Content uploaded by Winston R. Sieck
Author content
All content in this area was uploaded by Winston R. Sieck on Sep 30, 2024
Content may be subject to copyright.
Proceedings of Phoenix Challenge 2008 1
Cultural Network Analysis: Characterizing Target Audience Cognition
Winston R. Sieck, PhD Louise J. Rasmussen, PhD
Applied Research Associates
1750 Commerce Center Blvd, Fairborn, OH 45324
Phone: 937.873.8166, Fax: 937.873.8258, wsieck@ara.com
INTRODUCTION
As described by Major General Scales
(Marine Corp., Ret.), the U.S. military’s Achilles
heel is knowledge of the enemy, and technology is
no replacement for understanding the enemy’s
mind. This extends to understanding civilians, as
well, since winning their “hearts and minds” or
influencing their attitudes and psychological states
will play an important role in the wars we can
expect over the coming decades. Influence is a key
aspect of information operations (IO), where the
goal is to affect enemy and other decision makers to
achieve specific objectives. Also, the purpose of
PSYOP as a component of IO is to influence the
attitudes and behavior of foreign governments,
organizations, groups, and individuals.
Furthermore, past research has identified the
characterization of target audiences as one of the
most critical and challenging aspects of IO (Sieck,
Stevens, & Shafer, 2004).
An inherent challenge in understanding
foreign target audiences rests in gathering,
analyzing, and representing the relevant cultural
concepts, beliefs, and values that drive decisions in
those populations. In this paper, we present
Cultural Network Analysis (CNA) as a broad
approach that aids in providing the most relevant
cognitive aspects of cultural groups for decision
influence. CNA comprises a collection of
methodologies for eliciting, analyzing, and
representing the beliefs, values, and cognitive
concepts that are shared by members of cultural
groups (Sieck & Rasmussen, 2007). This paper will
provide a detailed description of CNA as well as a
discussion of how CNA can be applied to support
challenges in characterizing target audiences.
Culture as Shared Knowledge
Within cognitive anthropology, culture is
typically defined as involving shared knowledge
(D’Andrade, 1995). One specific theoretical
approach to culture that characterizes culture in
terms of knowledge is the epidemiological view.
Here, “Epidemiology” is used in the general sense
of describing and explaining the distributions of any
property within a population. Cultural
epidemiology regards culture in terms of the ideas
that are widely distributed throughout a population
(Sperber, 1996).
The emphasis on “ideas” or content knowledge
is consistent with work in cognitive field research
and naturalistic decision making that has
consistently found experiences and mental models
to have a primary influence on real-world decision
making. The research from this community clearly
identifies the contents of cognition, as opposed to
microlevel cognitive processes often studied in
laboratory experiments (such as working memory),
as the major driving force of decisions.
CNA leverages what has been learned about
mental models from cognitive field research in
order to more concretely define the kinds of ideas
and their interrelationships that matter most in
human decision processes. Mental models are
experience-based, causal explanations of how
things work that guide a person’s assessments,
judgments, and their decision-making. Mental
models depend heavily on culture.
To take a concrete example, consider
intermarriage between U. S. military personnel and
Iraqis. Intermarriage has been taken as one visible
indicator of the extent to which we are winning the
hearts and minds campaign. Several reasons have
been cited for why so few U. S. personnel have
taken home Iraqi spouses, as compared with wars
past. One potential reason has to do with the
cultural differences in how Americans and Iraqis
think about romantic relationships. We can
appreciate the differences by developing an explicit
model of Iraqi romantic relationships.
