Diversity Exposure in Social Recommender Systems: A Social Capital Theory Perspective Chun-Hua Tsai Jukka Huhtamäki Penn State University Tampere University University Park, PA, USA Tampere, Finland ctsai@psu.edu jukka.huhtamaki@tuni.fi Thomas Olsson Peter Brusilovsky Tampere University University of Pittsburgh Tampere, Finland Pittsburgh, PA, USA thomas.olsson@tuni.fi peterb@pitt.edu ABSTRACT research stating that social capital is a key asset in knowledge Meeting other scholars at conferences is often a stochastic, work [5, 11, 22]. intuition-driven process. Social recommender systems can However, various human tendencies and cultural norms can support identifying new collaboration partners that one might hinder the process of finding the most relevant interaction not naturally choose. However, to boost the accumulation of partners for knowledge work [24]. Existing social structure, social capital, such systems must be designed for diversifying or the lack of one, tends to direct the emergence of new ties. social connections. This paper draws from the extant theory For junior scholars and other newcomers just entering the con- on social capital and diversity exposure in recommendation ference community, the identification of relevant individuals systems to discuss the importance of social diversity expo- or cliques is laborious and characterized by chance [23]. For sure and presents design directions for social recommender more senior scholars, a key issue, perhaps counter-intuitively, systems for building social capital. As preliminary empirical is the existence of their strong connecting tissue to the core of insights, we report the results of a field study of two diversity- the community that limits their networking capability. enhancing interfaces in an academic conference. Interestingly, we identified contradictory results between the subjective user In this light, conferences are seemingly fruitful contexts for feedback on the user interface quality and the objective analy- deploying information technology that could introduce oppor- sis of clicking and viewing the recommendations. This implies tune new ties. Social recommender systems are recommender that assessing the overall quality of a diversity-enhancing so- systems that, instead of items, seek to identify social connec- cial recommender system requires careful design of suitable tions, new or existing, relevant to the system user [15]. The measurements. objective of developing social recommender systems is to mit- igate issues related to information overload by ranking the Author Keywords potential connections according to their relevance. However, Interactive Recommender Systems; Social Recommendation; these relevance-first social recommenders have been shown Social Capital; User Interface; Diversity to narrow down users’ recommendation selection diversity [16]. We argue that increasing social diversity is a meaningful CCS Concepts goal in this context: the capability to create novel ideas and •Human-centered computing → Field studies; User inter- innovations has been shown to result from complementary face design; •Information systems → Social recommenda- viewpoints and heterogeneity of knowledge among a diverse tion; Recommender systems; group of actors [28]. To this end, we want to introduce two types of connections to the discussion on what could be rele- INTRODUCTION vant ties: weak ties that serve as bridging capital and strong Academic conferences are socially vibrant events where new ties for bonding capital [14, 27]. We see particular value in acquaintances are made and existing relationships are strength- forming weak ties in social events such as academic confer- ened. Scholars with different cultural and scholarly back- ences. grounds come together to discuss topics of their interests. From economic and sociological perspectives, conferences In this paper, we discuss the role of diversity exposure in so- are thus not only about information dissemination but oppor- cial recommender interfaces for supporting the formation of tunities for building social capital. We subscribe to extant new social connections in academic conferences. We believe that this line of research can help build social capital both for Copyright (c) 2020 for this paper by its authors. Use permitted under Creative Com- individual scholars (i.e., dyadic ties) as well as for the research mons License Attribution 4.0 International (CC BY 4.0). community as a whole (i.e., the overall social fabric of the IntRS ’20 - Joint Workshop on Interfaces and Human Decision Making for Recom- community). We claim that the objective to form weak ties mender Systems, September 26, 2020, Virtual Event is particularly important in serving both the conference core From social psychological perspective, the formation of new community and newcomers and give arguments to support the connections is governed by two key mechanisms, namely claim. Moreover, we demonstrate that