Abstract:

The paper presents a newly developped and implemented system for segmenting respondents 'on the fly' during a computer assisted telephone interview, while at the same time minimizing the number of questions to be asked of the individual respondents. This is accomplished by using a model based segmentation scheme combined with ideas taken from artificial intelligence, especially expert systems based on probabilistic nets. In the paper the underlying the segmentation model is related to artificial intelligence and various ways knowledge representations. The opportunity of making the data collection intelligent or adaptive is explored and the implementation is demonstrated by way of a case study. The system offers several advantages. It makes segmentation based on a large number of variables, e.g. life style segmentation, operational in relation to follow-up surveys. Then, it reduces the costs of the interview in terms of money and respondent fatigue. Finally, it makes the segment variable accessible to the questionnaire designer for purposes of branching, skipping, and other conditional instructions.