A mental model of romantic relationships
contains a person’s concepts as well as their
understanding of the causal relationships between
concepts, i.e. the antecedents and consequences of
romantic activities and their outcomes. This mental
model influences the individual’s expectations for
how romantic relationships should unfold and
provides a framework for selecting behaviors and
goals within romantic situations. For example,
individuals may hold the idea that a date is a social
engagement to go out alone with another person,
usually with romantic intentions. Their minds may
also be inhabited by the idea that dates should be
avoided at any cost. As an example, consider
Figure 1, a pictorial representation that might
describe an Iraqi’s mental model of romantic
relationship pathways. The set of ideas represented
in Figure 1 were extracted from a single newspaper
article on Arab-Americans, and so it should be
treated as largely notional for illustrative purposes
(MacFarquhar, 2006). Figure 1 depicts a number of
ideas using circles, lines, and color. These ideas
include simple concepts such as dating and
marriage, represented as circles. It also includes
causal ideas such as that dating decreases ones
chances of marriage, and of staying on an Islamic
Path. These are represented as lines in the figure,
with +/- indicating the direction of the causal
relation. Finally, Figure 1 portrays ideas of desired
states or value using color, as well as a logical flow
across desired states. Staying on the Islamic path is
a good thing, something one should do. Finding a
marital partner is likewise valued.
Email
Online
Chat Dating
Pre-
marital
sex
Talking
Morally
Marked
Marriage
Islamic
Path
Parent
Influence Scripted
Meeting
−
+
+
+
+
+
+
+
+
−
Email
Online
Chat Dating
Pre-
marital
sex
Talking
Morally
Marked
Marriage
Islamic
Path
Parent
Influence Scripted
Meeting
−
+
+
+
+
+
+
+
+
−
Figure 1. Notional representation of a culturally-shared mental model.
Since dating increases the risk that one will be
toppled off of the Islamic path, as well as
hampering ones chances of getting married, it
should be avoided. Hence, holding this mental
model is likely to have fairly strong consequences
for how a person will decide and act.
As implied by the name, mental models reside
inside the heads of individuals. However, when
people engage in activities such as speaking,
writing, drawing, and modifying their environment
in any way, their mental models leave observable
traces in the form of physical artifacts and
representations. Figure 1 is one example of an
external trace of a mental model, just as was the
article from which it was derived. When
externalized traces are encountered by another
person, that person’s mind produces thoughts that
are similar to the originating mental models, at least
Proceedings of Phoenix Challenge 2008 2
Proceedings of Phoenix Challenge 2008 3
to some extent. While looking over Figure 1, you
probably entertained some thoughts that were
somewhat like those of the people quoted in the
article. On a broader scale, people who come into
contact with similar traces can thus develop mental
models that resemble one another. Mental models
can spread widely throughout a population,
becoming “cultural” in the sense of being shared by
many of its members. They can also endure within
the population for very long periods of time.
At this point, it is useful to summarize and
define a few related terms. First, the term culture
refers to mental models, and other contents of the
mind, that are shared by members of a population
over a period of time. It also includes the resulting
behaviors and other traces that foster prolonged
survival of the shared ideas by providing “habitats”
for them.
Cultural group refers to a self-identified group
of people that constitute the population of interest.
Traditionally, members of cultural groups were
connected in many different spheres, including
being neighbors, engaging in the same work, and
participating in the same social and religious
activities. High overlap in experiences like those,
clearly leads to shared ideas within a large number
of domains. More and more, people often identify
with an increasingly wide assortment of groups that
vary considerably in aspects such as purpose, size,
and cohesion. Modern cultural groups may be best
defined and described using tools such as social
network analysis.
The relevant cultural group for a study will
depend on the cultural domain, that is, the kind and
topic of knowledge of interest. Further, despite the
redundancy, we sometimes use cultural knowledge
in place of culture to refer to the networks of mental
content for which there is some level of
concordance among members in the cultural group.
Finally, cultural model refers to an external
representation of a culture that is constructed by a
researcher. A cultural model represents a consensus
of the mental models for a particular cultural group
and domain. Hence, to the extent that its elements
are shared among Iraqis, Figure 1 serves as an Iraqi
cultural model in the domain of romantic
relationships.
Considering Figure 1 as the cultural model for
some target audience within Iraq gives us a precise
way of identifying cognitive vulnerabilities to
influence cultural change. For example, suppose
marriage is the most tangible perceived outcome
that is negatively influenced by dating. We can
then affect a change in the cultural model by
targeting the specific causal chain of beliefs that
dating will decrease the chances of becoming
married. This could be done by developing
messages, including concrete images, that show
Iraqi daughters dating and then getting married.