taking a relevance-first choice homophily and triadic closure. Homophily refers to the approach to building social recommender interfaces will not preference to connect to individuals that are similar [19, 21]. support weak tie formation. The contribution of the paper Similarity in background, knowledge, and interests contributes is two-fold. First, the key contribution is a theory-based dis- to the ease of forming a connection and starting a meaningful cussion on the design directions of diversity-exposing social exchange of views and information. Triadic closure states that recommender interfaces. Second, we present and discuss the new connections are likely to form between actors that share findings of a preliminary field study of two visual interfaces a strong tie, e.g., between friends of friends [14]. In combi- built to expose users of social recommender systems to so- nation, these two mechanisms produce “striking patterns of cial diversity, that is, social connections that are outside their observed homophily” [19] and form echo chambers [31], that existing social circles. is, groups of densely interconnected actors that have limited connections to other such groups. THEORETICAL FRAMEWORK AND RELATED WORK An individual builds social capital one connection at a time. Both strong and weak ties have their role in sourcing informa- Social Capital in Knowledge Work tion [2]. We see particular value in forming weak ties in social Social capital is one of the components of human capital or events such as academic conferences. First, for system level broader cultural capital [5]. In their seminal article (cf., [29]), benefits, weak ties connect existing social clusters and sup- Nahapiet and Ghoshal [22] define “social capital as the sum of port the flow of information. Second, as networking strategy, the actual and potential resources embedded within, available forming weak ties serves both the community core members through, and derived from the network of relationships pos- and newcomers through the spread of ideas and knowledge. sessed by an individual or social unit” and point to previous Optimizing for weak ties points new connections away from research stressing the importance of including both the social the community core, toward newcomers. network and the resources available through the network un- der social capital [5]. Social capital is embedded in the social network and exists “in the relations among persons” [11]. Social Recommendations at Academic Conferences Starting in early 2000, some pioneer conference support sys- More specifically, two types of social capital are identified in tems offered conference attendees an opportunity to create extant research, bonding and bridging [27, 7]. Conceptually, profiles that provide relevant information about themselves bonding and bridging social capital are close to strong and and to explore profiles of other attendees [30, 10, 4]. This weak ties [14], respectively. Bonding capital consists of strong functionality is now offered by many commercial conference ties, i.e., actor’s strong connections to close colleagues, family, support systems, but support to social exploration remains and friends. Bridging capital is based on weak ties, connec- limited due to the passive nature of these systems and contin- tions beyond daily social life that form between clusters or ued challenges to find and recognize relevant attendees among groups of social actors. hundreds. To increase attendees’ chances to learn about each Why is social capital important in knowledge work? Simply other, some conference systems offered more proactive solu- put, social capital makes “possible the achievement of certain tions. For example, public displays and their combination with ends that in its absence would not be possible” [11]. Social sensors were used to stream information about random dele- capital is argued to be a key driver of organizational advantage gates or those who are nearby [30, 10]. Public displays were [22]. Bridging capital has a particular role for individual also used to allow two or more people to examine common actors, organizations, and communities. Weak ties, although topics of interest and future co-authors [18, 20]. limited in their bandwidth, serve as important conduits for More recently, recommender systems emerged as an attractive novel information [14]. Weak ties form across structural holes approach to make the process of accumulating social capital in social networks and put the connecting actors in the role at academic conferences more proactive [1]. Recommender of a broker. Brokerage “between groups provides a vision of systems are valuable in a time-sensitive conference context options otherwise unseen, which is the mechanism by which due to their ability to adapt to users’ interests and select list of brokerage becomes social capital” [8]. Further, weak ties are most relevant items or people from a large pool of candidates. shown to