Towards creating a non-synthetic group recommendation dataset

Matthijs Rijlaarsdam, Sebastiaan Scholten, Cynthia C.S. Liem

Research output: Chapter in Book/Conference proceedings/Edited volumeConference contributionScientificpeer-review

1 Citation (Scopus)
135 Downloads (Pure)


Recommender systems can be useful in group settings, e.g. when choosing a movie to watch with a group. However, while considerable research in group recommendation has been performed, we still lack truly ecological datasets on group recommendations in real life consumption scenarios. Much of the existing work considers hypothetical consumption scenarios, and commonly, individual ratings are aggregated, but no actual group consumption takes place in which situational differences per group are taken into account. In this paper, we outline a vision for acquiring more realistic and ecological group consumption data, based on a crowdsourcing application that will acquire individual ratings per group consumption event. We discuss various design decisions that will allow us to gather these ratings effectively from a large group of people, and demonstrate and evaluate the viability of our approach towards reaching group consensus through rating session simulations.

Original languageEnglish
Title of host publicationImpactRS 2019 Impact of Recommender Systems 2019
Subtitle of host publicationProceedings of the 1st Workshop on the Impact of Recommender Systems co-located with 13th ACM Conference on Recommender Systems (ACM RecSys 2019)
EditorsOren Sar Shalom, Dietmar Jannach, Ido Guy
Number of pages5
Publication statusPublished - 2019
Event1st Workshop on the Impact of Recommender Systems, ImpactRS 2019 - Copenhagen, Denmark
Duration: 19 Sept 201919 Sept 2019

Publication series

NameCEUR Workshop Proceedings
ISSN (Print)1613-0073


Conference1st Workshop on the Impact of Recommender Systems, ImpactRS 2019


  • Crowd sourcing
  • Datasets
  • Group recommendation


Dive into the research topics of 'Towards creating a non-synthetic group recommendation dataset'. Together they form a unique fingerprint.

Cite this