This example also highlights the interrelation
between causal beliefs and values. That is,
changing the causal belief chain so that dating is
seen as increasing the chances of marriage can also
affect the relevant value (or attitude) towards
dating.
Why Cultural Models?
Cultural models are formal descriptions of the
knowledge possessed by members of particular
groups. Cultural models describe and represent
how the world is understood by the members of
these cultural groups. A key premise is that cultural
knowledge comprises many networks of causally-
interconnected ideas. These mental models become
activated within particular situations to drive
thinking and decision making, and can change
under suitable conditions. Cultural models also
seek to account for intracultural, as well as
intercultural variation in cultural knowledge,
relationships between cultural knowledge and social
networks, and cultural change. Cultural dynamics
across social networks is especially useful for
planning IO, and anticipating influence effects.
Figure 2: Cultural models represent a statistical consensus of the mental models for a cultural group.
Cultural Models vs. Cultural Dimensions
Cultural psychologists have often
conceptualized culture in terms of lists of domain
general, stable traits, such as individualist-
collectivist value orientations. The intent of this
program is to find a core set of dimensions for
characterizing cultures that are believed by some
researchers to be important across a wide variety of
domains. The promise of this approach is to
provide a priori, purely analytical predictions about
cultural groups that are widely applicable to many
particular problems. The enterprise is successful if
the same small set of dimensions is predictive
across a wide variety of cultural domains and
groups.
There is some evidence at this point that general
cultural dimensions may not be as useful as one
might expect to predict cognitive or social patterns
within the context of specific situations. For
example, Sieck, Smith, & McHugh (2007) found
patterns of a work team orientation dimension that
were reversed from the predictions of
individualism/collectivism. Tinsley and Brett
(2001) also did not find individualism/collectivism
to be useful for predicting outcomes in US and
Chinese negotiations, but found that specific beliefs
about negotiation were useful. Osland & Bird
(2000) point to a number of cultural paradoxes that
arise in particular contexts from cultural
characterization in terms of general value
dimensions. The a priori analytical promise of the
dimensions approach is tempting, but so far the real
value is at best unclear.
CULTURAL NETWORK ANALYSIS
Cultural Network Analysis (CNA) refers to a
collection of methodologies for building cultural
models. CNA includes methods to:
• elicit and analyze the mental models of a sample
of individuals within the population
Proceedings of Phoenix Challenge 2008 4
Proceedings of Phoenix Challenge 2008 5
• measure the degree to which elements of the
mental models are shared across individuals and
develop cultural models
• represent the cultural model in accessible format
Figure 2 provides an abstract representation that
illustrates individuals’ mental models within a
cultural group, along with external cultural models
that have been extracted from the group using
Cultural Network Analysis. CNA is based on a
view of culture as comprising distributions of
networks of causally-interconnected ideas within
populations of investigation. CNA builds on a
synthesis of conceptually related methods for
knowledge elicitation, analysis, and representation
that stem from the diverse fields of naturalistic
decision making, cognitive anthropology, cognitive
psychology, and decision analysis. None of these
fields alone offers a comprehensive, end-to-end
approach for cultural modeling. CNA fills that gap.
Cultural Network Analysis comprises a
discovery phase and a consolidation phase. In the
discovery phase, concepts and mental models are
extracted from qualitative sources, such as
interviews and open source media (web news,
blogs, email), with little presupposition regarding
the elicited contents.
One goal of this phase is to develop an initial
understanding of the concepts and characteristics
that are culturally relevant within the domain. A
second objective is to obtain initial graphical
representations of target audience members’ mental
models in forms that closely match their own
natural representational structure. Qualitative
analysis and representation at this stage yield
insights that can be captured in initial cultural
models. Often, qualitative analysis may be all that
is needed for applications.
