%0 Conference Proceedings
%T Service-Oriented Justification of Recommender System Suggestions
%+ Università degli studi di Torino = University of Turin (UNITO)
%A Mauro, Noemi
%A Hu, Zhongli, Filippo
%A Ardissono, Liliana
%Z Part 3: Human-Centered AI
%< avec comité de lecture
%@ 978-3-030-85612-0
%( Lecture Notes in Computer Science
%B 18th IFIP Conference on Human-Computer Interaction (INTERACT)
%C Bari, Italy
%Y Carmelo Ardito
%Y Rosa Lanzilotti
%Y Alessio Malizia
%Y Helen Petrie
%Y Antonio Piccinno
%Y Giuseppe Desolda
%Y Kori Inkpen
%I Springer International Publishing
%3 Human-Computer Interaction – INTERACT 2021
%V LNCS-12934
%N Part III
%P 321-330
%8 2021-08-30
%D 2021
%R 10.1007/978-3-030-85613-7_23
%K Summarization of recommendation lists
%K Service blueprints
%K Explainable AI
%K Sentiment analysis
%Z Computer Science [cs]Conference papers
%X In the selection of products or services, overviewing the list of options to identify the most promising ones is key to decision-making. However, current models for the justification of recommender systems results poorly support this task because, as they exclusively focus on item properties, they generate detailed justifications that are lengthy to skim. Moreover, they overlook the existence of a complex item fruition process which can impact customer satisfaction as well. For instance, consumer feedback shows that relevant factors in home booking include both the properties of apartments, and previous customers’ perceptions of the interaction with the personnel who manages the homes. To address this issue, we propose a visual model that exploits an explicit representation of the service underlying item fruition to generate a high-level, holistic summary of previous consumers’ opinions about the suggested items. From this overview, the user can identify the relevant items and retrieve detailed information about them, in a selective way, thus reducing information load. Our model is instantiated on the Airbnb experiences domain and uses the Service Blueprints to identify evaluation dimensions for the incremental presentation of data about items. A preliminary user study has shown that our model supports user awareness about items by enabling people to quickly filter out the unsuitable recommendations, so that they can analyze in detail the most relevant options.
%G English
%Z TC 13
%2 https://inria.hal.science/hal-04292365/document
%2 https://inria.hal.science/hal-04292365/file/520517_1_En_23_Chapter.pdf
%L hal-04292365
%U https://inria.hal.science/hal-04292365
%~ IFIP-LNCS
%~ IFIP
%~ IFIP-TC13
%~ IFIP-INTERACT
%~ IFIP-LNCS-12934