Be Inaccurate but Don’t Be Indecisive: How Error Distribution Can Affect User Experience

摘要: System accuracy is a crucial factor influencing user experience in intelligent interactive systems. Although known to be important, little about the role of system’s error distribution experience. In this paper we study, context background music selection for tabletop games, how an system affects user’s perceived particular, show that supervised learning algorithms solely optimize prediction can make “indecisive”. That is, it errors sparsely distributed throughout game session. We hypothesize harm users’ and preferable use model somewhat inaccurate but decisive, than accurate often indecisive. order test our hypothesis introduce ensemble approach with restrictive voting rule instead erring through time, errs consistently period time. A study which people watched videos Dungeons Dragons sessions supports hypothesis.