improve the innovativeness of managers [28]. While the majority of research on recommendation in confer- ence context focused on paper recommendations, a number of Social Capital Accumulation projects explored various approaches for people recommen- We agree with Dobson [12] and Archer [3] that social structure dation [9, 26, 34, 35]. The increased popularity of this topic is dynamic and evolves over time, that social structure is a key brought the issue of diversity in attendee recommendation to driver of social activity, and that social activity is a key driver the research agenda, e.g., social recommendations [15, 34]. of the evolution of social structure. Bourdieu [5] highlights the We examine this issue in the next subsection. processual nature of social capital, stating that “The existence of a network of connections is not a natural given, or even APPROACH a social given, constituted once and for all by an initial act Next, we move to describe our proposed solution to increas- of institution [...] It is the product of an endless effort at ing social diversity exposure in academic conferences. We institution.” claim that a straightforward strategy to depart from the two key mechanisms of social connection formation, that is, ho- Table 1: Post-experiment survey questions mophily and triadic closure, is to maintain similarity as the key relevance criteria and increase the social distance between the Q1 The interface helps me to explore various active user and recommendations. If successful, this strategy interesting people in the conference. will also support the accumulation of bridging social capital Q2 It is helpful to see people attributes like because the connections between socially distant actors are by Title, Country and Position when exploring definition weak ties. However, instead of taking an algorith- interesting people in the list mic approach to limit the users focus on actors that are similar Q3 The interface helps me to perceive the di- and socially distant, we propose that social recommendation versity of explored attendees. system designers should focus on increasing the controllability Q4 The interface helps me to improve my trust and transparency of the system as means to carefully nudge in the people recommendation result. the user toward options with potential long-term benefits for Q5 The interface helps me to understand why both the individual and the entire community. specific attendees were recommended. Q6 I like the people recommendation result We test two alternative user interface designs, Scatter Viz from the system. and Relevance Tuner, to explore how might these interfaces Q7 I became familiar with the system very expose the active user to social diversity. The two interfaces quickly. were originally proposed by [34, 35, 36] who evaluate their Q8 Overall, I am satisfied with the system. usefulness in devising recommendation selection in controlled Q10 It was useful to see the explanation of user studies at conferences. The interfaces are designed from scores produced by different recommenda- alternative perspectives to enhance recommendation selection tion components. diversity. Relevance Tuner is a straightforward extension of the Q11 It is fun to use the system. ranked list. Scatter Viz follows a more exploratory approach. Q12 The system has no real benefit for me. Interface 1: Scatter Viz (Figure 1) combines a scatter plot visualization with a standard ranked list to present recommen- dations. The scatter plot view presents the recommended item grates the AMiner dataset [32]. The live system is available at in two dimensions defined by the active user and has been http://halley.exp.sis.pitt.edu/cn3/. shown useful in helping people in the analysis of the large datasets [25]. The design concept is supporting the user’s This demonstration will extend the evaluation to explore the perception of two dimensions of a multi-relevance recommen- role of the two interfaces in exposing the active user to more dation. By selecting different dimensions to X-axis and Y-axis, diverse social connections in a real-world context through A/B the user can correlate multiple types of relevance among the testing. That is, how does the proposed interface design lead social recommendation. to social exposure and the possible change in social capital? Moreover, we are interested in users’ subjective experience Interface 2: Relevance Tuner (Figure 2) takes an alternative of the two proposed interfaces in terms of diversity exposure. approach to help the user to inspect the multi-relevance social Following the discussion earlier in this paper, we defined two recommendations. The user interface provides five relevance research questions for the demonstration as follows: RQ1: sliders that allow the active user to adjust the weighting of How is the effect of social diversity exposure perceived by the features and, by doing so, re-ranking the social recom- users? RQ2: What are the social