Group Recommendation with Latent Voting Mechanism

Novel automatic group identification approaches for group recommendation

Group recommendation has attracted significant research efforts for its importance in benefiting a group of users. This paper investigates the Group Recommendation problem from a novel aspect, which tries to maximize the satisfaction of each group member while minimizing the unfairness between them. In this work, we present several semantics of the individual utility and propose two concepts of social welfare and fairness for modeling the overall utilities and the balance between group members. We formulate the problem as a multiple objective optimization problem and show that it is NP-Hard in different semantics. Given the multiple-objective nature of fairness-aware group recommendation problem, we provide an optimization framework for fairness-aware group recommendation from the perspective of Pareto Efficiency. We conduct extensive experiments on real-world datasets and evaluate our algorithm in terms of standard accuracy metrics. The results indicate that our algorithm achieves superior performances and considering fairness in group recommendation can enhance the recommendation accuracy.

Group recommendation is a special service type which has the ability to satisfy a group’s common interest and find the preferred items for group users. Deep mining of trust relationship between group members can contribute to the improvement of accuracy during group recommendation. Most of the existing trust-based group recommendation methods pay little attention to the diversity of trust sources, resulting in poor recommendation accuracy. To address the problem above, this paper proposes a group recommendation method based on a hybrid trust metric (GR-HTM). Firstly, GR-HTM creates an attribute trust matrix and a social trust matrix based on user attributes and social relationships, respectively. Secondly, GR-HTM accomplishes a hybrid trust matrix based on the integration of these two matrices with the employment of the Tanimoto coefficient. Finally, GR-HTM calculates weights for each item in the hybrid trust matrix based on weighted-meanlist and proceeds to group recommendation with a given trust threshold. Simulation experiments demonstrate that the proposed GR-HTM has better performance for group recommendation in accuracy and effectiveness.

A group recommender system (GRS) is a system that collectively recommends items to a group of users based on their preferences. The GRS and the individual RS challenge lies in a very small and incompleteness of user-item ratings. Such incompleteness resulted in the data sparsity problem. The issues of data sparsity in a group negatively affect the quality of recommendations to the group. It occurs due to the inefficient formation of groups, which usually involves individuals with sparse data in their user profiles. Most of the current studies focus on this issue after the formation of groups. However, this study focused before the group formation, based on the intuition that it will be more efficient if the data sparsity at the individual level is addressed before the group formation process takes place. Therefore, applying the approach through Linked Open Data (LOD) technology is proposed to ensure that the data sparsity issues can be overcome before the group formation process is implemented. We proposed a GRS-LOD model. The experimental evaluations relating to the prediction accuracy and recommendation relevancy of the proposed model were implemented on three aspects: comparison with the basic approach or baselines; comparison with the current approaches, and comparison in terms of group size and aggregation strategies. The aggregation strategies used were the Average (AV), Most Pleasure (MP), Average without Misery (AVM), and Least Misery (LM). The metrics for prediction accuracy were based on the RMSE and MAE, whereas for relevancy, precision, recall, and F1-score were considered. The results show that the prediction accuracy and relevancy of the developed model’s recommendations is better than the baseline study by adapting the Average (AV) strategy with the individual profile aggregation approach. Meanwhile, for the evaluation in terms of group size, the results show larger group size exhibits better prediction accuracy for the four used aggregation strategies. On the other hand, in terms of recommendation relevancy, the result shows that relevancy decreases with the increase in group size for the MP, AV and AVM strategies.

