Supporting Material is available for this work. For more information, follow the link from
the Table of Contents to "Accessing Supporting Material".
Table of Contents
Models of Attitude and Belief Change from the Perspective of
System Dynamics
Ralph L. Levine, Ph.D.
Professor Emeritus, On Call
Departments of Resource Development
and Department of Psychology
314 Natural Resources Building
East Lansing, MI 48824
Ph. 517-353-6441 (O)
Ph. 517-332-2317 (H)
leviner@ msu.edu
ABSTRACT
This paper describes models of attitude and opinion change from several theoretical
perspectives. Current research in persuasion, argumentation, opinion setting, and
attitude change emphasize either cognitive, emotional, or behavioral factors, which
determine how people change their attitudes. The paper first reviews the pioneer
mathematical work of John E. Hunter and his colleagues and then assesses his models
for its loop structural characteristics. Simulation output as well as the structural
characteristics of these models indicates that behavioral approaches, such as imitation
and conditioning, are problematic in controlling attitudes through arguments, messages,
and behavior. Cognitive dissonance and information processing models appear to be
more effective in controlling attitudes. Finally, the paper concludes with an
embellishment of these models to show how cognitive searching processes can give time
to think about counterarguments and thus be used as a coping mechanism to resist
persuasive messages.
KEY WORDS
Attitude Change, behavioral, cognitive, dissonance approaches, credibility, source
derogation.
INTRODUCTION
Scientific interest in studying the dynamics of attitudes and beliefs has come from a
variety of disciplines, including social psychology, sociology, communication, and
marketing. Over the years, a number of theories of attitude change have emerged,
developed, and perhaps fallen by the wayside, as additional empirical studies cumulated.
There have been some efforts to model attitude change, but most of the theories in the
area are presented at the verbal level. Currently there are some efforts to use system
dynamics to capture basic generic structures in social psychology, which would include
the dynamics of attitude change (see, e.g., Levine and Doyle, 2002). In the past, one
pioneer in modeling attitude and belief change was the late John E. Hunter, a colleague
and friend, who also was one of the earliest developers of meta-analysis (Schmidt and
Hunter, 1977; Hunter, Schmidt, and Jackson, 1982; Hunter and Schmidt, 1990). Hunter
and his associates, namely Stanley Cohen, Jeffrey Danes, Scot Sayers, and the author
developed a series of models from around 1970 to 1985 that mathematically captured the
essence of most if not all of the theories of the day (e.g., see Cohen, 1971; Hunter,
Cohen, and Danes, 1984). The collection of models are very interesting from the point
of view of loop structure, and some of them are as relevant theoretically today as they
were thirty years ago. The purpose of this paper is to view a few of the models that are
still at the cutting edge of the theory of attitude change from the perspective of system
dynamics.
A Different Approach
The work of Hunter et. al. primarily focused on developing a set of difference equations
and change/state graphs representing a variety of approaches to attitude change,
including, behavioral, informational, balance, conformity and dissonance theories of
attitude change. Although much of their work is exciting, from the author’s point of
view, one limitation of their modeling effort was that Hunter et. al. did not publish
solutions to their equations. In the Hunter, Cohen, and Danes book (Hunter, Cohen, and
Danes, 1984), they showed a number of rate/state graphs that were relatively static in
nature. System dynamicists apply computer simulation to integrate the rate equations to
show behavior over time. In preparing for this paper, the author first carefully converted
Hunter et. al’s. difference equations to differential equations and then simulated the
model. Some fascinating effects missed by Hunter et. al. came out of simulating the
models.
What affects attitude change? Among social psychologists, there is general agreement
that credibility is a necessary condition for attitude change to occur. Credibility itself in
these models is primarily related to how the receiver feels about the source of the
messages being received. In addition, there might be individual difference or personality
variables, such as persuasibility, which is related to the time constant that amplifies or
inhibits attitude change.
Most of the models focused on how a person, the receiver, reacts to a set of messages
given by the second party, the source person. Each family of theories have a different set
of assumptions concerning how the receiver’s attitudes toward some object or other
person, such as a movie star, changes over time. In addition, each model attempts to
represent how the receiver’s attitude toward the source of the messages might change as
well as time goes on. All the models have two stocks, namely the receiver’s Attitude
toward the Object, a, and the receiver’s Attitude toward the source, s.