The discovery phase also generates a wealth of
material for constructing subsequent structured data
collection in a consolidation phase useful in
strategic communications situations involving
longer-term monitoring and evaluation of cultural
changes. In the consolidation phase of CNA,
structured interviews, field experiments, and
automated semantic mining of web-based sources
are used to obtain systematic data that is more
amenable to statistical analysis. Statistical models
used by cognitive anthropologists, such as cultural
consensus theory, are employed to assess the
patterns of agreement and derive statistics
describing the distribution of concepts, causal
beliefs, and values. Finally, formal representations
of the cultural models are constructed that illustrate
the statistical and qualitative information in
diagrams. Influence diagrams are an important
representation format for cultural models. Formal
representation makes it possible to use cultural
models in a variety of applied contexts.
Discovery Phase
Mental models are explanations about how
things work, and these explanations vary across
cultures. Mental models entail culture-specific
knowledge of the elementary concepts, as well as
how they are causally related. A cognitive
anthropological study by Garro (2000) provides an
example of obtaining mental models from cultural
groups through interviews. In a study examining
the cultural knowledge and understandings relating
to diabetes causation in a Native American
community, Garro conducted interviews following
an “explanatory model framework.” All of the
participants were members of the Anishinaabe
community who had been previously diagnosed
with diabetes. The researchers ensured that the
following aspects of their experiences were covered
in the interview:
• The cause of their illness
• Why it started and when it did
• The history of the illness
• The kinds of effects it has
• Possible and appropriate treatments for the
illness
Participants were also encouraged to talk more
generally about possible causes and ways of dealing
with diabetes, and to answer additional related
questions that arose from the responses given.
Based on the results, Garro constructed a
Proceedings of Phoenix Challenge 2008 6
hierarchically organized outline of the culturally
available understandings relevant to a cultural
model for sickness. The outline organized the most
common explanations of illness mentioned in the
interviews. First, the different types of sicknesses
and sickness explanations were identified. After
having inferred the major types of sicknesses the
causes, or perceived causes, were sorted into
sickness categories. The level of detail and
abstraction of the cause descriptions was dictated
by the level at which informants naturally shared
information.
CNA generalizes these cognitive information
requirements to build cultural models. In particular,
a CNA study to characterize the cognition of a
target audience should aim to capture the following:
• Basic level of simple concepts
• Distinct states of the concepts
• Positive/negative value associated with distinct
concept states and outcomes
• Antecedent causal factors that influence the
states of the concepts
• Consequences of positive/negative valued states
of the concepts
• Synthetic conditions, such as cultural artifacts
and institutions that influence concept states
Although mental models are described in
abstract terms, gathering specific cases and
incidents can aid in achieving concrete grounding to
tease out clues to participant’s mental models. For
example, Sieck, McHugh, & Smith (2006) elicited
incidents from participants in Lebanon and the US
who had participated in protests as a means to gain
access to Arab crowd members’ understandings and
expectations of how crowds work, and the decisions
that are made within crowds.
The difficulty inherent in getting information
from what people say, especially when they are
challenged with talking about very abstract
concepts, can also be circumvented by attending
more to the metaphors they use to say it. Systematic
analysis of metaphor use can provide reliable access
to tacit knowledge. The metaphorical concepts
employed by an individual can then be compared to
those employed by the group. Importantly, it is
possible to compare whether the metaphors used by
one cultural group have the same or different
implications for action (Quinn, 2005). If two
cultural groups use metaphorical concepts that have
different implications for actions to describe the
same domain, this entails that they conceptualize
the domain very differently. For example, in cross-
cultural studies of AIDS concepts, Wolf (1996) has
compared the war metaphors of the first world to
conceptualize the AIDS virus (e.g., "combating the
disease" and "killer cells") with the metaphors
employed in Malawi, where the virus is
conceptualized using metaphors of eating; the virus,
conceived of as a worm, eats up human beings.
