Personal Attribute Judgments - Modeling Users' Experiences with Interactive Systems - page 22

does not take properly into account the diverse ways in which individuals construe
attributes to refer to internal concepts (Kelly, 1955). It is only when two attributes
agree both on semantics and in terms of ratings that one might be confident that
the two participants refer to the same concept. One could argue that in certain cases
two attributes refer to the same concept, but different participants value products
differently on the same concept. For instance, one participant might consider prod-
uct A more easy to use than product B with a second participant considering the
inverse, sharing however the same notion of ease of use. Even in this extreme case,
one should not group two attributes with diverse ratings. As it will become apparent
in the next chapters, we place ratings at higher importance than semantics, i.e. two
attributes are checked for semantics only when there is evidence that, within the
limited set of stimuli, they (quantitatively) represent the same internal concepts.
Second, techniques such as Principal Components Analysis (PCA) or Multi-
Dimensional Scaling (MDS) that are typically employed in the analysis of reper-
tory grid data assume homogeneity in the way people perceive the stimuli offered
to them. This is sometimes referred to as the principle of homogeneity of perception
(Martens, 2003). To our knowledge, however, all RGT approaches up to date have
been employing averaging techniques for the quantitative analysis of personal at-
tribute judgments. We believe this to be due to a lack of more advanced techniques
that can account for diversity in users' subjective judgments, eventually undermin-
ing the core value of the RGT, i.e. to account for diversity in individuals' subjective
judgments. We propose a quantitative, exploratory MDS procedure that aims at ac-
counting for the diverse views that one or more individuals may have on a set of
products. It will be demonstrated that even single participants can handle more than
one view on a set of stimul i. It will be shown that by averaging interesting views
are overlooked due to majorization bias.

2.3.2 On Bipolarity

A central notion in the RGT is the bipolarity of the idiosyncratic constructs (i.e.
attributes). Kelly, in his theory of Personal Constructs, postulated that individuals
perceive the world around them through the construction of dichotomous constructs.
It is our experience, however, that participants often need to be probed in order to
derive a truly bipolar attribute. This raises concerns with respect to whether indi-
viduals actually do think in bipolar terms. Lyons (1977) posited that “categorizing
experience in dichotomous contrasts is a universal human tendency which is only
secondarily reflected in language” . He identified three different types of bipolarity:
negation (i.e. practical-impractical), opposition (i.e. professional - amateurish) and
non-contiguous , where the opposite pole does not constitute a negation or linguistic
opposition (i.e. easy - powerful) (c.f. Yorke, 2001).
In a meta-analysis of a RGT study on early concept evaluation (Karapanos and
Martens, 2007), we found that the majority of the elicited constructs (67%) were
negation constructs, while 17% were opposition constructs and only 16% were
non-contiguous. This deviates substantially from what was observed in a study by