System dynamicists might be very interested in some of the behavioral mechanisms
represented in these models. For example, there is a possibility of a “boomerang effect”
that might occur when someone who you mistrust gives you a message that is totally
contrary to what you believe. In that circumstance, one might be prone to actually move
in the opposite direction from the source’s message. Also, if the receiver gets a message
that is contrary to his/her current attitude, one reaction might be to spend time finding
counterarguments rather than listening carefully to the line of arguments given by the
source person. This may inhibit the effectiveness of the message. Note that this second
mechanism is quite different from the boomerang effect where the receiver actually
changes attitudes in the opposite direction from the intended message. Finally, some of
the models displayed a polarity, which affects attitude change for those who harbor
extreme opinions and attitudes. This effect assumes that it is harder to change people
with extreme attitudes than people who have neutral attitudes toward a person or issue.
The same message generates more change for people who are neutral than for those who
hold extremely positive or negative attitudes and opinions.
In theory, attitudes can be either positive or negative, and yet cumulate over time so that
they play the role of levels or stocks. Unlike many material stocks used in system
dynamic models, attitudes must be considered as vector objects, having both magnitude
and direction. The value of the level could legitimately go negative if the source gives
the receiver a series of negative messages about another person, object, or issue.
BEHAVIORAL PERSPECTIVE
The first set of model suggests that attitude change obeys the rule that, if the message
agrees with the receiver’s attitude, then the receiver’s attitude is bolstered and intensified.
On the other hand, if the message disagrees with the receiver’s attitude, then the
receiver’s attitude is shaken and changes take place to lower or increase the receiver’s
attitude towards the value of the message. Some behavioral theorist stress the role of
imitation in attitude change. They assume that the receiver is passively taking the
source’s modeling behavior as a message. If the message and attitude match each other,
the receiver’s attitude is “reinforced, and increases the receiver’s attitude value in the
same direction. On the other hand, suppose the message and the receiver’s attitude
disagreed, i.e., one was positive and the other was negative. Then the receiver’s attitude
would be “punished”, and change would occur in the direction of the message.
A simple model of the agreement/disagreement principle is represented by a flow
diagram in figure | below. Although this may look very much like a simple stock
adjustment process because of the negative loop through the upper and lower bound
effect variable, one should be warned that the dynamics of change expressed in this
model is very different. It is also important to know that I assume that attitudes are
bounded. Something happens at the extremes. I have incorporat that assumption in this
model through the use of the variable, “upper and lower bound effect”. The upper and
lower bound effect is modeled as a fuzzy max and min function (see Sterman, 2000) to
keep the scales between the values of -10 to +10. The choice of those values was
arbitrary, although not in conflict with the psychological literature.
Figure 1. A simple version of the agreement-
disagreement reinforcement model
persuasability
OF S = > Attitude Toward
change in Object
attitude
sil ee
*
message from upper and lower
source
bound effect
Figure 2, below, which is in the form of a causal loop diagram may help in understanding
the implications of a “blind” reaction to either agreeing or disagreeing with the messages
given by the source person. As one can see, the loop deals with the fact that attitudes
normally have upper and lower bounds. In all of these models, I have defined the
attitudes as ranging from —10, extremely negative, to plus 10, extremely positive about a
person, an object, or an issue. In Figure 2, the only loop that comes into play is one
which only begins to dominate as the receiver’s attitudes become closer and closer to one
of the bounds.
persuasability
£ ninile toward
change in _>\Obieet
, attitude
message from
source cea |
ih and lower
bound effect
Figure 2. Causal loop diagram of the
agreement-disagreement model
Note the message affects the rate variable, change in attitude, directly. It does not play
the role of a goal in a typical stock adjustment process. The equation for the rate variable
is as follows:
Change in attitude = message from source*persuasibility*upper and
lower bound effect.
The rate equation, as an expression of policy, implies that strong messages, in either
direction, give the most change in the receiver’s attitude toward an object or issue. Note
also that this version of the behavioral approach has no boomerang effect.