Consolidation Phase
One issue with purely qualitative approaches to
the development of cultural models is the lack of
transparency or consistent guidelines in what
knowledge was deemed sufficiently shared to
include in the model. Strauss and Quinn state, “At
what point in the continuum of sharedness we
decide to call a given schema ‘cultural’ is simply a
matter of taste,” (p 122). Computational
approaches are required for consolidating the
qualitative discoveries about culturally shared
mental models, and further analyzing and
representing their distributions within and between
populations. Cultural consensus theory and
Cultural Mixture Modeling (CCM) are two tools
that can be usefully employed to meet those needs.
Cultural consensus theory is a collection of
formal statistical models designed to assess
concordance in knowledge and beliefs among a set
of respondents (Romney, Weller, & Batchelder,
1986). When a cultural consensus is found, it
provides the consensual responses that indicate
culturally shared knowledge and estimates of the
strength of consensus for those responses.
Individuals will also vary in the extent to which
their responses agree with the consensus, and that
variation is captured explicitly for each individual
under the rubric of “cultural competence.” Cultural
competence should not be confused with expertise,
Proceedings of Phoenix Challenge 2008 7
but rather with the degree of concordance with the
culturally shared model.
Additional analyses can be performed to
understand individual variability in cultural
competence, for example, by relating those CCT
measures to social network analysis measures. The
theory also enables the calculation of the minimum
number of respondents needed to assess the degree
of agreement. Assuming the data collection taps
into reasonably well-shared cultural knowledge,
then the number or respondents can be quite small,
e.g., 10 or fewer respondents. This is an important
feature for field use, which often aims to draw
conclusions from relatively small samples.
Cultural Mixture Modeling (CMM) is a more
recently developed statistical technique for
identifying shared cultural beliefs (Mueller, Sieck
& Veinott, 2007). CMM uses model-based
clustering techniques that allows one to determine
multiple distinct shared beliefs among segments
within a sampled population, as well as what those
beliefs are. It also provides a wider number of
metrics for assessing the cultural coherence of the
segments within the target audience, as well as the
cultural competence of each member.
Graphical Representation
We have developed a default approach to
representation for CNA so as to accomplish the
following:
1. Provide a standard pictorial form that shows
the concepts and causal linkages in a manner
that can be readily digested by IO end users who
need to routinely comprehend cultural models in
varied domains
2. Permit a direct means of representing the
statistical distributions of cultural knowledge,
rather than just the shared knowledge
3. Yield representations in a useful form for
developers of agent-based simulation and
analysis systems
Figure 1 presents a standard representation
format, illustrated with an initial Arab-American
cultural model of romantic relationships. It is an
influence diagram. In it, each node-link-node
combination represents influence, in the sense that
the value of the concept at the beginning of an
arrow affects the value of the concept at the arrow’s
point.
An influence diagram can present a relatively
simple and useful representation of a cultural model
of a domain that is tied to key judgments and
decisions that are of importance for an IO mission.
The diagram represents the “culturally correct”
concepts and linkages as determined by CCT or
CMM. Furthermore, CMM results can be used to
populate the numerical probability values in the
model for developing a technically useful
representation of a cultural model. The result in
this case is a summary of not only the shared
influence links across the population, but rather the
full distribution of beliefs, with probabilities
indicating the consensus on any particular link.
Although numerical analyses inform the final result,
the use of influence diagrams to represent cultural
models only requires that individual members of the
target audience be able to convey the qualitative
components and directions of the influences in the
diagram. They do not have to report quantitative
information.
APPLICATIONS
Culture is made up of contagious ideas, that is,
ideas that propogate effectively and durably within
a population (Sperber, 1996). Two broad objectives
of research within this cultural epidemiology
viewpoint are to characterize the current
distribution of mental models within the cultural
group and to understand the dynamics of culture.
Enhanced understanding of the current distribution
of mental models within a culture can form a solid
foundation for shaping and effecting cultural
change.
Fundamental cultural research seeks to address
why some ideas are more infectious than others, and
to explain the most widely distributed and long-
lasting ideas within a population. Research for
practical purposes of IO has a slightly different
Proceedings of Phoenix Challenge 2008 8
focus, as it is directed to influence decision making.
