Recommender SystemsThe full title of the project is: Recommender Systems for Group Consumption of Hedonic Media Products: Theoretic Modelling, System Development and Quality Testing.
DescriptionThe goal of the project is to significantly improve recommender systems by means of three conjoint approaches:
- Consideration of group consumption: Although hedonic media products are often consumed in groups, existing recommender systems implicate the isolated consumption by a single person. In contrast, our system also takes the group consumption into account.
- Consumerism theoretic foundation: Existing recommender system approaches are based on pure informatics criteria and neglect the user behaviour. Our aim is to deepen consumption research in the area of group behaviourthat yet only exists in attempts−in order to develop a consumerism theoretic model of group consumption of hedonic media products.
- Usability-Perspective: Although the relevance of electronic recommendations is substantially determined by the interaction- and user interface design, the current marketing and informatics research is almost exclusively focussed on the algorithms that are the base for the generation of recommendations and neglects the usability perspective. We therefore develop innovative group interactions models—focussing a participative decision-making process—for negotiating compromises in groups and prototypical implement these models for a user evaluation.
This research project develops a novel Web-based recommender system, which−by means of mathematical algorithms−analyses hitherto product ratings to automatically generate individualised recommendations for consumers in order to support the selection of hedonic media products like feature films, and so becomes an important aid for decision-making.
ApproachThe project DFG_Reco develops−in cooperation with all participating partners−innovative concepts and systems to support groups of users with personalised recommendations of choices to make that facilitate decision-making. The CML contributes to the DFG_Reco project with a domain-specific process model, which departs from a thorough literature review and user studies. It facilitates easy interaction among the group members and with group recommender systems. Our work is in three areas:
- Participative group interaction (i.e., all group members take part in the decision-making process actively)
- Simplification of the group recommendation process (i.e., groups only take relevant steps towards an effortless and casual decision-making)
We concentrate on the interaction- and user interface design, where we employ admitted design procedures (e.g., design principles specified in ISO 13407) of the field of Human-Computer Interaction and Computer-Supported Cooperative Work. The complexity of group recommender systems requires a sophisticated iterative development process including multiple detailed low- and high-fidelity prototyping cycles and user studies.
- Literature review of existing group recommender systems and approaches, as well as elaboration of a domain-specific process model
- Low-fidelity and high-fidelity prototyping with concluding user studies
- Implementation of a Web-based and mobile group recommender system, as well as a desktop tool for evaluators of group recommender systems
- Documentation and publication of the insights and requirements for new models, concepts, and applications for group recommender systems
Current State and ResultsThe project DFG_Reco has two project phases with individual research and prototyping cycles.
First Project Phase (Feb. 2009—Aug. 2011)First, the project DFG_Reco has performed a thorough literature review. The review revealed building blocks of group recommender systems and commonly known approaches like list-merging strategies for joining individual predictions to group recommendations. We aligned the user interaction- and interface design according to the domain-specific process model. In the first prototyping cycle, we created low- and high-fidelity prototypes that support ad-hoc groups in a mobile scenario. We built a distributed group recommender platform, which leverages on advanced concepts for group recommendation merging strategy exploration, performance optimisation, session persistence and mobility, as well as open interfaces. We implemented and tested a Web-based mobile group recommender application along the process model. The first prototyping cycle concluded with an evaluation in terms of expert interviews and a user study. The conducted user study revealed insights about the process model, the interaction design as well as the need for evaluator tools of group recommender studies.
In the second prototyping cycle we have conducted a low-fidelity prototyping with a concluding user study. We contributed the insights of this user study to improve future prototypes and implementations of our distributed group recommender platform.
Second Project Phase (Aug. 2011—ongoing)We started the second project phase with a literature study to ground for concepts from social psychology that are relevant for the decision-making process.
In the first prototyping cycle we started by observing the decision behaviour of groups going to cinemas. We analysed the ethnographically surveyed data in a creativity workshop for the purpose of enriching the process model and easing the interaction- and user interface design. Using low-fidelity prototypes explored the best conversion of the process model intro a user interface design. Therefore, we built team of two for designing three different flavours for a mobile application and compiled a final design in several small sessions with experts of different domains (design practitioner, organisational psychologist, and computer scientists). The resulting high-fidelity prototype—as a mobile application—implements the advanced group recommendation process. It concentrates on distributed group members, which want to reach consensus at the same time. Currently, we a conducting a user study on the first prototyping cycle as well as preparing the second prototyping phase, which will emphasise research on co-located group members with foot-size devices.