Results of the Behavioral Model
The first run of this model shows what happens when initially, the source had a relatively
strong negative attitude toward an object, person, or issue, but the message was very
positive. Figure 3 shows the response of the receiver. The receiver reverses direction
toward the upper and lower bound of the message. However, and this is an important
point about this model, one can see that the attitude moves steadily and linearly to the
value of the message and then continued to grow in the positive direction beyond the
value of the message. There is very little control over the final value of the receiver’s
attitude by the message, unlike a typical proportional control negative loop process. To
see this even more clearly, consider the source that is only moderately positive about the
object and gives a series of moderate messages presumably to move the receiver’s
attitude level to the value of the message. One would think that the source would want
the receiver to become more moderate, i.e., to lower the attitude from its initial high
positive value. However, as seen in Figure 4, since the source and the receiver are both
positive, i.e., in agreement, the effect of the moderate message is to push the receiver’s
attitude toward the extremity, not down toward the moderate message value. This is the
problem with the reinforcement, behavior approach to attitude change. The source
cannot control the attitudes of the receiver very well. This is open loop control of sorts,
and as a social mechanism, one would think that human’s have evolved a better system of
managing social opinions and attitudes.
Response to an incomparible message
10
5
a \
\ Message from Source
5
Attitude toward Object
-10
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Time (Month)
message from source : m3 a-7 Attitude Units
Attitude Toward Object : m3a-7 Attitude Units
Figure 3. The attitudinal response to a positive
5 message when there is disagreement
Response to a compatible moderate message
10
— attitude toward Object
5
0
Message from Source
-5
-10
0 2 4 6 8 #10 12 14 16 18 20 22 24 26 28 30
Time (Month)
message from source : Message moderate positive agree—H—"—"
Attitude Toward Object : Message moderate positive agree
Figure 4. Response of receiver to a moderately positive message
Further Embellishment of the Behavioral Model
Most if not all of the models developed by Hunter et. al. dealt with changes in the
receiver’s attitude about the source itself, and at the same time dealt with the credibility
of the source as a major factor in speeding up or slowing down attitude change. Figure 5
shows a flow diagram of the situation where the receiver’s attitude toward the source also
affects the receiver’s attitude toward other people, objects, or issues and vice versa.
source
atu ay
change in attitude Toward
fo toward object yh
persuasability-alpha bound effect on ra as
towards Obj change in att
message from agreementowards source
source — indicator
bound effect on att
ease of change in
toward source
att toward source
Figure 5. Flow diagram of behavioral model
6 that includes attitudes toward the source
In this model, the rate equations are
ds/dt = agreement indicator*ease of change in att toward
source* bound effect on att toward source (1)
and
da/dt = Persuasibility*source credibility* bound effect on Att toward obj*m (2)
where the agreement indicator is the product of the message and the attitude toward the
object,
s = the Attitude toward the source
a= Attitude toward the object, and
m= message
The agreement indicator is the product of the Attitude Toward the Object. When the
message and the object attitude match, the rate of change in Att Toward Source is
positive. When the message and object attitude are mismatched, the Att Toward Source
will decrease. If the source and the receiver are not in agreement, then the receive will
lower his or her opinion of the source, according to this model.
This time there is a positive loop connecting the two stocks, which indicates that an
increase in the attitude toward the source increase credibility, which leads to an increase
in the attitude toward the object. An increase in the attitude toward the object would lead
to an increase in the agreement indicator, which would again increase the attitude toward
the source.
Results of the Embellished Behavioral Model
The question then becomes whether the control problem pointed out with the behavioral
approach to changing and controlling attitudes would remain with this additional
structure.
Message, Attitude towards object and attitude towards source
12
P <— Attitude toward the Source
0
Message from the Source
-6
-12
0 50 100 150 200 250 300 350 400 450 500
Time (Month)
message from source : a= 7 m=3 s =-6-————————_ att toward ojbect units
Attitude Toward Object : a= 7 m:
Att Toward Source : a= 7 m=3 s =-
=-6————_ att toward ojbect units
att toward ojbect units
Figure 6. Change in attitude toward the object, for a moderate
positive message and an initial negative attitude toward
source
As one can see from Figure 6, since both the value of the message about the object and
attitude toward the object are initially compatible, the source attitude moves to positive
ground. However, again object attitude becomes more extreme as time goes on, rather
than coming closer to the value of the message. The role of credibility in this model is
somewhat interesting, because credibility, which can enhance or inhibit object attitude
change, can really slow down the change in object attitudes. However, in general, the
open loop problem of trying to have the receiver’s attitude conform to the message is not
solved. At least from the perspective of this model, the behavioral approach is flawed.
Hunter et. al. did not emphasize or point out this problem of controlling attitudes with
messages. Let us now go on to look at other non-behavioral approaches to attitude change
that might solve this control problem.
