Toward Context-Aware Mobile Social Networks

ACCEPTED FROM OPEN CALL Toward Context-Aware Mobile Social Networks Zhiyong Yu, Daqing Zhang, Zhu Wang, Bin Guo, Ioanna Roussaki, Kevin Doolin, and Ethel Claffey The authors classify CA-MSNs into four categories, and divide their life cycle into four phases: discovery, connection, interaction, and organization. They then introduce personal and community context, and discuss the corresponding taxonomy. They also discuss how such context can be leveraged to enhance each life cycle phase. ABSTRACT CA-MSNs are more intelligent and user-friendly than conventional online or mobile social networks. We first classify CA-MSNs into four categories, and divide their life cycle into four phases: discovery, connection, interaction, and organization. We then introduce personal and community context, and discuss the corresponding taxonomy. Subsequently, we elaborate how such context can be leveraged to enhance each life cycle phase. We also present our practices on designing various CA-MSN applications. Finally, future research directions are identified to shed light on the next generation MSNs from the context awareness perspective. INTRODUCTION Mobile social networks (MSNs) are becoming killer applications that can show the power of combining mobile computing with social networking [1]. MSNs are not only an elementary extension of existing online social networks (i.e., conventional MSNs), but also revolutionizing social networks by bringing anywhere anytime social interaction with higher-level intelligence. The former is reached by smart mobile phones’ inherent property via wireless communication, whereas the latter is enabled through utilizing the comprehensive users’ context acquired or inferred from fertile data sources such as web services, social networking sites, wearable/mobile devices, and environmental wired/wireless sensor networks [2, 3]. We name the latter context-aware mobile social networks (CA-MSNs). Although online social networking services have been very successful in attracting billions of users to socialize in cyber space in a short time, none of them tap into the vast amount of context affiliated with users who shuttle constantly across the physical and virtual worlds with feature-rich smartphones. Nowadays there is an unparalleled chance to comprehensively understand the context surrounding individuals or communities in almost any scene [2]. Motivated by this observation, we aim to exploit new facets of context that are vital to MSNs, and examine how context awareness will shape the future MSN paradigm. More specifically, this article: • Characterizes CA-MSNs by comparing them to conventional MSNs and provides a taxonomy for CA-MSNs with corresponding application scenarios Digital Object Identifier: 10.1109/MCOM.2017.1700037 168 • Proposes a methodology of creating CA-MSN applications by investigating the usage of personal and community context in their lifecycle phases • Designs three CA-MSN applications with guidance from the methodology, which shows the power of context awareness through our practices and outputs some visions of future MSNs CHARACTERISTICS AND TAXONOMY OF CA-MSNS An increasing number of people are socializing and grouping in cyber space by using online social networks regularly. With the quick penetration of sensor-equipped smartphones, social networking services (e.g., Twitter and Facebook) tend to create mobile phone applications, which can provide online users with “here and now” access from their smartphones. In turn, native MSNs (e.g., Foursquare) have been developed to construct communities for real-world mobile users. The line between social network services on the web and mobile applications is being blurred. As a result, two trends are joined: online social networks are extended for mobile access and localization through mobile phone browsers or applications, and native MSNs utilize user profiles, activities, and contents generated via online social networks. This way, MSNs can be maintained remotely and virtually just like traditional online social networks, and can also be leveraged to support face-to-face and spontaneous interaction. However, these two trends have different genes. Compared to conventional MSNs, CA-MSNs use rich and high-semantic-level context to support either long- or short-term/range communities, as summarized in Table 1. Considering spontaneous MSNs at one extreme and online social networking services at the other, we can classify MSNs into four categories in accordance with their temporal and spatial features. Short-Term Short-Range MSNs: They can be built on an ad hoc network that adopts wireless point-to-point communication protocols (e.g., WiFi Direct, Bluetooth), and may appear in a coffee shop, an airport departure lounge, or a moving bus. The goal is to boost face-to-face conversation or facilitate information sharing in the physical world (for both acquaintances and strangers). For instance, Meetup allows members Zhiyong Yu is with Fuzhou University; Daqing Zhang (corresponding author) is with Telecom SudParis; Zhu Wang and Bin Guo are with Northwestern Polytechnical University; Ioanna Roussaki is with National Technical University of Athens; Kevin Doolin and Ethel Claffey are with Waterford Institute of Technology. 