On the Generation of Diverse Compound Critiques?

Critiquing is a powerful form of feedback often used by conversational recommender systems. There are two main types of critiquing; unit and compound. Unit critiques allow the user to provide limited feedback at the feature-level by constraining a single feature’s value space. Compound critiques, on the other hand, allow the user to manipulate multiple features simultaneously and therefore can help the user to locate the product they are looking for more efficiently. However, the usefulness of the compound critiquing approach is compromised when all the options that are presented to the user are very similar to each-other. In this paper we propose the idea of presenting diverse compound critiques, and evaluate the effectiveness of two alternative approaches in terms of their recommendation performance. Specifically, we look at the degree to which critique diversity can be improved, the effect this may have on user interaction, and its expected impact on recommendation efficiency.