COGNITIVE THEORY: INFORMATION PROCESSING
The information processing approach to changing attitudes conforms very nicely to many
of the elements seen in system dynamic models. Here the emphasis is on having the
receiver’s object attitude change as a function of the discrepancy between the message
and the attitude as shown in Figure 7. This is a case of the use of the classical stock
adjustment process, which is a form of proportional control. The information processing
approach originates from the work of Hovland and Pritzker (1957), Anderson (1959,
1964, 1971), Levine, Hunter, and Sayers, (1972). Hunter, Levine, and Sayers (1976).
message from
source
attitude
message gap
o;
Ct >| Att
change in att Towards
7 towards object Object
Persuasibilty
process
Figure 7. The simple version of the information processing
model of attitude change
The Receiver’s Attitude Towards the Source, and Source Credibility
The Mirror Model of Source Change.
The trajectory for this simple version of the information processing model of attitude
change is the well known logarithmic curve that converges to value of the message. Let
us consider a second stock, the receiver’s attitude toward the source, or source attitude, s.
How does the attitude toward the source change in an information theory framework?
There has been little said about the process of source change. However, Hunter et. al.
suggested several different possible ways that the receiver’s attitude toward the source
could change over time. The first approach is subsumed under what Hunter et al, p.52,
call the “mirror model.” It assumes that the only thing the receiver knows about the
source of the messages is the message itself, so the message not only tells the receiver
what the source thinks about the object, but the message also lets the receiver know
something about the source. The mirror model assumes that, if the source says something
nice, i.e. positive, then the source is nice. On the other hand, if the source says something
nasty, i.e. m is negative, then the source is not nice, i.e. nasty. The rate equation for
source change is
ds/dt = m —s (3)
The model can be seen in Figure 8.
attitude messag&“* __--—~ message
gap from
persuasibility source
Att Towards
change in att Object
towards object
—
message source
source credibility gap
Attitude
Toward }tt mS)
Source change in att
toward source beta adjustment fract
for source change
Figure 8. The mirror version of the information processing model
Initial results of running the mirror model. Does this model solve the problem of
controlling the receiver’s Attitude toward the Object, “a?” Figure 9 shows the results of
the test run utilizing the mirror model of source change.
10
Attitude Towards Object and Source for the Mirror Model
Attitude toward Object
TN
Attitude toward Source
-10
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
Att Towards Object : Huntmirror m3 a7 s-6
message from source : Huntmirror m3 a7 s-6
Attitude Toward Source : Huntmirror m3 a7 s-6
Figure 9. The mirror model of source change reacting to
a message that was less than the initial object attitude.
To see why one calls this a mirror model, consider the following situation: Let the source
attitude, s, equal —8, m = -3, and a=-6. Now, although m is negative, it is higher than s,
so s should move upward toward m.
The output of the model is found in Figure 10. As one can see, the mirror model simply
assumes that the attitude toward the source would go up if it is below the message value
or go down if its initial value were above the message value.
11
Attitude Towards Object and Source for the Mirror Model
Message from the Source
0 "4
<—_ Attitude toward the Source
-10
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
Att Towards Object : mirror m-3 a- 6s-8
message from source : mirror m-3 a- 6s-8
Attitude Toward Source : mirror m-3 a- 6s-8
Figure 10. The mirror model when the message is still negative,
But less negative than the initial value of the attitude toward the source
Cohen’s Model of Source Change.
Cohen (1971) developed a version of the information processing model that made a
different set of assumptions concerning the dynamics of source change. He focused on
the direction of the differences between the message and the attitude toward the object.
He assumed that the receiver’s attitude toward the source would increase under only two
conditions, as shown in Figure 11. First, assuming that the receiver’s current attitude
toward the object were positive, the attitude toward the source, “s,” would only increase
if the message, m, were greater than the object attitude, “a.” The second condition for
positive source change would be if the current Attitude Toward Object were negative, but
the message was more negative than the attitude toward the object. In this situation,
positive change in attitude toward the source would also occur.
When these conditions are not met, then the receiver would degrade the source. For
example, suppose the receiver hated the object or issue being evaluated, so variable, “a,”
might be —9.0. Now suppose the source sent a message, “m,” about the object that was
negative, but not that extreme, say m = -2. The receiver, according to Cohen’s model,
would interpret the source as being “wish-washy,” or too lenient, and would lower his or
her opinion of the source.
