A Semi-Automated Approach to the Content Analysis of Experience Narratives - Modeling Users' Experiences with Interactive Systems - page 119

Ta b l e 6 . 1 Domain-specific knowledge employed for the analysis of the 329 narratives. The
domain-specific concepts were derived from Hassenzahl (2004) and Lavie and Tractinsky
(2004) psychometric scales.

Concept Individual terms

Pragmatic Quality technical, human, complicated, simple, impractical, prac-

tical, cumbersome, direct, unpredictable, predictable, Con-
fusing, clear, Unruly, manageable

HQ-Stimulation typical, original, standard, creative, cautious, courageous,

conservative, innovative, lame, exciting, Easy, challenging,
Commonplace, new

HQ-Identification isolating, integrating, amateurish, professional, gaudy,

classy, cheap, valuable, non-inclusive, inclusive, unpre-
sentable, presentable

Classic aesthetics aesthetic, pleasant, clear, clean, symmetric, artistic

Expressive aesthetics creative, fascinating, special effects, original, sophisticated

6.3.2 Iterative Open Coding

Next to existing domain-specific knowledge, the researcher may want to annotate
the data using traditional coding practices (Strauss and Corbin, 1998). Coding con-
sists of two steps. In open coding the researcher derives concepts from terms or
phrases in the raw data. A concept may be defined through a direct relation to a spe-
cific term in the raw data ( in-vivo coding ; Glaser and Strauss (1967), for instance
the researcher may create a concept termed 'friend' referring to all terms 'friend' in
the raw data), or may relate to terms through an intermediate step of interpretation
(for instance, the researcher might desire to group terms 'friend' , 'family' and 'col-
league' under the concept named 'relevant others' ). In a second step, termed axial
coding , the researcher may define a hierarchical structure between concepts. The
creation of this hierarchical structure is realized through the definition of a superor-
dinate concept that includes all terms that appear in its subordinate concepts.
For instance, table 6.2 displays the full list of concepts along with coded exam-
ples, derived from the analysis of 329 narratives of the study described in chapter
4. Figure 6.1 displays a two-dimensional visualization of the similarity between
all concepts including those deduced from domain-specific knowledge (table 6.1)
and those derived from coding the data (table 6.2). Distances between concepts
were derived from equation 6.7 and submitted to Multi-Dimensional Scaling. A
two-dimensional solution was extracted. A-priori defined concepts are depicted in
italics, while superordinate concepts are denoted in bold. Note that similar concepts
(e.g. Novelty - Aesthetics in Interaction) often co-occur in the same narratives as
displayed by the high similarity in the two-dimensional space.
One problem often met in coding procedures is the fixation of the researcher onto
a certain perspective leading to increased biases in the interpretation process, what
Kahneman et al. (1982) call anchoring bias. Strauss and Corbin (1998) proposed a
number of techniques aimed at supporting the researcher in taking different perspec-