Analyzing Personal Attribute Judgments - Modeling Users' Experiences with Interactive Systems - page 55

space that is optimized for a small fraction of the dataset. We thus argue that despite
the limitations of the R k measure, averaging analysis fails to model a substantial
fraction of the dataset.

3.5 Conclusion

This chapter argued against averaging in the analysis of personal attribute judg-
ments. It was illustrated that when using averaging only 1/6th of the attributes in
our study, i.e. 18 out of 118, were taken into account. A new MDS procedure that
can better account for diversity in judgments was developed and its added value was
illustrated through the reanalysis of published data. The analysis resulted in three
diverse views on the data which were directly compared to the average view that is
the common practice in RGT studies. The diverse models were found a) to account
for more than double of the attributes accounted for by the average model, b) to
provide a better model fit even for the attributes that were adequately predicted by
the average model, and c) to result in semantically richer insights, since the diverse
models can account for more semantically different attributes.
It was further illustrated that diversity exists not only across different individuals,
but also within a single individual, in the sense that different attribute judgments of
a participant may reveal different, complementary, views. At any point in time in-
dividuals can have different, seemingly conflicting views. For instance, individuals
may regard one car as beautiful, but at the same time expensive. Individuals' overall
evaluations of the car might thus be modulated by contextual aspects such as their
motivational orientation (whether they just saw it in a newspaper on a Sunday morn-
ing or they are in the process of purchasing it, see Hassenzahl et al., 2008). Thus,
being able to understand individuals' conflicting views is crucial for understanding
how individuals infer the overall value of a product.
These insights strongly advocate the view that the analysis of quality judgments
of interactive products should not stop on a group level, but must be extended to the
relations between the attribute judgments within an individual. The Repertory Grid
combined with the suggested technique to analyze the resulting quantitative data
is an important step towards the adequate account of homogeneity and especially
diversity in individual quality judgments.