exposure patterns implied by mendations with a standard ranked list. The user can tune the the two proposed user interfaces in an academic conference? weighting for each recommendation feature and the ranked list will reorder on the fly. Next, we will put these designs into Study Procedure play in a real-life setting to see what kinds of social diversity exposure patterns their use results in. 1) We sent out an invitation email three days before the con- ference date to introduce the social recommendation feature EXPERIMENT: FIELD STUDY available in CN3 to all the 170 attendees of the conference. The email contained the login and ID/Password information, Study Design so the conference attendees can click and link to the experi- To explore the social diversity exposure impact of the two mental system. 2) The subjects were assigned to the Scatter interface designs, Scatter Viz and Relevance Tuner, we orga- Viz (Scatter group) and Relevance Tuner (Tuner group) inter- nized an in-the-wild experiment as a field study in EC-TEL face, respectively. We equally split the conference attendees 2017 conference held in Tallinn, Estonia in September 12–15, into the two groups based on the system-generated user ID 2017. The two interfaces are integrated into Conference Nav- (odd or even), resulting to 85 users in each group with a cus- igator (CN3) [33], a social recommendation system for aca- tomized link to the assigned testing interface. 3) The assigned demic conferences. The recommendations are mostly based users were free to explore the system for four days during the on data collected by the CN3 system [6]. To mitigate the cold conference event. We collected the system log for the four start problem that occurs when users have no publications conference days. The log data includes the frequency of users or co-authorship information related to the event for which logging in to the system, clicking on the social recommenda- the recommendations are produced for [33], the system inte- tions, and the duration of each session. 4) A post-experiment Figure 1: Scatter Viz: (A) Scatter Plot; (B) Control Panel; (C) Ranked List; (D) User Profile Page. The user can inspect the recommendations with two relevance dimensions in the scatter plot. Figure 2: Relevance Tuner: (a) Relevance Sliders; (B) Stackable Score Bar; (C) User Profiles. The user can inspect the recommendations with multi-relevance dimensions. online survey (5-point scale) was sent three days after the Social network analysis provides an expressive way to inves- conference date to collect the subjective feedback from the tigate complex interaction patterns at a macro level [37]. We users. take a visual network analytic approach [17] to investigate the diversity exposure implied by the two interfaces. The ap- proach allows the investigators to observe patterns in social structure and to share their findings [13]. We create and visual- Data Analysis and Measurements ize users’ pre-existing social networks and project their social The social recommender systems should support the accumu- exploration on top of this network. The resulting network vi- lation of social capital, but it is hard to measure the effects sualization enables inspecting the patterns of existing and new that the system had to this end, particularly in such a lively social links. Here we define the “existing social connections” empirical context. We assumed the user must first find and using the co-authorship of the conference publications. We view the profile of a potential connection in the social recom- assume that conference paper co-authors have an existing so- mendation system. Hence, for the purposes of this experiment, cial connection. Moreover, we define “new social connections” we operationalize “social exposure” as clicks on the social rec- using user clicks logged when the users were using the CN3 ommendations that inspecting the profile of the recommended system. The click pattern helps to observe the social exposure scholars. That is, if the user clicks on the scholar who is of new connections and their topological placement in the recommended by the system, we will consider it as social network. exposure. Figure 3: Post-Experiment Survey result shows that the Relevance Tuner interface was preferred by users in almost all aspects except perceived diversity. We did not perform significance tests due to the small sample size. RESULTS co-authorship. 2) Green nodes are active users in the experi- The demonstration produced a total of 97 social recommen- ment and green edges are the clicks of social recommendations dation clicks by 32 participants that used the system under (social exposure). 3) Light gray nodes are conference atten- the two different treatments. There were 21 active users in dees that were clicked during the experiment. 