Conclusions - Modeling Users' Experiences with Interactive Systems - page 138

product and context, or they might simply lack validated measurement scales for a
relevant dimension. Secondly, such approaches assume that participants are able to
interpret correctly and position a given statement, that is defined by the researcher,
in their context. Recent literature has challenged this assumption, suggesting that
in certain cases participants are unable to interpret the personal relevance of the
statement in their own context; instead, they employ shallow processing (Sanford
et al., 2006), i.e. responding to surface features of the language rather than attaching
personal relevance to the question.
We argued for an alternative approach that lies in a combination of structured
interviewing, that aims at eliciting the attributes that are personally meaningful for
each individual, with a subsequent rating process performed on the attributes that
were elicited during the interview. The Repertory Grid Technique was proposed
as a promising attribute elicitation technique as it aligns well with parallel design
practices that are typically met in the concept design phase (Nielsen and Desurvire,
1993; Hassenzahl and Wessler, 2000).
Despite the promising nature of the Repertory Grid Technique, it was argued that
current analysis procedures are not suited for interpreting such data. Two dominant
approaches in the analysis of repertory grid data were identified: a) a qualitative, id-
iosyncratic approach with a narrative summarization, and b) a quantitative approach
that employs averaging procedures using exploratory multivariate techniques. We
argued that these two approaches are limited in two respects. Qualitative approaches
do not take properly into account the elicited attribute ratings and thus do not fully
exploit the true value of the Repertory Grid Technique, which is to quantify rich
qualitative insights. Averaging procedures, on the other hand, treat diversity among
participants as error and thereby contradict the basic idea of The Repertory Grid and
relevant personal attribute elicitation techniques.
Thus the lack of appropriate statistical analysis techniques was established and
two Multi-Dimensional Scaling procedures that aim at quantifying the diversity of
repertory grid data were proposed. The inadequacy of traditional averaging practices
was established as in two case studies they were found to account for 15% and 43%
of the available attributes only; the remaining attributes had minimal contribution
to the configuration space and thus could not be adequately modeled by the average
model. The technique proposed in chapter 3 was found to account for more than
double of the attributes accounted for by the average model, to provide a better
model fit even for the attributes that were adequately predicted by the average model,
and to result in semantically richer insights, since the diverse models can account
for more semantically different attributes.

7.1.3.2 Understanding the Dynamics of Experience through Experience

Narratives

Traditional approaches to measuring the dynamics of experience over time typi-
cally employ validated measurement and structural models across different phases
in the adoption of a system (e.g. Venkatesh and Davis, 2000; Venkatesh and John-
son, 2002; Kim and Malhotra, 2005). We highlighted a number of limitations of