12
Let attitude towards Object =a
Let message = m
Let attitude toward source = s
The Attitude Toward the Source, s,
would increase under these two
conditions
-10 07 {+0
10 J f 0 +10
Figure 11 The assumptions of Cohen’s model of source change
Within an information process framework
Cohen’s model is displayed in Figure 12. He found a nonlinear function for the rate of
change in the source that was compatible with his assumptions. The rate variable, change
in att toward source, is a nonlinear, dimensionally correct function containing a second
degree polynomial, namely,
ds/dt = (alpha*a* gap) /SQRT( 1+(a‘2) )),
13
Where
alpha = adjustment fract for source change
a= Att Towards Object, and
gap = attitude message gap
attitude message message
from
gap
source
persuasibility
Att Towards
change in att Object
towards object
source credibility
Attitude
Toward |} 7)
Source change in att
toward source adjustment fract for
source change
Figure 12. Cohen’s model of source change.
Simulating Cohen’s information process model of attitude change. Figure 13 shows
the results of the first run of Cohen’s model. As one can see, Cohen’s model generates
very different trajectories from either the behavioral models or the mirror model. There
is a steady state error for the receiver’s attitude toward the object. This is because, since
the message, “m,” was below “a,” the receiver deceased his/her attitude to the lowest
attitude value, and since credibility is based on the source attitude, “s,” credibility went to
zero and the attitude toward the object went into equilibrium before it reached the
message value.
14
Attitude Towards Object and Source
0 '—_ Attitude toward the Object
5)
9 Message from the Source
5 eral Attitude toward the Source
-10
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
Att Towards Object : Cohen m-3 a7 s -6
message from source : Cohen m-3 a7 s -6
Attitude Toward Source : Cohen m-3 a7 s -6
Figure 13. The results of Cohen’s model of source change for
the test run
Cohen’s reversal effect. The model has an interesting prediction that, under suitable
conditions, the receiver’s attitude toward the source may start out one way, but reverse
itself as conditions change. Figure 14 shows the reversal phenomenon. Initially, “m”
was set to be quite negative, namely —7, a was set to be moderately positive, +5, and
finally “s” was set to be set to be barely positive, namely +1. Under this condition, the
attitude toward the source should fall, because it does not meet either of the two
conditions illustrated in Figure 11. Now two things are happening at once. The Attitude
toward the Object, “a,” is dominated by its negative loop process and begins to decrease
over time. Although simultaneously “s” is dropping, the key thing to look for is when a
goes negative around the fourth month. When that happens, both “a” and “m” are
negative and m<a. This meets one of the conditions for the Attitude toward the Source
to increase. Thus one observes a reversal in this variable as “a” continues to be more and
more negative, going into equilibrium when “a “equals “m.
This behavior was not seen before. It provides a very strong test of the differences
between these models. That is the beauty of model building. They frequently provide a
theoretical framework for good theory-based empirical research. This set of test
conditions will show up again when assessing the behavior of the models described in the
last part of this paper.
15
-10
Att Towards Object : Cohen m-7 a5 s 1
message from source : Cohen m-7 a5 s |
Attitude Toward Source : Cohen m-7 a5 s I:
Attitude Towards Object and Source
<< Att toward Object
Attitude toward Source
Message from the Source
20 30 40 50 60 70 80 90 100
Time (Month)
Figure 14. Reversing directions. The behavior of Cohen’s model when
the attitude toward the source initially decreases.
DISSONANCE THEORY
The last model of attitude change deals with the dynamics of dissonance, which is still
very much considered a viable theory and in vogue (e.g., see Meyers, 2000). Today,
much of the application of dissonance theory deals with how people cope with the
discrepancy between attitudes and action. Nevertheless the essential theory is the same
as thirty years ago.
Dissonance theory originated with the pioneering work of Leon Festinger (1957). The
main idea behind dissonance theory is that, when the receiver gets a message that is
discrepant from his/her own attitude or belief, the receiver begins to experience
dissonance. In order to dissipate this dissonance, in the situation described in this paper,
the receiver can do only two things to lower dissonance, namely either change the
attitude to decrease the gap, or change his/her attitude, toward the source, “‘s,” in the
negative direction. The direction of source change is very important in dissonance theory.
A dissonance model would never generate the trajectories found in Figures 10 and 14 for
example. Dissonance theory makes some very strong predictions about the direction of
change in the attitude toward the source, “s.
