Conclusions - Modeling Users' Experiences with Interactive Systems - page 142

7.2.2 The End of Specifications?

The starting point of this research project was the observation that about half of re-
turned products are attributed to so called soft reliability problems, i.e. cases where
the user complaints despite the product meeting its specifications. The question
raised then is, can quality models rely on specification compliance as a metric for
the quality of a product? Brombacher et al. (2005) criticized traditional quality and
reliability models in that they assume a) the existence of a full set of specifications
that provide a complete description of the functional behavior of a product, and b)
that these specifications provide complete coverage of all requirements of the prod-
uct. This work further supported this view by illustrating that user requirements
may substantially vary across different individuals and that these requirements may
develop over time as the users appropriate the product in their own context.

7.3 Avenues for Future Research

We proposed a number of methods for quantifying the diversity in users' experi-
ences with interactive products. A number of limitations may however be noted,
leading to interesting avenues for future research.

7.3.1 Leveraging Insights across Different Exploratory Studies

One of the advantages of the Repertory Grid Technique is that it provides rich in-
sights into individuals' idiosyncratic views on the set of stimuli. We argued for
exploratory quantitative analysis techniques that can adequately model the differ-
ent views that one or more individuals have. A limitation of the Repertory Grid
Technique however lies in its lack of scalability across different empirical studies.
Each study yields a wealth of relations between stimuli, perceptions (i.e. personal
constructs) and evaluations (e.g. preference judgments). Leveraging the insights of
different empirical studies is cumbersome, if not impossible.
Structured interview techniques, however, such as triading (Kelly, 1955), ladder-
ing (Reynolds and Gutman, 1988) and pyramiding (Fransella et al., 2003), that are
typically employed in repertory grid interviews have the advantage of imposing a
set of relations between the constructs that are being elicited, and are thus computa-
tional friendly . Empirical insights could thus be modeled through a graph G=(V,E)
where the vertices V reflect the identified constructs and the edges E reflect the em-
pirically identified relations between the constructs, and each edge, i.e. a relation
between two constructs, could be weighted by its frequency of appearance in em-
pirical studies. This might lead to a new set of tools that could have a substantial
impact on science and design.