PartnersThorsten Hennig-Thurau, Department of Marketing and Media Research, University of Muenster, Germany
moviepliot GmbH, Berlin, Germany
PeopleTom Gross, Principle Investigator
Christoph Beckmann, Researcher
Sascha Herr, Researcher
Andreas Rösch, Researcher
Maximilian Schirmer, Researcher
FundingThe project is partly funded by the Deutsche Forschungsgemeinschaft (German Research Foundation); the funding period is Feb. 2009 to Sept. 2013.
PublicationsBeckmann, Christoph, Herr, Sascha and Gross, Tom. Gemeinsam einig: Partizipation in Gruppenempfehlungssystemen (Jointly Agreed: Participation in Group Recommender Systems; in German). In Mensch & Computer - 12. Fachuebergreifende Konferenz fuer interaktive und kooperative Medien - M&C 2012 (Sept. 9-12, Konstanz, Germany). Reiterer, H. and Deussen, O., eds. Oldenbourg, Munich, 2012. pp. 113-122. (ISBN: 978-3-486-71879-9).
Herr, S., Roesch, A.G., Beckmann, C. and Gross, T. Work-in-Progress: Informing the Design of Group Recommender Systems. In Extended Abstracts of the Conference on Human Factors in Computing Systems - CHI 2012 (May 5-10, Austin, TX). ACM, N.Y., 2012. pp. 2507-2512. (ISBN: 978-1-4503-1016-1/12/05).
Beckmann, C. and Gross, T. AGReMo: Filmempfehlungen fuer Ad-Hoc Gruppen auf mobilen Endgeraeten (AGReMo: Movie Recommendations for Ad-Hoc Groups on Mobile Devices; in German). In Mensch & Computer - 2011. Fachuebergreifende Konferenz fuer interaktive und kooperative Medien - M&C 2011 (Sept. 11-14, Chemnitz, Germany). Oldenbourg, Munich, 2011. pp. 335-338. (ISBN: 978-3-486-71235-3).
Beckmann, C. and Gross, T. AGReMo: Providing Ad-Hoc Groups with On-Demand Recommendations on Mobile Devices. In Proceedings of the European Conference on Cognitive Ergonomics - ECCE 2011 (Aug. 24-26, Rostock, Germany). ACM, N.Y., 2011. pp. 179-183. (ISBN: 987-3-86009-113-5).
Beckmann, C. and Gross, T. Towards Group Recommendations at Work. In Workshop Social Media at Work (WS 7) of the Twelfth European Conference on Computer-Supported Cooperative Work - ECSCW 2011 (Sept. 24-28, Aarhus, Denmark).
Gross, T., Beckmann, C. and Schirmer, M. GroupRecoPF: A Distributed Platform for Innovative Group Recommendations. In Proceedings of the Nineteenth Euromicro Conference on Parallel, Distributed, and Network-Based Processing - PDP 2011 (Feb. 9-11, Ayia Napa, Cyprus). IEEE Computer Society Press, Los Alamitos, 2011. pp. 293-300. (ISBN: 978-0-7695-3939-3).
Beckmann, C. and Gross, T. Towards a Group Recommender Process Model for ad-hoc Groups and On-Demand Recommendations. In Proceedings of the 2010 International ACM Conference on Supporting Group Work - Group 2010 (Nov. 7-10, Sanibel Island, FL). ACM, N.Y., 2010. pp. 329-330. (ISBN: 978-1-4503-0387-3).
Beckmann, C. and Gross, T. Leichtgewichtige Auswahlentscheidungen fuer Gruppen am Beispiel eines Recommender-Systems fuer Filme (Light-Weight Selection Decisions in Group Expemplified in a Recommender-System for Movies; in German). In Workshop-Proceedings der Tagung Mensch & Computer 2009 - 9. Fachuebergreifende Konferenz - Grenzenlos frei!? (Workshop-Proceedings of the Mensch & Computer Conference; in German)- M&C 2009 (Sept. 6-9, Berlin, Germany). Logos Verlag, Berlin, 2009. pp. 213. (ISBN: 978-3-8325-2181-3).