Sequential recommendation and group recommendation are two important branches in the field of recommender system. While considerable efforts have been devoted to these two branches in an independent way, we combine them by proposing the novel sequential group recommendation problem which enables modeling group dynamic representations and is crucial for achieving better group recommendation performance. The major challenge of the problem is how to effectively learn dynamic group representations based on the sequential user-item interactions of group members in the past time frames. To address this, we devise a Group-aware Long- and Short-term Graph Representation Learning approach, namely GLS-GRL, for sequential group recommendation. Specifically, for a target group, we construct a group-aware long-term graph to capture user-item interactions and item-item co-occurrence in the whole history, and a group-aware short-term graph to contain the same information regarding only the current time frame. Based on the graphs, GLS-GRL performs graph representation learning to obtain long-term and short-term user representations, and further adaptively fuse them to gain integrated user representations. Finally, group representations are obtained by a constrained user-interacted attention mechanism which encodes the correlations between group members. Comprehensive experiments demonstrate that GLS-GRL achieves better performance than several strong alternatives coming from sequential recommendation and group recommendation methods, validating the effectiveness of the core components in GLS-GRL.

A POI group recommendation method in location-based social networks based on user influence

Group recommendation aims to recommend items to a group of users. In this work, we study group recommendation in a particular scenario, namely occasional group recommendation, where groups are formed ad hoc and users may just constitute a group for the first time—that is, the historical group-item interaction records are highly limited. Most state-of-the-art works have addressed the challenge by aggregating group members’ personal preferences to learn the group representation. However, the representation learning for a group is most complex beyond the aggregation or fusion of group member representation, as the personal preferences and group preferences may be in different spaces and even orthogonal. In addition, the learned user representation is not accurate due to the sparsity of users’ interaction data. Moreover, the group similarity in terms of common group members has been overlooked, which, however, has the great potential to improve the group representation learning. In this work, we focus on addressing the aforementioned challenges in the group representation learning task, and devise a hierarchical hyperedge embedding-based group recommender, namely HyperGroup. Specifically, we propose to leverage the user-user interactions to alleviate the sparsity issue of user-item interactions, and design a graph neural network-based representation learning network to enhance the learning of individuals’ preferences from their friends’ preferences, which provides a solid foundation for learning groups’ preferences. To exploit the group similarity (i.e., overlapping relationships among groups) to learn a more accurate group representation from highly limited group-item interactions, we connect all groups as a network of overlapping sets (a.k.a. hypergraph), and treat the task of group preference learning as embedding hyperedges (i.e., user sets/groups) in a hypergraph, where an inductive hyperedge embedding method is proposed. To further enhance the group-level preference modeling, we develop a joint training strategy to learn both user-item and group-item interactions in the same process. We conduct extensive experiments on two real-world datasets, and the experimental results demonstrate the superiority of our proposed HyperGroup in comparison to the state-of-the-art baselines.

Group recommendation plays a significant role in Social TV systems, where online friends form into temporary groups to enjoy watching video together and interact with each other. Online microblogging systems introduce the following relationship that reflects the common interests between users in a group and external representative followees outside the group. Traditional group recommendation only considers internal group members' preferences and their relationship. In our study, we measure the external followees' impact on group interest and establish group preference model based on external experts' guidance for group recommendation. In addition, we take advantage of the current watching video to improve context-aware recommendations. Experimental results show that our solution works much better in situations of high group dynamic and inactive group members than traditional approaches.

KTPGN: Novel event-based group recommendation method considering implicit social trust and knowledge propagation

The group polarization effect (risky shift) has often been identified in ad hoc leaderless discussion groups but never tested in a group with a leader. The present experiment extended the research by examining the effect on decision making of discussion in groups with and without a leader. Ninety-four college students' decisions on Choice Dilemma Questionnaire items (Kogan & Wallach. 1964) were examined three times: (1) individually prior to discussion, (2) as members of a group, and (3) individually after group discussion. In leaderless groups, members discussed the problems and came to a consensus decision. In led groups, the leader discussed the problems with advisors before making a decision. Instructions to the led groups stressed that the task of the members was to advise the leader whose choice was final. It is current belief that exchange of information alone and not group decision is what influences people's decisions (Myers & Lamm, 1976). If assigning the role of making a final decision to one person in a group does not significantly alter the information flow during discussion, then the decision of a leader should be equally polarized as the decisions of his advisors and the decisions of members in leaderless groups. Thus, it was hypothesized that post-discussion decision, using either a group or individual measure, would be more polarized than individual decision made prior to discussion for all people: members of leaderless discussion groups, leaders of discussion groups, and members (advisors) of led discussion groups.