Although there could be several approaches to modeling dissonance theory, in this paper
we will show one relatively simple version that captures many of the features of the
verbal theory. The model assumes that the change in the attitude toward the object, “a,”
is identical in form to the information processing model previously discussed. Thus,
da/dt = credibility*(m-a), (4)
where credibility takes on the form found in the table function below.(Figure 15).
16
Graph Lookup - source credibility
1 a
.-
5 F
a
0 =
“10 0 1
Attitude toward Source
Figure 15. Table function showing source credibility as
a function of the Attitude toward Source
Derogating the source is also a possibility. It too should be a function of the gap between
the message, “m,” and the value of the Attitude toward the Object, i.e. (m—a). However,
the direction of this difference is not important, so the rate of change of “s,” is assumed to
be proportional to |m-a|. Another important assumption is that, if a source is completely
credible, there is no source derogation. This implies that the rate of source change is
inversely related to source credibility.
Putting these assumptions together, the following rate equation was used to capture the
dynamics of source change.
ds/dt = -B(1— credibility)|m — al, (5)
where f is the symbol for the model parameter, beta adjustment fraction for source
change.
Dissonance as an auxiliary. Finally, there is the matter of defining the concept of
dissonance per se. Hunter et. al. unfortunately did not treat dissonance as an emotional
state that increases when the receiver gets a discrepant message from the source and is
dissipated though either through attitude change or through source derogation, or a
17
combination of both processes. Instead, they chose to treat dissonance as an auxiliary
variable, or converter, defined as
Dissonance = a*credibility*|m —al, (6)
where for all runs « was set to 1.0.
A flow diagram representation of the dissonance model is shown in Figure 16 below.
Dissonance
attitude message
persuasibility gap
Att Towards
Object message
= from
change in att
fo towards object
source credibility
source
Attitude
Toward | — oD)
Source change in att
foward source: beta adjustment fract
for source change
effect of lower bound
. _ .. of att scale
source incredibility
Figure 16. A flow diagram representing the dissonance model
of object and source change.
Results of Runs for the Dissonance Model.
Base run: Source has no credibility, ie.,s =-10,a =+5,m=-7
As an initial run, let us start with the situation where there is a substantial gap between
the message and the receiver’s attitude toward the object, but the source has zero
credibility. Under these circumstances, even though there is a very large gap between the
message and the attitude, there should be no dissonance generated that has to be
dissipated. Thus, one would not expect any change in the attitude. Since the attitude
toward the source is set at its lowest level, -10, so that source credibility, which is a
18
logistic function of “s,” also equals zero.. There is no dissonance in this particular case.
The results of this run are shown in Figure 17.
RB Att Toward Object
6
0
Message from Source
-6
-12
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
Att Towards Object : dissec m-7 a5 s -10
message from source : dissec m-7 a5 s -10
Attitude Toward Source : disse m-7 a5 s -10-
Figure 17. Output of the model where source had no
credibility
As one can see, since there was no dissonance the message was completely ignored.
Run # 2. Source Has Perfect Creditability: s =+10, a =5,m=-7.
In this run, “s” is set at its maximum value, which in turn implies that the source is
perceived as perfectly credible. If that is the case, our rate equations indicate that there
should be no source derogation, only change in the attitude toward the object. Indeed, one
can see from Figure 18 that there was no source change, only the change in the attitude
toward the object.
19
12
6 Attitude toward Source av
0 err Att Towards Object
-6
“12 ~ Message from source
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
Att Towards Object : dissec m-7 a5 s 10
message from source : dissec m-7 a5 s 10-
Attitude Toward Source : dissec m-7 a5 s 10
Figure 18. The results of a simulation run when s = 10.
Was there any dissipation of dissonance? Figure 19 shows that dissonance decreased
over time because the receiver lowers his/her attitude toward the object to match the
message.
20
Dissonance
20
15
10
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
Dissonance : dissec m-7 a5 s 10
Figure 19. The dissipation of dissonance through changing
the attitude toward the object
Run #3. Simulating the dissonance model in the test run situation.
In the test situation, i.e, m = -7, a = 5, and s = 1, the theory would say that if source
change occurs at all, it will always be in the negative direction. The dissonance model is
incapable of showing a reversal effect, which was one of the most startling features of
Cohen’s information processing model. The results of the test run for the dissonance
model are seen in Figure 20. With this parameterization, both attitude and source change
occurred. However, as “s” became more negative, credibility went down to a point
where it inhibited the rate equation to a point where the variable, “a”, went into
equilibrium. The point of equilibrium, however, is in negative territory, according to
Figure 20. That would have been sufficient to generate a reversal effect on the attitude
toward the source, “‘s”., in Cohen’s model. The reversal never occurred in the dissonance
case.
