Eliciting Preferences to Find Your Perfect Laptop: A Usability Study

Abstract

Most e-commerce websites that sell technological items provide a simplified Queryby-Example preference elicitation approach to enable users to search for the product that they desire by specifying constraints over criteria and filtering results. Such an interface may be adequate when the technical specification is known. In this work, the usability of an alternative preference elicitation approach, Pairwise Comparisons, that enable users to specify trade-offs between various potentially conflicting criteria, is compared and contrasted to the traditional Query-by-Example approach. This is done by implementing two search tools that implement each preference elicitation approach to collect data about user behaviour. An experiment is carried out to evaluate performance and usability to search for laptops over a web scraped data set for varied tasks. It is found that while Query-by-Example tends to be preferred for cases where the technical specification criteria are precisely specified, the result sets generated for the Pairwise Comparisons approach are preferred when it is less clear what the criteria values should be. As such, the Pairwise Comparisons preference elicitation is deemed to be a worthwhile complementary approach for such e-commerce websites.