A Semi-Automated Approach to the Content Analysis of Experience Narratives - Modeling Users' Experiences with Interactive Systems - page 134

concepts while a wider spectrum of relations would enable a richer exploration of
data. For instance, causal relations between concepts might be explicitly identified
by the researcher or automatically established by the relative position of concepts
within the unit of analysis.

6.6 Conclusion

This chapter highlighted two problems in the qualitative analysis of experience nar-
ratives. First, qualitative analysis is a labor intensive activity which becomes in-
creasingly a concern when qualitative data may be elicited from a large amount of
participants as in the case of iScale. Second, qualitative analysis has been shown to
be prone to researcher bias as humans often rely on heuristics in forming judgments
about the relevance or similarity of two or more data instances (Kahneman et al.,
1982).
This chapter proposed a semi-automated approach that aims at supporting the re-
searcher in the content analysis of experience narratives. This approach relies on a
combination of traditional qualitative coding procedures (Strauss and Corbin, 1998)
with computational approaches to the assessment of semantic similarity of docu-
ments (Salton et al., 1975). The approach shares a number of advantages over tra-
ditional content analysis procedures as the coding scheme derived in the analysis
of a small set of data is used to characterize all remaining data. Thus, through an
iterative process of coding and visualization of insights, the approach enables the
researcher in moving between highly idiosyncratic insights and generalized knowl-
edged. Secondly, as the researcher is forced to examine the use of language under
different contexts, it minimizes the risk of over-interpretation which is common in
traditional qualitative analysis practices.
Using data from chapter 4, the performance of the proposed approach was com-
pared to the one of a fully automated semantic analysis procedure, the Latent-
Semantic Analysis Deerwester et al. (1990). The proposed approach was found to
display a substantially closer fit to the results of manual clustering of narratives in
comparison to Latent-Semantic Analysis.