Rhetoric of Truthfulness in the Battle Between Social Attributions and Empathic Emotions - Conflict and Multimodal Communication: Social Research and Machine Intelligence - page 298

PURE EMPATHY
R 1 =.64

.48

EMOTIONAL IMPACT
R 2 =.33

TRUTHFULNESS .58

.40

.64

SOCIAL
DESIRABILITY
R 1 =.40

.72

PARALLEL EMPATHY
R 1 =.52

Fig. 15.6 Path diagram on effects of truthfulness, emotional impact of story, and social desirability
on pure and parallel empathy in ingroup condition. Note: All paths were significant for p < 0.05.
For the endogenous variables, the estimated R -squared is given in the squares

pure empathy (Cronbach's alpha D 0.87), and parallel empathy (Cronbach's
alpha D 0.92). The model was first separately run on two groups of respondents
corresponding to the two categories of the target's group membership (i.e., ingroup
and outgroup) in order to assess the model fit for each condition. Subsequently,
a multisample analysis in which the two models were directly compared to each
other was performed on the data. The SEM approach (Bollen 1989 ) was applied
using the software Lisrel (Joreskog and Sorbom 1996 ), and the hypothesized
pattern of relationships among variables was estimated through a path analysis
model. In such a model, explanatory and dependent variables are all observed, and
constructs are considered to be directly measured, not assessed through a specific
measurement model. As a consequence, no measurement errors are estimated
by the procedure because all theoretical constructs are supposed to be fully
captured by the corresponding observed indicators. Relationships among variables
in the model might be directional or nondirectional. Directional relationships are
estimated through regression coefficients. Nondirectional relationships correspond
to covariances among variables. For each endogenous variable, the procedure gives
an estimate of the proportion of variance unexplained by structural relationships
(Bollen 1989 ; Joreskog and Sorbom 1996 ). Finally, all computed coefficients are
associated with a significance test, which allows the researcher to draw inferences
about the generalization of results to the population.
Figure 15.6 summarizes the path analysis model applied to the ingroup condi-
tion . The nonsignificant

2 value indicates that the proposed model does not differ
significantly from the observed data. This means that the model has a good fit to
the data in analysis [

2 (5, N D 33) D 1.72, p D 0.89, RMSEA D 0.00, GFI D 0.97,
AGFI D 0.93]. As expected, truthfulness affects emotional impact, which then
directly affects pure empathy and, through social desirability, the parallel empathy
dimension. Overall, as displayed in the diagram, the R -squared for the endogenous
variables in the model appears to be satisfactory.