Infoscience

Thesis

Social Interface and Interaction Design for Group Recommender Systems

Group recommender systems suggest items of interest to a group of people. Traditionally, group recommenders provide recommendations by aggregation the group members’ preferences. Nowadays, there is a trend of decentralized group recommendation process that leverages the group dynamics and reaches the recommendation goal by allowing group members to influence and persuade each other. So far, the research on group recommender systems mainly focuses on the how to optimize the preference aggregation and enhance the accuracy of recommendations. There is a lack of emphasis on the users’ social experience, such as interpersonal relationship, emotion exchange, group dynamics, etc. We define the space where user-user interaction occurs in social software as social interfaces. In this thesis, we aim to design and evaluate social interfaces and interactions for group recommender systems. We start with surveying the state-of-the-art of user issues in group recommender systems and interface and interaction design in the broad sense of social applications. We present ten applications and their evaluation via user studies, which lead to a preliminary set of social interface and interaction design guidelines. Based on these guidelines, we develop group recommender systems to investigate the design issues. We then study social interfaces for group recommender systems. We present the design and development ofan experimental platformcalled GroupFun that recommends music to a group of users. We then study the impact of emotion awareness in group recommender systems. More concretely, we design and implement two different methods for emotion awareness: CoFeel and ACTI that visualize emotions using color wheels, and empatheticons that present emotions using dynamic animations of users’ profile pictures. Our user studies show that emotion awareness tools can help users familiarize with other members’ preferences, enhance their interpersonal relationships, increase the sense of connectedness in distributed social interactions, and result in higher consensus and satisfaction in group recommendations. We also examine social interactions for persuasive technologies. We design and develop a mobile social game called HealthyTogether that enables dyads to exercise together. With this platform, we study how different social interaction mechanisms, such as social accountability, competition, cooperation, and team spirits, can help usersmotivate and influence each other in physical exercises. We conducted three user studies lasting for up to ten weeks with a total of 80 users. Being accountable for each other’s performance enhances interpersonal relationships. Supporting users to cooperate on health goals significantly improve their number of steps. When designing competition in the applications, it is crucial to help users to choose comparable buddies. Finally, teamwork in exercises not only helps users to increase their steps, but also help them sustain in exercise. Furthermore, we present an evaluation framework for social persuasive applications. The framework aims at modeling how social strategies and social influence affect user attitudes and behavioral intentions towards the system. Finally, we derive a set of guidelines for social interface and interaction design for group recommender systems. The guidelines can help researchers and practitioners effectively design social experiences for not only group recommenders but also other social software [...]

Related material