21
12
6
+ Att Towards Object
0
6 Attitude toward Source
“2 N Message
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
Att Towards Object : disse m-7 a5 s1
message from source : dissc m-7 a5 sl
Attitude Toward Source : disse m-7 a5 s}:
Figure 20. The results of the test run for the dissonance model,
where m = -7,a=5, ands=1.
MODERN EMBELLISHMENTS.
This paper has presented some models that summarize theoretical frameworks that are
still in contention and seemingly interesting to researchers working on attitude change.
Currently, there is a strong emphasis on cognitive processes and other emotional
processes that might affect attitude change. The literature still talks about classical
conditioning theories and imitative “modeling” approaches to attitude change (Aiken,
2002). From a cognitive approach, one frequently sees research studies that show how,
under some circumstances, people will focus on the quality of the argument or go in the
opposite direction by spending time trying to find a counter-argument. Finding time for a
counter-argument takes time away from the persuasive aspect of the messages. This is a
mechanism for not feeling the full persuasive force of a message. It is a type of
mechanism to discount the message if the gap, m — a, is large. There is some interest in
modeling the dynamics of argumentation by system dynamicists. Having a model that
includes discounting might be useful in looking at argumentation dynamics.
An interesting place for the introduction of this cognitive process of discounting would
be to see if the reversal effect in Cohen’s model, in which initially s decreased over time
and then reversed itself when the attitude toward the object became negative. The
extended model is shown in Figure 21. If discounting occurs, and it is a strong
22:
mechanism used by someone to resist the influence of the message, then perhaps there
will be no reversal, and the receiver will continue to like the source less and less.
The equation for the discounting variable is
Effect of cognitive discounting = 1/(1 + *(m — ay), (7)
where i is the strength of the tendency to look for counter-arguments.
“a attitude messag
_» effect of cognitive gap
strength of processing message
counter arg from
source
Att Towards
change in att Object
towards object
persuasibility
adjustment fract fo
effects of bounds on source change
source credibility att toward source
Attitude
Toward > 75]
Source change in att
toward source
Figure 21. A flow diagram of Cohen's model of attitude change, in
which the receiver takes time out to think about counterarguments
Run 1. Thinking about counterarguments-persuasibility reset to .7 and J =0.1
Figure 22 below shows what happens to Cohen’s model when a cognitive inattention
process is introduced. For this run, the author increased the persuasibility parameter from
.3 to a very high value of .7. The parameter associated with the strength of the use of this
cognitive mechanism was set to .1 Figure 22 shows the steady decline of the receiver’s
attitude toward the source person. Note again that the attitude toward the object reached
equilibrium considerably above the message and in positive territory. Note that the
receiver’s attitude toward the source continued to go down after it crossed into negative
territory. There was not reversal effect found in this run.
23:
Attitude Towards Object and Source
10
5 ‘ Att Towards Object
0
<< _ Attitude Toward Source
-5 message from source
-10
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
Att Towards Object : Cohen m-7 a5 s Lhigh pers
message from source : Cohen m-7 a5 s Lhigh pers
Attitude Toward Source : Cohen m-7 a5 s Lhigh pers—— +"
Figure 22. The effect of the receiver’s inattention to the message
While searching for counter arguments.
Run #2 Checking For a Reversal Effect when i is extremely small
In the next run, the strength of the tendency to look for counterarguments was set to .01, a
relatively small value. Persuasibility was again set high at a value of .7. Figure 23 shows
the results. Initially, looking for counterarguments has a small effect of delaying the
movement of variable a towards the message. However, the variable, “a,” eventually
moved into negative territory and the attitude toward the source changed directions at that
point. Again, under those circumstances, i.e., when is extremely small, the model
generates a reversal effect.
Indeed, is seems to be an excellent potential policy lever if one can affect the tendency to
find counterarguments. From both sides, this could be useful for training be better
listeners as well as perhaps find ways to prevent people from using this mechanism when
messages are delivered.
