3.3.1 Attitude Theory
Attitudes are general evaluations people hold in regard to themselves, other people,
objects and issues (Petty and Cacioppo, 1986). As an example, consider a simulation
project that is investigating several options for improving the speed at which patients
are treated in an accident and emergency department. Assume one of these options
is to employ an additional nurse at peak times. The attitude towards the additional
nurse would be the degree to which a person has a favourable or unfavourable
evaluation or appraisal (Ajzen, 1991) of employing an additional nurse.
Attitudes are formed from the salient beliefs, the small number of beliefs that an individual can access at a given moment (Ajzen, 1988), about the object of the
attitude. The beliefs about the improvement option, in the example above, may have
been constructed from ‘direct observation, self-generated by inference processes, or
formed directly by accepting information from outside sources such as friends or
media’ (Ajzen, 1988, :7). For example, a manager within the A&E department may
believe that:
• The current nurses could work harder at a higher utilisation and improve performance to the same extent as employing an additional nurse.
The measurement of attitudes follows an expectancy-value model (Ajzen, 1991).
Attitudes are the product of the subjective likelihood that a belief is true and an
evaluation of outcome desirability. In the example above the manager may feel it is
very likely that working the nurses harder will improve performance and also finds
it very desirable to have both improved performance and higher value-for-money
from resources.
Figure 3.3 illustrates that the expectancy-value model takes account of multiple
beliefs. The measure of attitude is calculated as simply the summation of the n
beliefs (Ajzen, 1991). The only requirement for this calculation is that the beliefs
are a compatible sample. The usual way to achieve this is to sample beliefs for
a specific time period. For example, a simulation study may be concentrating on
improving performance of an A&E department over a six month period. Thus all
measurement instruments (typically questionnaires) would focus on eliciting beliefs
relevant to this time period.
Figure 3.3: Expectancy Value Model
Figure 3.3 also illustrates that attitude is a significant predictor of (management)
intentions. Large literature reviews have substantiated this claim (Ajzen, 1991, 2001;
Rouwette, 2003). In fact, this construct is typically referred to as an attitude towards
behaviour (Ajzen, 1991). This is relevant in a simulation study as the behaviour
we are interested in predicting is implementation action (Rouwette et al., 2010;
Rouwette, 2003).
behaviour is taken from a larger psychological theory of behaviour called the The-
ory of Planned Behaviour (TPB) (Ajzen, 1991). In the TPB there are two other
constructs that predict intention: perceived social pressure to perform or not to per-
form the behaviour (subjective norm) and perceived ease or difficulty of performing
the behaviour (perceived behavioural control). These are both relevant in a real
simulation studies involving multiple stakeholders (Rouwette et al., 2010). How-
ever, as the current study does not require this level of detail and will not involve
multiple decision makers, the discussion of these variables will not be taken further
at this point. Full details can be found in Ajzen (1991) and Rouwette (2003) and
are discussed in the limitations of the experiment in Section 11.5.5.
3.3.2 Persuasion Theory
Chapter 1 highlighted credibility as an issue within model reuse studies. That is,
decision makers (and modellers) may not have sufficient confidence to use a model as
an aid to decision making if it has not been built in-house. Chapter 2 reviewed the
DES, HCI, media and psychology credibility literature. As part of the psychology
literature review the concept persuasion through systematic and heuristic processing
of information was discussed. Systematic processing is a comprehensive, analytic
orientation in which individuals scrutinise all available information for its relevance
to their judgemental task (Chaiken et al., 1989). Heuristic processing lies at the
other end of the spectrum, requiring much less effort, and is theory driven (e.g.
trust experts or consensus implies correctness) (Chaiken et al., 1989).
The most relevant aspect of persuasion theory for credibility in model building or
reuse studies is the Sufficiency Principle taken from the Heuristic-Systematic Model
(Chaiken et al., 1989). The principle has two important assumptions. Firstly, it
assumes that people are economy minded and prefer less effortful to more effort-
Secondly, it assumes that in general people cannot know if their attitudes are com-
pletely correct. For example, it is well accepted in DES literature that there is no
such things as general validity of a simulation model; instead validation should be
viewed as a confidence building exercise (Robinson, 2004).
Given these assumptions, the sufficiency principle introduces the concept of a
sufficiency (confidence) threshold. Figure 3.4 illustrates the principle using a hypo-
thetical scale of confidence in a simulation model. For a given simulation project
individuals may have different confidence level thresholds that they deem sufficient
before the model can be used. In Figure 3.4 there are two individuals with thresholds
ST1 and ST2 respectively. Clearly what is sufficient for person one is not sufficient
for person two. This difference is important as it affects the type of processing and
effort the individuals will give. For example, as person two’s threshold is relatively
high compared to person one it is more difficult to achieve. Person two is much
more likely to need to undertake systematic processing to achieve his or her thresh-
old than person one. Person one will only expend as much effort as necessary to
achieve his or her threshold. As his or her threshold is lower it is more likely that
heuristic assessment will be enough (e.g the model visually looks correct).
Figure 3.4: Example of Sufficiency Thresholds
To illustrate the link to simulation studies and model reuse a dynamic hypothesis
is depicted using a causal loop diagram in Figure 3.5. The direction of the arrow
head in causal loop diagrams indicates the direction of causality while the sign of
the arrowhead indicates the effect of the causality (Pidd, 2004). The small loops
containing positive and negative labels indicate the reinforcing or balancing feedback
Verification and Validation (V&V) of a model using sufficiency thresholds.
Figure 3.5: Sufficiency Thresholds for Verification and Validation
A key variable affecting sufficiency thresholds is the personal importance of the simulation study to the decision maker (Chaiken and Maheswaran, 1994). This will be affected by an individual’s natural motivation to engage and possibly any com-
peting models of reality (e.g. I think the system works in this manner) that can also
explain system performance. High involvement in the conceptualisation and building
of the model increases personal importance of the study above what it would have
been. This in turn increases the threshold for where V&V of the model is viewed as
sufficient. Importantly a higher V&V sufficiency threshold increases the systematic
processing of information (Chaiken et al., 1989): model assumptions, logic and data.
This reinforces involvement in the study as re-conceptualisation may be required.
Balancing this loop is confidence in the validity of credibility assessments. The
higher the systematic processing of information the more ‘errors’ or inconsitencies
participate in this way the higher their ‘thought confidence’ becomes (Petty et al.,
2002). This continues until the V&V sufficiency threshold is reached. Thus it is pre-
dicted that clients with high involvement in building will apply more scrutiny to the
model and have high confidence in the validity of their credibility assessments. It is
also predicted that participants with low involvement in development will use more
heuristics in forming their attitude of credibility (e.g. the model has been built by
an expert) as their sufficiency threshold for V&V is lower. In other words, although
both model building and model reuse clients may rate the model as credible, model
reuse clients may have low confidence in the validity of their assessment of credibility
- a sort of heuristic or surface credibility (Tseng and Fogg, 1999). Model building
clients may hold a more systematic high confidence attitude to the credibility of the
model and its results as they have scrutinised the model to a greater extent.
This depiction agrees within one conclusion from Rouwette (2003): ’if the issue
[being modelled] is not sufficiently important [to the decision maker], learning effects
may be absent and thus implementation of modelling conclusions hampered’. That
is, systematic processing often confers greater attitude persistence than heuristic
processing (Chaiken et al., 1989). Personal importance aids V&V and the chances
of implementation.