4) Edge weight Scatter Group with 75 observations (M=3.33, SD=3.92) and encodes the number of clicks and co-authored papers. The 11 active users in Tuner Group with 22 observations (M=2.00, network is directed with edges pointing from the active user SD=1.48; M=mean, SD=standard deviation) within a four- to the scholars they clicked. 5) Node size is proportional to day period (period of the main conference events). The data its weighted indegree, that is, the sum of the weights of con- shows that the Scatter Viz users performed more clicks when nections pointing to a node. Indegree is the simplest network exploring the social recommendation compared to Relevance metric for authority. Tuner users. The demonstration results are consistent with the The visual network analytic can be summarized in four-fold. previous study in that Relevance Tuner requires fewer clicks First, the identical pattern in the two networks is the high den- to explore the desired conference attendees [34]. We found sity of the co-authorship network. In such a network, should the users in Scatter Group tended to try a few clicks to interact the social recommender system follow the relevance-first ap- with the visualized interface. This result also supports our proach, the active users would be more likely to get exposed to assumption on the steeper learning curve of the Scatter Viz the scholars at the core of the network. For newcomers, this is interface, due to the higher number of clicks required to look not a major issue because most of the conference participants up information on recommended scholars. are new to them. However, senior scholars would be likely User Perception: We sent out the post-experiment question- to get exposed to each other, resulting in triadic closure and naire three days after the conference and received a total of 13 choice homophily. responses, three from Scatter Group and ten from Tuner Group. Second, compared to Relevance Tuner, the clicking pattern in Interestingly, the response rate of the Tuner Group is much Scatter Viz is more visible. A few of the conference attendees higher than the Scatter Group (14% vs. 90%). The response that have not co-authored a paper are brought in through clicks; rate may indicate the usability of the user interface—users the click-based connections form a couple of new bridges into may be more likely to submit their feedback when the user the already densely connected network. However, it is possible experience is positive. The post-experiment survey (shown that the difference in the pattern is first and foremost a function in Table 1) gives further support for the usability of the Rel- of the number of clicks. evance Tuner over Scatter Viz. However, post-experiment survey results also suggest that users self-reported to perceive Third, the few clicks that the Relevance Tuner users performed higher social diversity using the Scatter Viz. do not merit any analysis. However, an interesting contradic- tion is that the Relevance Tuner users did not view essentially Social Exposure Pattern: In order to explore their actual be- any of the recommendations and yet the questionnaire indi- havioral patterns, we project the clicks that the users made cated a preference toward Relevance Tuner. to recommended scholars on the conference co-authorship network. Figure 4a & 4b represents the social diversity expo- Fourth, the visible structural differences in the two networks sure pattern for Scatter Viz and Relevance Tuner, respectively. also show in network metrics. The Scatter Viz network has 4 The visualization can be interpreted by the color and size components and 103 nodes in total out of which 96 (93.2%) of network nodes representing the scholars: 1) Dark nodes in the giant component. The Relevance Tuner network has 10 represent authors in the conference and dark edges are their (a) Social exposure with Scatter Viz (b) Social exposure with Relevance Tuner Figure 4: Conference Co-Authorship (pre-existing network) and Click Networks (new social exposure) components and 94 nodes in total out of which 71 (75.5%) in that the click pattern is visible on both strong and weak ties. the giant component. The Relevance Tuner interface enables the user to re-rank social recommendations according to five recommendation DISCUSSION features. The users can diversify the recommendation expo- The field study results hint that users might be prone to favor sure through re-tuning the sliders. A table style presentation the Relevance Tuner, an interactive extension of the ranked decreased the interface learning costs as well as the clicks list, that is, the relevance-first approach for implementing a so- on the recommendations. These findings help to provide de- cial recommendation system. At the same time, the observed sign directions for diversity-exposing user interface in social social diversity exposure pattern in Scatter Viz seems to better recommender systems. meet our design objective to support social diversity exposure and weak tie formation. This contradiction might suggest that CONCLUSION the perceived usability and