation
which commits the speaker to a course of action, in contrast to the models
discussed above, does not focus on the receiver's or speaker's mental state,
but instead "shows the conversation as a dance" (64). Based on Austin's
and Searle's work on speech act theory, Winograd analyzes the structure of
utterances and “patterns of commitment entered into by speaker and hearer by
virtue of taking part in a conversation"(59).
Winograd emphasizes that:
1. he is concerned with the structure, not the content of the
conversation
2. the acts are linguistic, representing utterances and silences
between the requestor and hearer
3. conditions of satisfaction depend on the "
speaker and hearer"
See figure 5 below.
nterpretation of
they are not “objective realities"(66).
y J
A: Dec
AG tome are A: Oectare
8
'B: Renege
Lo)
A: Withdraw
8: Reject
As Withdraw
A: Witnaraw
A: Reject
8: Witnaraw
ro)
Figure 5. : The basic conversation for action
For example, Winograd correctly observes that "what is mot said is listed to
as much as what is said," and that some acts are taken for granted, if, for
example, there is no declaration to the contrary"(66).
Some limitations of Winograd's model are as follows:
1. It identifies only one very limited type of human conversation-
-a "commissive," i.e., one that commits the speaker to an
action
2. It does not take into account non-verbal elements, or such
element as the environment and status as identified in the
Targowski-Bowman model.
3. Most important, it does not acknowledge that language is
primarily somd of the voice, of the words, of the tone, that
creates and conveys meaning, that stimulates response.
System Dynamics '90 301
Despite these limitations, Winograd's model is superior to any other I know
of dyadic communication. The power of Winograd's model is its grounding in
three salient truths:
1. Interpersonal communication is recursive in nature and can be
analyzed in terms of cybernetics and the effect of feedback.
2. Human communication evolved from the need for survival, and the
study of information processing in other animals using
communication to survive can elucidate the process of
information processing in human beings.
3. Meaning derives not merely from "a systematic manipulations of
representations" (10), (as suggested by Mitchell and Lehnert,
for example), but from the experience of individuals living and
interacting with other individuals in a social environment, and
within a culture and a tradition.
His model is based on the view that an understanding of communication
and cognition requires solid understanding of living organisms as they
function in their natural environment. Winograd's model is built on the
solid foundation of a wholistic, systems approach to communication.
Specifically, Winograd builds his model on Humberto Maturana's rigorous
scientific studies of information processing in frogs. According to
Winograd, “Maturana provides two useful insights... : the role of the
observer in creating phenomenal domains; and the concept of structural
coupling, which is to understand behaviour that is mechanically generated
but not programmed" (10). Maturana's emphasis on survival (49) and on the
ways in which organisms use language, actions, and behavior in response to
their environment and in their interactions with other organisms from the
perspective of a neurophysiologist elucidates information processing in
human beings. His work "Neurophysiology of cognition"(1970) and "Biology of
language: The epistemology of reality" (1979) have enormously important
implications for the understanding of human communication.
The fact that Maturana illustrates his thesis with examples from mating
rituals, and that Winograd builds on Maturana's description of "the mating
dance as a pattern of mutual interactions" (49) serves to emphasize that
successful communication and information processing is basic to the survival
of all living species.
IV. CONCLUSION
Scholars and researchers from various disciplines have attempted to
understand and represent graphically the process of human information
processing and interpersonal communication. Models from business
communication, information process in advertising, and AI research on memory
access discussed in this paper contribute some insights for constructing a
model, although no adequate model presently exists. The most promising
model of those discussed here is the one by Winograd, who incorporates into
it knowledge from cybernetics, as well as from neurophysiology and from
semiotics.
302 System Dynamics '90
A more complete, more perfect model of interpersonal communication would
combine the best features of the Targowski-Bowman layered model (i.e., the
influence of the physical medium used for the exchange; the attitudes of the
speakers towards themselves and towards each other; and the effects of the
time and place of the session on the outcome of the dialog) with the
emphasis on commitment and on the structure of the interactions, as suggested
by Winograd's model. Rather than the rigid, hierarchical layered model or
the deliberately restricted domain of discourse described by Winograd, this
more perfect model of dyadic communication would include more elements of
observed behavior, represented graphically by feedback loops to suggest the
non-linearity and the complexity of the process. Such a model, with the
appropriate influence diagrams, is the subject of my next paper.
"A fundamental condition of successful communication," observes Winograd, is
"a standing commitment by both speaker and listener to enter into dialog in
the face of a breakdown"(63). It is only by a recursive process of
interactions between the sender and receiver that organisms succeed in
approaching each other's version of reality, for reality is neither entirely
subjective nor entirely objective. "Reality" is negotiated when two
individuals use feedback to "adjust their future conduct by past
performance." And when both are committed to continue the dialog until they
arrive at consensus. While they may not see the same tree, they have
constructed, by their engaged conversation, some semblance of shared
meaning. Perhaps, more importantly, they have affirmed or re-affirmed some
bond of trust, thereby strengthening the social relationships on which all
language and discourse are founded.
System Dynamics '90 303
WORKS CITED
Bowman, J. P. and A. S. Targowski. 1987. "Modeling the Communication
Process." The Journal of Business Communication. v. 24, no. 4, 21-34.
Eco, Umberto. 1988. "On Truth. A Fiction." In Meaning and Mental
Representations. ed. Umberto Eco et al. Bloomington, Ind. ‘Indiana
University Press.
Lehnert, Wendy G. "The Analysis of Nominal Compounds." 155-179. In Meaning
and Mental Representations. ed. Umberto Eco et al. Bloomington, Ind.:
Indiana University Press.
Mitchell, Andrew A. 1983. "Cognitive Processes Initiated by Exposure to
Advertising." In Information Processing Research in Advertising. ed. Umberto
Eco et al. Bloomington, Ind. :Indiana University Press.
Targowski, A. S. and J. P. Bowman. 1988. "The Layer-Based Pragmatic model of
the Communication Process." The Journal of Business Communication. v.25,
no.l, 5-24.
Wiener, Norbert. 1954. The Human Use of Human Beings: Cybernetics and
Society. Garden City, N.Y. Doubleday Anchor.
Winograd, Terry and Fernando Flores. 1986. Understanding Computers and
Cognition: A New Foundation for Design. Norwood, N.J. Ablex.
CC BY-NC-SA 4.0