Humans are generally biased to show implicit favoritism for in-group over out-group members, but developmental experiences may alter this process in important ways. Prior work has elucidated associations of family (i.e., in-group) violence exposure in childhood with risk for internalizing symptoms through weakened implicit favoritism for novel in-group members. The present study probes whether childhood violence exposure influences implicit bias and psychopathology differentially depending on the participant's relationship with the perpetrator (i.e., in-group vs. out-group member) at the time of exposure. We administered a minimal group assignment paradigm and implicit association test to 455 young adults aged 18-25. Young adults who experienced out-group violence in childhood showed stronger implicit in-group favoritism compared to those who experienced in-group or no violence. Implicit out-group favoritism was associated with increased alcohol use. Early-life experiences may shape innate preferences for novel in-group vs. out-group members in ways that have lasting implications for mental health.

Humans are generally biased to show implicit favoritism for in-group over out-group members, but developmental experiences may alter this process in important ways. Prior work has elucidated associations of family (i.e., in-group) violence exposure in childhood with risk for internalizing symptoms through weakened implicit favoritism for novel in-group members. The present study probes whether childhood violence exposure influences implicit bias and psychopathology differentially depending on the participant’s relationship with the perpetrator (i.e., in-group vs. out-group member) at the time of exposure. We administered a minimal group assignment paradigm and implicit association test to 455 young adults aged 18-25. Young adults who experienced out-group violence in childhood showed stronger implicit in-group favoritism compared to those who experienced in-group or no violence. Implicit out-group favoritism was associated with increased alcohol use. Early-life experiences may shape innate preferences for novel in-group vs. out-group members in ways that have lasting implications for mental health.

We present PolyLens, a new collaborative filtering recommender system designed to recommend items for groups of users, rather than for individuals. A group recommender is more appropriate and useful for domains in which several people participate in a single activity, as is often the case with movies and restaurants. We present an analysis of the primary design issues for group recommenders, including questions about the nature of groups, the rights of group members, social value functions for groups, and interfaces for displaying group recommendations. We then report on our PolyLens prototype and the lessons we learned from usage logs and surveys from a nine-month trial that included 819 users. We found that users not only valued group recommendations, but were willing to yield some privacy to get the benefits of group recommendations. Users valued an extension to the group recommender system that enabled them to invite non-members to participate, via email.

Many of the most important decisions in our society are made within groups, yet we know little about how the physiological responses of group members predict the decisions that groups make. In the current work, we examine whether physiological linkage from “senders” to “receivers”—which occurs when a sender’s physiological response predicts a receiver’s physiological response—is associated with senders’ success at persuading the group to make a decision in their favor. We also examine whether experimentally manipulated status—an important predictor of social behavior—is associated with physiological linkage. In groups of 5, we randomly assigned 1 person to be high status, 1 low status, and 3 middle status. Groups completed a collaborative decision-making task that required them to come to a consensus on a decision to hire 1 of 5 firms. Unbeknownst to the 3 middle-status members, high- and low-status members surreptitiously were told to each argue for different firms. We measured cardiac interbeat intervals of all group members throughout the decision-making process to assess physiological linkage. We found that the more receivers were physiologically linked to senders, the more likely groups were to make a decision in favor of the senders. We did not find that people were physiologically linked to their group members as a function of their fellow group members’ status. This work identifies physiological linkage as a novel correlate of persuasion and highlights the need to understand the relationship between group members’ physiological responses during group decision-making.

The methods and techniques introduced in the previous chapters provide a basic means to aggregate the preferences of individual group members and to determine recommendations suitable for the whole group. However, preference aggregation can go beyond the integration of the preferences of individual group members. In this chapter, we show how to take into account the aspects of personality, emotions, and group dynamics when determining item predictions for groups. We summarize research related to the integration of these aspects into recommender systems, and provide some selected examples.