familiarity of the user interface In this paper, we discussed and explored the ways social rec- weigh more in the overall estimation of the system quality. ommender systems, in particular their user interfaces, impact Operationalizing the social effect as questionnaire statements social diversity exposure, and therefore the accumulation of about the interface is challenging particularly in this context: social capital. Although the small size of the population that the participants seem to have focused on the quality of the took part in the experiment, we argue that focusing on the for- interface as a decision-support system, while the notion of so- mation of weak ties is a recommendation strategy that should cial effect would call for objective measurements like the click be explored further. For conference newcomers, the lack of analysis in Figures 4a and 4b. However, drawing conclusions an existing network implies that new connections are weak requires further research and new ways of studying both the ties by default. For senior scholars, weak ties provide an op- subjective perceptions and objective measurements about the portunity to escape their existing dense web of connections. long-term effects of providing social recommendations. We point to extant literature for evidence on the importance of weak tie-based bonding capital in knowledge work. We make The click-based new connections seem to form in bridging po- two main contributions. First, we draw from extant theory sitions and draw in peripheral scholars, including newcomers on social capital and diversity exposure in recommendation and attendees without a paper at the conference. However, it systems to suggest design directions for social diversity expo- is noteworthy that the two interface solutions are based on dif- sure in social recommendation systems. Second, we run an ferent types of information architecture and interaction flows, in-the-wild online field study in an academic conference to which means that different numbers of clicks are required for reflect on our theoretical discussion and to guide the design the same actions. The Scatter Viz interface allows the user to of controlled user experiments and the future user interface explore the social recommendation in two dimensions. This design of social systems. interface design required the user to click more for inspecting social recommendations. The network visualization showed REFERENCES [15] Ido Guy. 2015. Social Recommender Systems. In [1] Saeed Amal, Chun-Hua Tsai, Peter Brusilovsky, Tsvi Recommender Systems Handbook. Springer US, Kuflik, and Einat Minkov. 2019. Relational social 511–543. recommendation: Application to the academic domain. [16] Natali Helberger, Kari Karppinen, and Lucia D’Acunto. Expert Systems with Applications 124 (2019), 182–195. 2018. Exposure diversity as a design principle for [2] Sinan Aral. 2016. The Future of Weak Ties. Amer. J. recommender systems. Information, Communication & Sociology 121, 6 (2016), 1931–1939. Society 21, 2 (2018), 191–207. [3] Margaret S. Archer. 1995. Realist social theory: the [17] Jukka Huhtamäki, Martha G. Russell, Neil Rubens, and morphogenetic approach. Cambridge University Press, Kaisa Still. 2015. Ostinato: The Cambridge. 354 pages. Exploration-Automation Cycle of User-Centric, [4] Martin Atzmüller, Dominik Benz, Stephan Doerfel, Process-Automated Data-Driven Visual Network Andreas Hotho, Robert Jäschke, Bjoern Elmar Macek, Analytics. In Transparency in Social Media: Tools, Folke Mitzlaff, Christoph Scholz, and Gerd Stumme. Methods and Algorithms for Mediating Online 2011. Enhancing Social Interactions at Conferences. it - Interactions. Springer International Publishing Information Technology 53, 3 (2011), 101–107. Switzerland, 197–222. [5] Pierre Bourdieu. 1986. The Forms of Capital. In [18] S. Konomi, S. Inoue, T. Kobayashi, M. Tsuchida, and M. Handbook of Theory and Research for the Sociology of Kitsuregawa. 2006. Supporting Colocated Interactions Education, John G. Richardson (Ed.). Greenwood Press, Using RFID and Social Network Displays. IEEE New York, New York, USA, 241–258. Pervasive Computing 5, 3 (2006), 48–56. [6] Peter Brusilovsky, Jung Sun Oh, Claudia López, Denis [19] Gueorgi Kossinets and Duncan J. Watts. 2009. Origins Parra, and Wei Jeng. 2017. Linking information and of Homophily in an Evolving Social Network. Amer. J. people in a social system for academic conferences. New Sociology 115, 2 (2009), 405–450. Review of Hypermedia and Multimedia 23, 2 (2017), http://doi.org/10.1086/599247 81–111. [20] David W. McDonald, Joseph F. McCarthy, Suzanne [7] Moira Burke, Robert Kraut, and Cameron Marlow. 2011. Soroczak, David H. Nguyen, and Al M. Rashid. 