24
Attitude Towards Object and Source
10
5 xm Attitude Toward Source
0
5 4—_ Attitude Toward Object
-10 + Message
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
Att Towards Object : Cohen m-7 a5 s Lhigh pers
message from source : Cohen m-7 a5 s Lhigh pers:
Attitude Toward Source : Cohen m-7 a5 s Lhigh pers ——_——+—"_
Figure 23. Cohen’s model when the was set to .01, implying
the receiver did not have a strong tendency to look for counterarguments
CONCLUSIONS
Three different approaches to attitude change have been presented. The first, which dealt
with behavioral explanations of how people change their attitudes, opinions and values
proved to be problematic. Developing a model and looking at its loop structure helped to
focus on the matter of control. If one of the purposes of the source person is to influence
and persuade the receiver with a set of messages, one would think that it should be
possible to move the person to the level of the message, not below or above the message.
The models showed the inability of the behavioral approach to have that degree of
control.
Information processing models easily solved the control problem. Both the mirror and
Cohen’s model utilized proportional control to persuade the receiver to endorse the value
of the message. Cohen’s model in particular, came up with an interesting potential social
phenomenon, where under certain conditions, the attitude toward the source would
reverse itself and move in the opposite direction from its initial response to the message.
The author independently derived a model that also gave a reversal effect. In any event,
we saw that the negative loop associated with the tendency to think about counter-
arguments could moderate this effect quite severely. Nevertheless, it is an interesting
effect. For example, students, who go into required courses, may come in with a set of
negative attitudes toward the course and perhaps the instructor. Sometime those students
25:
find that they like the course and shift their assessment of the instructor as well. It does
happen on occasion.
Finally, a third current approach to attitude change is to embrace a cognitive dissonance
perspective. The dissonance model, although interesting, perhaps needs some work on
developing the notion of the generation and dissipation of dissonance. On the one hand,
the verbal theory indicates that dissonance is an emotional state. Yet dissonance was not
treated as a level, accumulating and dissipating over time. One direction would be to
include an independent dissonance stock in future models
REFERENCES
Aiken, L. R. 2002. Attitudes and related psychosocial constructs: Theories, assessment,
and research. Thousand Oaks, CA: Sage Publications.
Anderson, N. H. 1959. Test of a model for opinion change. Journal of Abnormal and
Social Psychology. 59:371-381.
Anderson, N. H. 1964. Linear models for responses measured on a continuous scale.
Journal of Mathematical Psychology. 1:121-142.
Anderson, N. H. 1971. Integration theory and attitude change. Psychological Review.
78:171-206.
Cohen, S. H. 1971. Models of attitude change in passive communication paradigm:
Information processing, social judgment, dissonance, balance, and congruity.
Dissertation research, Michigan State University.
Festinger, L. 1957. A theory of cognitive dissonance. New York: Harper and Row.
Hovland, C. I., & Pritzker, H. A. 1957. Extent of opinion change as a function of amount
of change advocated. Journal of abnormal and Social Psychology. 54:257-261.
Hunter, J. E., Danes, J. E., & Cohen, S. H. 1984. Mathematical models of attitude
change: Change in single attitudes and cognitive structure, Vol. 1. Orlando, FL:
Academic Press.
Hunter, J. E., Levine, R. L., & Sayers, S. E. 1972. Attitude change in hierarchical belief
systems an its relationships to persuasibility, dogmatism, and rigidity. Human
Communication Research. 3:3-28.
Hunter, J. E., Schmidt, F., & Jackson, G. B. 1982. Meta-analysis: Cumulating research
findings across studies. Beverly Hills, CA: Sage Publications
Hunter, J. E., & Schmidt, F. 1990. Methods of meta-analysis: Correcting error and bias
in research findings. Newbury Park, CA: Sage Publications.
26
Levine, R. L., & Doyle, J. 2002. Modeling generic structures and patterns in social
psychology. Paper delivered at the 20" International Conference of the System Dynamics
Society, Palermo, Italy, July 28-August 1, 2002.
Levine, R. L., Hunter, J. E., & Sayers, S. E. 1972. Hierarchical structure and attitude
change: A multivariate approach. Research report, Computer Institute for Social Science
Research, Michigan State University.
Meyers, D. G. 2000. Exploring social psychology (2" Ed). Boston: McGraw-Hill.
Schmidt, F., & Hunter, J. E. 1977. Development of a general solution to the problem of
validity generalization. Journal of Applied Psychology. 62:529-540.
27
CC BY-NC-SA 4.0