The effects of transparency on perceived and actual competence of a content-based recommender

Abstract

Perceptions of a system’s competence influence acceptance
\nof that system [31]. Ideally, users’ perception of
\ncompetence matches the actual competence of a system.
\nThis paper investigates the relation between actual and
\nperceived competence of transparent Semantic Web
\nrecommender systems that explain recommendations in
\nterms of shared item concepts. We report an experiment
\ncomparing non-transparent and transparent versions of a
\ncontent-based recommender. Results indicate that in the
\ntransparent condition, perceived competence and actual
\ncompetence (in specific recall) were related, while in the
\nnon-transparent condition they were not. Providing insight
\nin what aspects of items triggered their recommendation, by
\nshowing the concepts that were the basis for a
\nrecommendation, gave users a better assessment of how
\nwell the system worked.