2008. Social Capital on Facebook: Differentiating Uses and Proactive displays: Supporting awareness in fluid social Users. In Proceedings of the 2011 annual conference on environments. ACM Transactions on Computer-Human Human factors in computing systems - CHI ’11. ACM Interaction 14, 4 (2008), 1–31. Press, New York, New York, USA, 571. [21] Miller McPherson, Lynn Smith-Lovin, and James M. [8] Ronald S. Burt. 2004. Structural Holes and Good Ideas. Cook. 2001. Birds of a Feather: Homophily in Social Amer. J. Sociology 110, 2 (2004), 349–399. Networks. Annual Review of Sociology 27, 1 (2001), 415–444. [9] Alvin Chin, Bin Xu, Fangxi Yin, Xia Wang, Wei Wang, http://dx.doi.org/10.1146/annurev.soc.27.1.415 Xiaoguang Fan, Dezhi Hong, and Ying Wang. 2012. Using Proximity and Homophily to Connect Conference [22] Janine Nahapiet and Sumantra Ghoshal. 1998. Social Attendees in a Mobile Social Network. In 2012 32nd Capital, Intellectual Capital, and the Organizational International Conference on Distributed Computing Advantage. The Academy of Management Review 23, 2 Systems Workshops. IEEE, 79–87. (1998), 242–266. [10] Elizabeth Churchill, Andreas Girgensohn, Les Nelson, [23] Ekaterina Olshannikova, Thomas Olsson, Jukka and Alison Lee. 2004. Blending digital and physical Huhtamäki, Susanna Paasovaara, and Hannu spaces for ubiquitous community participation. Kärkkäinen. 2020. From Chance to Serendipity: Commun. ACM 47, 2 (2004), 38. Knowledge Workers’ Experiences of Serendipitous Social Encounters. Advances in Human-Computer [11] James S. Coleman. 1988. Social Capital in the Creation Interaction 2020 (2020), 18. of Human Capital. Amer. J. Sociology 94 (1988), S95–S120. http://www.jstor.org/stable/2780243 [24] Thomas Olsson, Jukka Huhtamäki, and Hannu Kärkkäinen. 2020. Directions for Professional Social [12] Philip J. Dobson. 2001. The Philosophy of Critical Matching Systems. Commun. ACM 63, 2 (2020), 60–69. Realism—An Opportunity for Information Systems Research. Information Systems Frontiers 3, 2 (2001), [25] Anshul Vikram Pandey, Josua Krause, Cristian Felix, 199–210. Jeremy Boy, and Enrico Bertini. 2016. Towards Understanding Human Similarity Perception in the [13] Linton C Freeman. 2000. Visualizing Social Networks. Analysis of Large Sets of Scatter Plots. In Proceedings Journal of Social Structure 1, 1 (2000), [np]. http://www. of the 2016 CHI Conference on Human Factors in cmu.edu/joss/content/articles/volume1/Freeman.html Computing Systems - CHI ’16. ACM Press, New York, [14] Mark Granovetter. 1973. The Strength of Weak Ties. New York, USA, 3659–3669. American journal of sociology 78, 6 (1973), 1360–1380. [26] Manh Cuong Pham, D. Kovachev, Yiwei Cao, G. M. of academic social networks. In Proceeding of the 14th Mbogos, and R. Klamma. 2012. Enhancing Academic ACM SIGKDD international conference on Knowledge Event Participation with Context-aware and Social discovery and data mining - KDD 08. ACM Press, New Recommendations. In 2012 IEEE/ACM International York, New York, USA, 990. Conference on Advances in Social Networks Analysis [33] Chun-Hua Tsai and Peter Brusilovsky. 2016. A and Mining. IEEE, 464–471. personalized people recommender system using global [27] Robert D. Putnam. 2000. Bowling Alone: The Collapse search approach. In Proceedigns of the IConference and Revival of American Community. Simon & Schuster. 2016, March 20-23, 2016, Philadelphia, Pennsylvania, 541 pages. USA. 5. http://d-scholarship.pitt.edu/28984/ [28] Simon Rodan and Charles Galunic. 2004. More than [34] Chun-Hua Tsai and Peter Brusilovsky. 2018. Beyond the network structure: how knowledge heterogeneity Ranked List: User-Driven Exploration and influences managerial performance and innovativeness. Diversification of Social Recommendation. In Strategic Management Journal 25, 6 (2004), 541–562. Proceedings of the 2018 Conference on Human [29] Kaisa Still, Jukka Huhtamäki, and Martha G. Russell. Information Interaction&Retrieval - IUI ’18. ACM 2013. Relational Capital and Social Capital: One or two Press, New York, New York, USA, 239–250. Fields of Research?. In Proceedings of the 10th [35] Chun-Hua Tsai and Peter Brusilovsky. 2019a. International Conference on Intellectual Capital, Explaining recommendations in an interactive hybrid Knowledge Management and Organisational Learning, social recommender. In Proceedings of the 24th The George Washington University, Washington, DC, International Conference on Intelligent User Interfaces. USA, 24-25 October 2013. 420–428. 391–396. [30] Yasuyuki Sumi and Kenji Mase. 2002. Conference assistant system for supporting knowledge sharing in [36] Chun-Hua Tsai and Peter Brusilovsky. 2019b. Exploring academic communities. Interacting with Computers 14, Social Recommendations with Visual 6 (2002), 713–737. Diversity-Promoting Interfaces. ACM Transactions on Interactive Intelligent Systems (TiiS) 10, 1 (2019), 1–34. [31] Cass R. Sunstein. 2007. Republic.com 2.0. Princeton University Press. 251 pages. [37] J. Yang and J. Leskovec. 2014. Overlapping Communities Explain Core - Periphery Organization of [32] Jie Tang, Jing Zhang, Limin Yao, Juanzi Li, Li Zhang, Networks. Proc. IEEE 102, 12 (2014), 1892–1902. and Zhong Su. 2008. ArnetMiner: extraction and mining