From a decision-making standpoint, we recognize
that many ideas may be pervasive but
inconsequential to decisions of practical interest.
Hence, CNA applied to IO begins by identifying the
critical judgments and decisions that meet IO
objectives. We then direct Cultural Network
Analysis to characterize the networks of causally-
interconnected ideas that are relevant to those
decisions in order to answer the questions most
pertinent to the goals of designing information
campaigns:
• How are networks of ideas organized in mental
models for this target audience?
• What is the distribution of mental models in this
target audience?
• In what ways are the distributions changing over
time?
• What ideas are especially successful/vulnerable
in the target audience?
By addressing these questions, CNA can serve
as a basis for developing influence operations and
composing effective strategic communications.
Using CNA to make explicit maps of a target
audiences’ cultural understanding related to specific
decisions can serve as a basis for inferring
otherwise implicit goals and intentions, and
determining what makes for culturally relevant
messages. Cultural models also allow for making
predictions concerning the effectiveness of a
message by providing the opportunity to assess
potential unintended inferences that individuals
with a certain knowledge structure might make. The
explicit content knowledge obtained using CNA can
also provide a starting point for modeling and
simulating the dynamics of within a given culture.
REFERENCES
D’Andrade, R. G. (1995). The development of cognitive
anthropology. Cambridge: Cambridge University Press.
Garro, L. C. (2000). Remembering what one knows and the
construction of the past: A comparison of cultural
consensus theory and cultural schema theory. Ethos,
28(3), 275-319.
McFarquhar, N. (2006). It’s Muslim boy meets girl, yes, but
please don’t call it dating. The New York Times, Sept. 19.
Mueller, S. T., Sieck, W. R., & Veinott, E. (2007). Cultural
Metrics: A Finite Mixture Models Approach (Technical
Report DAAD19-01-2-0009). Fairborn, OH: Klein
Associates Inc.
Osland, J. S., & Bird, A. (2000). Beyond sophisticated
stereotyping: Cultural sensemaking in context. Academy
of Management Executive, 14(1), 65-79.
Quinn, N. (2005). Finding culture in talk. New York:
Palgrave MacMillan.
Romney, A. K., Weller, S. C., & Batchelder, W. H. (1986).
Culture as consensus: a theory of culture and informant
accuracy. American Anthropologist, 88, 313-338.
Sieck, W. R., McHugh, A. P., Smith, J. L. (2006). Use of
cognitive field research methods to investigate cultural
groups: The case of individual decision making in Middle
Eastern crowds. In R. Sun & N. Miyake (Eds.),
Proceedings of the 28th Annual Conference of the
Cognitive Science Society, 2164-2168.
Sieck, W. R., & Rasmussen, L. J. (2007, September). Cultural
network analysis. Proceedings of the First Annual
Conference of the US/UK International Technology
Alliance, Baltimore, MD.
Sieck, W. R., Stevens, L. M., & Shafer, J. L. (2004).
Understanding target audiences: Individual and system
change processes (Technical Report 79258DBS26).
Fairborn, OH: Klein Associates.
Sperber, D. (1996). Explaining culture: A naturalistic
approach. Malden, MA: Blackwell.
Wolf, A. (1996). Metaphors of eating in the context of HIV
and witchery in Malawi. In: A. Wolf & M. Stürzer (Eds.),
Die gesellschaftliche Konstruktion von Befindlichkeit. Ein
Sammelband zur Medizinethnologie (pp. 205-221).
Berlin: Verlag für Wissenschaft und Bildung.
This research was sponsored by the U.S. Army Research Laboratory
and the U.K. Ministry of Defence and was accomplished under
Agreement Number W911NF-06-3-0001. The views and conclusions
contained in this document are those of the author(s) and should not
be interpreted as representing the official policies, either expressed
or implied, of the U.S. Army Research Laboratory, the U.S.
Government, the U.K. Ministry of Defence or the U.K. Government.
The U.S. and U.K. Governments are authorized to reproduce and
distribute reprints for Government purposes notwithstanding any
copyright notation hereon.
