Log in / Register
Home arrow Economics arrow Bounded rational choice behaviour
< Prev   CONTENTS   Next >

6.2.2. Measuring Mental Representations

Many cognitive mapping and mind reading approaches exist ranging from verbal to non-verbal techniques and from recall to recognition techniques. In the context of the present chapter, only verbal techniques are relevant and the most prominent ones of this category will be presented here.

Mental representations and hierarchical value maps from means-end-chain theory (Gutman, 1982) show structural similarities as both can be mapped as causal networks. In order to measure means-end-chains, several techniques proved to work and could hence also be adapted for measuring MRs. A widely used qualitative technique in this regard is laddering (Reynolds & Gutman, 1988). The original laddering technique is a structured face-to-face interview for understanding consumer's values (ends) and how they are trying to attain them (means). The interviewer starts by asking respondents to name the most important attributes of some choice products which are subject to investigation. For each mentioned attribute they then are asked why they consider it. Ideally, the responses can be classified as consequences. Accordingly, the interviewer continues asking why these consequences are important for the respondent until a satisfying level of ends has been attained. The resulting ladder or means-end-chain does hence consist of more than two levels of abstractness. The exact number of levels depends on the interview depth and the desired precision determined by the interviewer. So, consequences can for instance also be grouped into the more concrete physical consequences and the next highest level of psychosocial consequences. A laddering interview performed in the described manner does not provide support in the memory retrieval process in terms of revealed attributes, consequences and values. Thus, laddering can be classified as recall-based technique.

Since the emergence of a recognition-based variant of laddering (Botschen & Thelen, 1998) both versions are distinguished as soft (the recall-based version) and hard (the new version) laddering. Hard laddering presents predefined attributes, consequences and values from which respondents are asked to select the relevant ones. The sequence of the interview in the laddering format remains however the same. Next to soft laddering Russell, Busson, et al. (2004) and Russell, Flight, et al. (2004) applied a paper-and-pencil and a computerised version of hard laddering in an experiment on mothers' opinions of the role of breakfast on their children's physical and psychological well-being. In contrast to what was stated above about soft laddering Russell asked his respondents to select one to three important attributes from a list. Consequences and values were however elicited without auxiliaries by recall. The results showed that the hard laddering techniques yielded more ladders than soft laddering; a fact which is attributed to differences in participants' cognitive processing (recall vs. recognition). While Russell, Busson, et al. (2004) recommend hard laddering if the focus of the research is on investigating strong links between certain predetermined elements, soft laddering would be more appropriate for gaining a fuller picture of participants' cognitive structure. Yet, the drawbacks of a face-to- face interview remain which make soft laddering not suitable for large-scale surveys.

Ter Hofstede, Audenaert, Steenkamp, and Wedel (1998) suggested another measurement technique, called the association pattern technique (APT). Similar to the hard laddering variants respondents are faced with revealed attributes, consequences and values. The difference is only that the variables are not shown in list format and that the ladders are not elicited one-by-one. Rather, APT consists of two matrices (one for attributes and consequences and one for consequences and values) where respondents can indicate causal links by ticking off the corresponding cells. Hence, all ladders are elicited simultaneously which makes this technique quite difficult. The high complexity of the matrix format with which respondents might struggle can hardly be outweighed by the short interview duration. The advantage of APT is due to its simple analysis and the convenience it brings for the researcher. Thanks to the predefined labelling of variables no postprocessing of the responses is necessary, thus, making responses conveniently comparable. Yet, the downside of this convenience is that respondents are limited in their response freedom and possibly influenced by the revealed presentation of attributes, consequences and values. Furthermore, APT does not allow for a variation of the level of abstractness of the means-end-chains.

To collect data on MRs specifically in the context of decision tasks, a semi- structured interview protocol has been developed and tested in face-to-face sessions by Arentze et al. (2008) and Dellaert et al. (2008). The method assumes a choice task that may include multiple decision variables (e.g. the transport mode, destination and departure time for a trip). The so-called CNET starts by confronting the respondents with the decision variables in a random arrangement. They are asked to select them in the sequence in which they prefer to deal with them, assuming they were to make decisions. Next, the interview proceeds through the list of decision variables in the order indicated by the respondent and, for each variable, the respondent is informed about the decision alternatives and asked 4What are your considerations when faced with these alternatives?' From a list of predefined attributes and benefits, that is not visible to the respondent, those variables are identified that correspond to the response. If the response variable is not on the predefined list, the new attribute or benefit will be added. In any case, it is verified whether the respondent agrees with the classification and determined whether the attribute or benefit is causally linked to the decision variable. In case of doubts, these links are checked with the respondent. Having identified the variable, the next step depends on the variable type. If the variable is an attribute, the interview proceeds with the question 'Why is this variable important in this case?' This 'why' question generally results in the identification of an underlying benefit generated by the attribute, in which case no further 'why' questions are needed. If another attribute is mentioned, the 'why' question gets repeated until an underlying benefit emerges. When the originally mentioned variable is a benefit, the interview proceeds with the question 'How is this variable influenced?' and this 'how' question leads to the identification of other situational or alternative attributes. The causal links are also established and verified if in doubt. Further considerations are prompted by repeating this procedure until the respondent has no further considerations to mention. After the first decision variable is processed, the entire procedure is repeated for the next decision variable, and so on, until all decision variables are processed. Ultimately, this procedure leads to a completed representation of the attributes and benefits involved in respondents' MR of the decision problem, as well as the causal links among these attributes and benefits and the action variables involved in the decision. Finally, after the MR is completed, the respondent is asked to select, for each decision variable, the alternative that he or she would choose in the given scenario. An example of an MR elicited in this way is shown in Figure 6.1.

This protocol implies already that the interview is quite intensive and time- consuming. Each variable is processed step-by-step so that all components of the MR are captured. However, the repeated prompts for consideration might possibly evoke too much deliberation on the respondent's side so that he or she gives answers only in order to satisfy the interviewer. A somewhat tricky property of CNET is connected to its response freedom. Because respondents are not instructed in the labelling of the predefined variables the interpretation of their responses is subject to the interviewer. Still, a common set of variable labels is necessary to enable comparing MRs between individuals. The possibility to include even not predefined variables makes the MRs however strongly individually tailored. To eliminate possible interview bias and to allow application to large samples, an automated version of CNET has been developed and tested in Horeni, Arentze, Dellaert, and Timmermans (2014). The online tool is implemented in a PHP-based algorithm where a string recognition algorithm in cooperation with a predefined MySQL database takes over the job of the human interviewer.

Originating from the semi-structured CNET protocol Kusumastuti, Hannes, Janssens, and Wets (2009) developed modifications in order to measure MRs underlying leisure-shopping trip decisions. Their first modification is called CNET card game as it works with revealed variables printed on cards. Instead of eliciting the components of the MR by recall the interviewee indicates thus the relevant variables from card stacks which he goes through one-by-one with the interviewer. The second modification is a computerised version of the card game (CB-CNET).

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science