Measuring mental representations underlying activity-travel choices

The technological and societal challenges connected with the direct and indirect consequences of the still increasing traffic volume are keeping many people in research and practice busy. While some try to develop alternatives and travel demand measures to keep the traffic volume low others work on the improvement of travel demand prediction. Both have in common that they target human choice behaviour with their work. An essential condition for the success of travel demand measures and transport models is therefore to understand how individuals make their (travel) decisions and which needs they want to fulfill with their choices. The investigation of mental representations seems to be the key to understand human decision making. Mental representations are in fact images individuals bear in mind to oversee the consequences of their choices. They are tailored to the specific task and contextual setting under concern and show a significant simplification of reality. Next to the nature of the considered choice alternatives, the temporal construal of the task, the severeness of consequences and the (un)certainty of necessary information are held among others as determinants of mental representations. Shifts in the composition of mental representations are thus expectable when these contextual settings are changing. A drawback connected with the investigation of MRs is that so far only a few techniques exist by which these latent constructs can be elicited from individuals. Yet, all these methods are limited in the sense that they either influence or restrict respondents in their statements or are inappropriate for large-scale applications. This thesis introduced therefore a new online instrument for measuring mental representations which is able to collect data fully automatically. The first application of that instrument has its origin in the semi-structured CNET interview protocol from Arentze et al. (2008) and Dellaert et al. (2008). While online CNET is due to its open format still able to elicit an unbiased picture of respondents’ spontaneous recalls, adaptations to the original interview protocol had to be made to ensure the elicitation of benefits. In order to allow for a methodological comparison to online CNET an alternative application (online HL) has been developed that works only with revealed response options. As experimental subject a fictive trivial activity-travel choice task was chosen that consisted of scheduling working and grocery shopping activities for a normal working day in a fictive urban environment. In sum, decisions for the shopping location, the transport mode and the time of the shopping activity had to be considered. Side information was given for situational settings depending on the scenario. Next to the basic task four scenarios were developed of which one implied uncertainty about the side information, one implied a temporal distance of five years between the moment of decision making and the fictive moment of action, one introduced an additional online shopping alternative, and one implied negative consequences when the activity-travel task could not be fulfilled successfully. Data on these scenarios were collected among households subscribed to the nationwide Dutch LISS panel. The survey took place in two waves in spring and autumn 2010. While the first survey collected data on the basic, the uncertain and the distant scenario with both online CNET and HL, the second wave of experiments for the ecommerce and risky task was conducted with CNET only. In total, 1745 mental representations could be measured successfully which were subsequently analysed in an explorative and model-based approach. The analysis of the collected data showed a significant smaller structure of mental representations elicited with CNET compared to mental representations elicited with HL as the former consisted of significantly fewer components than the latter. This finding suggests an influence of the revealed handling of variables in HL which is supported by the fact that no shifts between scenarios could be measured with this technique. CNET however turned out to be sensitive for shifts caused by contextual manipulations. The substantial analysis of the uncertain, e-commerce and risky scenarios showed thus increased frequency and centrality values for attributes which were targeted by the experimental situations. For instance, the available product assortment nearly doubled its centrality value in the risky and uncertain scenario compared to the basic setting. Disappointing was however the distant scenario. An expected shift towards benefits could not be measured. A stable finding that was made with both techniques and in all scenarios was the high importance of the benefits time savings, ease of shopping and ease of travelling. These are in fact the driving forces of people’s choices for the investigated activity-travel task. These findings were supported by means of a formal model application which estimated parameters for MR component activation and strength of causal relationships in light of varying contexts. The analysis revealed that significant differences in MRs occur that result from situation-dependent need activation. The attributes on which choice alternatives are evaluated and the underlying benefits appear to be sensitive to (un)certainty of task-relevant information, the severeness of anticipated choice consequences and the set of choice alternatives. In conclusion, this thesis confirms the ability of online CNET to measure mental representations in a more sensitive and less influencing manner than online HL. Besides this scientific advantage CNET provides still all amenities of automatic online surveys for both respondents and researchers. These circumstances speak to the appropriateness of online CNET as a tool to elicit mental representations from decision makers of any choice task and perhaps also to a better understanding of human travel behaviour