Personal Attribute Judgments - Modeling Users' Experiences with Interactive Systems - page 23

Ta b l e 2 . 1 Percentages of attributes types from Karapanos and Martens (2007) (Study 1) and
Hassenzahl and Wessler (2000) (Study 2) studies.

Type Study 1 Study 2

Negation 67% 35%

Opposition 17% 26%

Non-contiguous 16% 39%

Hassenzahl and Wessler (2000) where non-contiguous constructs constituted the
major category (39%) while negation accounted for 35% of the constructs and op-
position for the remaining 26%. This observed discrepancy is likely to have been
influenced by two aspects: a) the instructions of the experimenter (our study had a
more evaluative character aiming at eliciting attributes that can be validly used as
semantic differential scales, while the latter study's primary goal was to inform de-
sign through rich qualitative accounts) and b) the fidelity of the prototypes (in our
study early concepts were communicated in sketched scenarios, while in the latter
study users interacted with working prototypes).
This highlights a problem rooted in the dual nature of the RGT: it aims at elicit-
ing rich qualitative accounts which can also be quantified. Non-contiguous attributes
provide insight into the relationships that individuals perceive between design quali-
ties (i.e. beautiful - hard-to-use) and concrete product attributes (i.e. easy to use - has
many buttons). They provide rich information to design. They are however inher-
ently problematic when used in psychometric scales as the two poles do not underly
a single uni-dimensional construct, and thus they will evidently elicit distorted rat-
ings. In our experience we have seen cases where, during rating, participants cannot
recall the context in which the attributes were elicited. When both poles of a bipo-
lar construct are not (equally) evident to the participant, ratings may very well be
based mostly on one of the poles. We would thus suggest that attributes should be
validated by the participant before moving to the rating phase. In this attribute val-
idation phase, participants can be asked to remove duplicate attributes and rephrase
attributes when needed.
Negation and opposition bipolarity constitute the common practice in validated
questionnaires. In a small study we attempted to explore the difference, if any, be-
tween negation and opposition bipolarity in rating scales. Our interest was to test
the opposition hypothesis, i.e. ratings for the negative and positive pole should have
a linear correlation of -1. Fourteen participants, all students at the department of In-
dustrial Design, rated three concepts on two negation (i.e. secure-insecure, practical-
impractical) and two opposition (i.e. standard-creative, amateurish-professional) at-
tributes, using paired comparison scales. The two poles of each scale were split in
two distinct scales; ratings on these two scales (e.g. secure and insecure) should
approximate a correlation of -1. All attributes except the one referring to security
were selected from Attracdiff2 (see Hassenzahl, 2004), a validated user experi-
ence questionnaire. Attributes had been translated to Dutch (Frens, 2006). These
three attributes were also identified in a repertory grid study with the three concepts