Using fuzzy coding with qualitative data: example with subjective data in human-computer interaction

Abstract

ABSTRACTThis article shows the role that fuzzy sets may play in the prospect of analysing qualitative data. To underline this role, a human-computer interaction (HCI) study is presented. The data coming from 20 experts concerns their judgment regarding 33 questions related to the use of HCI approaches in order to support interactive system development phases. Each response scale features three main modalities, that is Agree, Partially agree and Disagree. The dataset example is analysed using multiple correspondence analysis (MCA) with both crisp and fuzzy coding models where the intermediate modality, Partially agree, is removed and considered with ½ membership values to the two extreme modalities. A comparative analysis is performed and the discussion states the interest of fuzzy coding with several kinds of qualitative factors or measurement variables. With qualitative measurement variables (our example), the main drawback of fuzzy coding could be the information loss, which is counterbalanced by the possibility of having fewer modalities and therefore of simplifying the multivariate analysis.