Recommender systems help users to identify goods or servicestypically by offering suitable items from a broad range of alternatives. They have successfully spread into many domains. Tourism is a domain that has a huge potential for simplifying selections and decisions (e.g., on destinations; on itineraries; on accommodation; on cultural activities). In this position paper I discuss how groups of tourists can benefit from group recommender systems and give some examples.
CCS CONCEPTS
• Human-centered computing →
computing devices
Collaborative and social
1 INTRODUCTION
The increasing diversity of information, goods, and services offers
consumers a huge choice. At the same time finding the preferred
item can be challenging. Recommender systems help users
making choices by offering suitable items. They have spread into
many domains (e.g., book recommendations; music and movie
recommendations) [
Tourism is a domain with a huge potential for recommender
systems to help users to reduce the complexity of planning and
deciding, since ‘planning a vacation usually involves searching
for a set of products that are interconnected (e.g., transportation,
lodging, attractions) with limited availability, and where
contextual aspects may have a major impact (e.g., spatiotemporal
context)’ [
Group recommender systems support groups of users who want to share information, experiences, or products. Private travelling and touristic activities often happen in pairs or groups: people travel with a partner, people travel with family, people travel to meet friends, but also in business travelling colleagues might travel together or travel to meet working partners or colleagues. Here, group recommender systems are particularly suited, since ‘a group recommender is more appropriate and useful for domains in which several people participate in a single activity’ [8, p. 199].
Decision making is a core aspect of recommender systems,
since the basic assumption is that the system provides suggestions
helping users to make informed decisions [
In this position paper I share two examples for our own work
on group recommender systems: the AGReMo process model for
recommendation and decision processes; and the MTEatSplore
interactive tabletop applications for groups.
2 THE AGREMO PROCESS MODEL
The AGReMo (Ad-hoc Group Recommendations Mobile) process
model was conceived to serve as a blueprint for our group
recommender systems that aim to support the full cycle of a
recommender process starting with a preparation, followed by a
decision, and leading to action. Since it has already been
published elsewhere [
As Figure 1 shows, AGReMo consists of three principal phases:
The Preparation Phase kicks off the process by collecting all the required data that are needed to later estimate the predictions and generate recommendations. Each group member creates a personal profile. In our case the process model originated from a group recommender systems for movies, so the individual users created a profile and rated movies that they had already seen. After that the group members meet and elect an agent who interacts with the group recommender system (i.e., the assumption is that the whole group meets face-to-face and therefore only needs one system). Then the group can optionally specify group preferences and set preferred parameters. In our case of movie recommendations the group can pre-select cinemas and movies in the region. The group members can furthermore also optionally specify vote weights (i.e., the default was that all group members have equal influence on the recommendation generation, but the group can assign stronger weights to a member, for instance, as a courtesy or due to different levels of expertise). The active agent then requests recommendations, and the system generates group recommendations.
In the Decision Phase the group members receive the
recommendations with the best prediction on top. The group
recommendations are ranked according to the least misery
aggregation strategy (i.e., maximising the minimal prediction in
the group) [
In the Action Phase the group would go to the cinema together and each member is then asked to rate the watched movie to further develop their own profile. The group might also dissolve if no consensus can be reached.
This model was then instantiated in multiple apps (e.g., on app
for Android; another app for iOS) and explored with users and
based on real-time movie data that were retrieved from our project
partner moviepilot [
TABLETOP APPLICATION FOR GROUPS In a different project on a group recommender system we explored the suitability and affordances of interactive tabletops for supporting groups of users in choosing from a set of recommendations generated and presented by the tabletop app.
Here the primary focus was on a concept for the user
interaction and user interface for the group decision phase. We
started by developing paper prototypes that allows each team
member to pick their personal favourite restaurant and to suggest
it to the group through a drag-and-drop gesture towards the centre
of the table. The table then clusters and aggregates and counts
nominations as well as allows the group members to drill down
for textual as well as visual background information for the
respective restaurant. Figure 2 shows a scenario of the multi-user
multi-touch interaction with the restaurant recommendations.
Further details on the design process and results can be found it
[
In the Workshop on Recommenders in Tourism at the 11th ACM Conference on Recommender Systems I would love to discuss ideas and concepts for future work on the whole process of group recommender systems—including technical aspects on how to generate recommendations that reach broad acceptability as well as conceptual aspects of group decision making based on group interaction with recommendations.
ACKNOWLEDGMENTS I would like to thank the members of the Cooperative Media Lab. Part of the work has been funded by the German Research Foundation (DFG GR 2055/2-1). Thanks to the anonymous reviewers for valuable comments.