Evaluating Preference-Based Feedback in Recommender Systems

Abstract

Anecdotal evidence suggests that while real-world customerbuying models can be translated for online e-commerce scenarios, perhaps they should not be (because of the effort these models impose on the user). Currently the vast majority of recommender systems require the user to engage in a lengthy dialogue of information elicitation in order to locate the best product/service for them. We suggest that such lengthy dialogues are inappropriate in many e-commerce settings, and that alternative models are needed that minimise the cost to the user, while maintaining recommendation quality.

The support of the Informatics Research Initiative of Enterprise Ireland is gratefully acknowledged.