Mixed-initiative Recommender Systems: Towards a Next Generation of Recommender Systems through User Involvement Katrien Verbert Department of Computer Science KU Leuven, Belgium katrien.verbert@cs.kuleuven.be ABSTRACT visualization and control elements to different personal character- Researchers have become more aware of the fact that effective- istics. We found that musical sophistication has significant effects ness of recommender systems goes beyond recommendation accu- in a recommender system providing different UI controls [3]. In racy. Thus, research on these human factors has gained increased addition, both visual memory and musical sophistication are more interest, for instance by combining interactive visualization tech- likely to influence perceived diversity with more sophisticated vi- niques with recommendation techniques to support transparency sualizations [2]. These effects were sustained when studying the and controllability of the recommendation process. In this talk, I combined effect of controls and visualizations. These results allow will present our work on interactive visualizations to enable end- us to extend the model for personalization in music recommender users to interact with recommender systems as a means to incor- systems by providing guidelines for interactive visualization de- porate user feedback and input and to help them steer this process. sign for music recommenders, both with regards to visualizations In addition, I will present the results of several user studies that in- and user control. vestigate how user controllability interacts with different personal characteristics. REFERENCES In our initial work in this area, we elaborated TalkExplorer [5, 6], [1] B. Cardoso, G. Sedrakyan, F. Gutierrez, D. Parra, P. Brusilovsky, and K. Verbert. a cluster map visualization that enables end-users to interleave 2018. IntersectionExplorer, a multi-perspective approach for exploring recom- mendations. International Journal of Human-computer Studies (2018), in press. the output of several recommender engines with human-generated [2] Y. Jin, N. Tintarev, and K. Verbert. 2018. Effects of Individual Traits on Diversity- data, such as user bookmarks and tags, as a basis to increase explo- Aware Music Recommender User Interfaces. In Proceedings of the 26th Confer- ence on User Modeling, Adaptation and Personalization (UMAP ’18). ACM, New ration and thereby enhance the potential to find relevant items. To York, NY, USA, 291–299. address scalability issues of the cluster map, we also proposed In- [3] Y. Jin, N. Tintarev, and K. Verbert. 2018. Effects of Personal Characteristics on tersectionExplorer [1], using the scalable relevance-based UpSet Music Recommender Systems with Different Levels of Controllability. In Pro- ceedings of the 12th ACM Recommender Systems Conference (RecSys ’18). ACM, visualization technique [4] to allow users to simultaneously ex- New York, NY, USA, in press. plore multiple sets of recommended items. We evaluated the vi- [4] A. Lex, N. Gehlenborg, H. Strobelt, R. Vuillemot, and H. Pfister. 2014. UpSet: ability of IntersectionExplorer and TalkExplorer in the context of visualization of intersecting sets. IEEE transactions on visualization and computer graphics 20, 12 (2014), 1983–1992. conference paper recommendations. Objective measures of perfor- [5] K. Verbert, D. Parra, and P. Brusilovsky. 2016. Agents Vs. Users: Visual Recom- mance linked to interaction showed that users were not only in- mendation of Research Talks with Multiple Dimension of Relevance. TiiS 6, 2 (2016), 1–42. terested in exploring combinations of machine-produced recom- [6] K. Verbert, D. Parra, P. Brusilovsky, and E. Duval. 2013. Visualizing Recommen- mendations with bookmarks of users and tags, but also that this dations to Support Exploration, Transparency and Controllability. In Proceed- “augmentation” actually resulted in increased likelihood of finding ings of the 2013 International Conference on Intelligent User Interfaces (IUI ’13). ACM, New York, NY, USA, 351–362. relevant papers in explorations. Overall, the findings indicate that our multi-perspective approach to exploring recommendations has great promise as a way of addressing the complex human- recom- BIO mender system interaction problem. Katrien Verbert is an Associate Professor at the HCI research group When conducting user studies with IntersectionExplorer, we of the department of Computer Science at KU Leuven. She obtained observed some key differences with less technically-oriented par- a doctoral degree in Computer Science in 2008 at KU Leuven, Bel- ticipants. As a result, we started researching the effect of differ- gium. She was a post-doctoral researcher of the Research Founda- ent personal characteristics on the effectiveness (e.g., acceptance tion - Flanders (FWO) at KU Leuven. She held Assistant Profes- of recommendations, diversity, cognitive load) of interactive in- sor positions at TU Eindhoven, the Netherlands (2013 - 2014) and terfaces for recommender systems. These user studies were con- the Vrije Universiteit Brussel, Belgium (2014 - 2015). Her research ducted in the music recommender systems domain. We studied interests include visualisation techniques, recommender systems, the influence of different characteristics on the design of (a) vi- visual analytics, and digital humanities. She has been involved in sualizations for enhancing recommendation diversity, and (b) the several European and Flemish projects on these topics, including optimal level of user controls while minimizing cognitive load. The the EU ROLE, STELLAR, STELA, ABLE, LALA and BigDataGrapes results of three experiments show a benefit for personalizing both projects. She is also involved in the organisation of several con- ferences and workshops (general chair EC-TEL 2017, program co- chair EC-TEL 2016, workshop co-chair EDM 2015, program co-chair Joint Workshop on Interfaces and Human Decision Making for Recommender Systems, LAK 2013 and program co-chair of the RecSysTEL workshop se- Vancouver, Canada 2018. ries).