| One of the aspects of recommendation systems that has only rudimentarily been explored is the effect of the preference elicitation process on the evaluation of the recommendations provided by such systems. The goal of this study is to investigate potential influences of the content, form, and extent of the elicitation process on a number of evaluation criteria (perceived fit, affective attitude, cognitive attitude, general attitude, intention to choose), while simultaneously considering user perceptions of the elicitation process (perceived enjoyment and personalization) as factors that can potentially mediate the impact of elicitation process characteristics. Further, the study seeks to examine whether explicit measurement effects result from cue-based priming. Finally, the study explores the role of additional variables including gender, need for cognition, and trust regarding recommendation systems within the overall process of interaction and evaluation of the recommendation system.;A total of 401 subjects participated in a Web-based experiment which involved using a destination recommendation system for Spring Break vacation destinations. The data was analyzed primarily using a structural equation modeling methodology. The results suggest that the content of the questions asked during the course of the preference elicitation process does not matter much. In contrast, form of measurement---and to some extent the number of questions asked---have small but significant direct effects on the evaluation of the recommendation provided by the system. Gender, need for cognition, and trust in recommendation systems are identified as important aspects to consider when examining the persuasive potential of the preference elicitation process. Importantly, the results indicate that explicit measurement seems to persuade more than implicit measurement and lead subjects to make more favorable judgments based on explicit attitudes constructed at the time of preference measurement. The theoretical and practical implications of the research are discussed, followed by potential limitations and directions for future research. Specifically, it is argued that preference measurement in recommendation systems plays an important role and that especially its impact on trust and perceived personalization needs to be more extensively considered in recommendation system design. |