0163-6804/17/$25.00 © 2017 IEEE IEEE Communications Magazine • October 2017 Conventional MSNs CA-MSNs Facebook, Twitter, Google+ Foursquare, Meetup, Instagram, WeChat Category Mainly long-term long-range, based on Internet All possible (detailed below), based on hybrid networks (Internet + opportunistic networks) recognized as a major Context richness No context, or only few and low-level context, such as time and location Rich and high-level context (detailed in “Personal Context and Community Context”) pervasive computing Manual Automatic or aided with context (detailed in “Life Cycle Management in CA-MSNs”) objective was to confer Stateless, data stored mainly on server side Stateful, historical context stored at both client and server sides pervasive services and Examples Life cycle management Context storage Table 1. Comparisons of conventional and context-aware MSNs. Context-aware computing has been research branch of since the late 1990s. Its more intelligence on systems by considering the relevant context that was not yet taken community context, including context representation, mining, and inference [8]. More noteworthy, new technologies such as community modeling, participatory sensing, and large-scale multi-modal data fusion are promising to fully empower context-aware MSNs. into account. Afterward, PERSONAL CONTEXT TAXONOMY attention of research communities. CONTEXT-AWARE COMPUTING IN MSNS Personal context describes all relevant information of a person that can characterize his/her situation. It can be classified into static personal context and dynamic personal context. Static personal context refers to an individual profile that remains almost unchanged. It includes one’s identity and affiliation, which is quite stable, and one’s preference/interest, available resources, and contact list, which might change slowly. In MSNs, what we care about most is one’s preferences concerning social activities, for example, likes some movies and dislikes a certain restaurant. Dynamic personal context refers to contextual information that changes from time to time, such as one’s location, physiological condition (e.g., blood pressure and heart rate), behavior (e.g., walking and laughing), activity (e.g., sleeping, meeting, in a certain mood), and intent (e.g., temporary goal for a task). Context-aware computing has been recognized as a major research branch of pervasive computing since the late 1990s [5]. Its objective was to confer more intelligence on the pervasive services and systems by considering the relevant context that was not yet taken into account. Afterward, with the development of social computing, the context from social aspects drew the attention of research communities [6, 7]. In MSNs, context refers to the information concerning not only individuals, but also multiple users and entire groups. Both personal context and community context are indispensable to intelligent decision making in each phase of the MSN life cycle. For example, if one intends to create a community by discovering nearby people with certain hobbies, it would be essential to be aware of their personal context such as interest and location. On the other hand, if one wishes to discover existing communities to join, community context such as community location and profile might be needed. A spectrum of existing technologies has been developed for the extraction of personal and Community context is able to help communities to function efficiently by exploiting and understanding the activities, similarities, and relationships of the entire community as a whole. Community context can also be static or dynamic. Static community context consists of information about community profile and community structure. More specifically, the profile of a community includes motivation, membership, demography, resources, and preferences. The structure of a community comprises relationship (i.e., inter-personal and inter-community relationships), social status, and structural metrics obtained based on social network analysis (i.e., connection, distribution, density, and segmentation). In case members have different personal preferences, there should be a method to determine the community preference, which may be not a simple average of each member’s preference. Dynamic community context refers to the time varying contextual information of a communi- to create or join offline group meetings by a common temporary interest, such as books, games, movies, or pets. Short-Term Long-Range MSNs: They usually aim to facilitate remote teamwork via the Internet to complete a large task before a given deadline, for example, voluntary support for disaster relief, like a crowdsourcing disaster support platform (CDSP) [4], which is detailed in “Our Practices on CA-MSNs.” Long-Term Long-Range MSNs: Users of social networking services extended with mobile accessibility (e.g., Facebook) form this category of MSNs. The goal is to facilitate instant messaging and information dissemination globally. Long-Term Short-Range MSNs: They are confined to a group of people living/working together in limited physical spaces. The goal of these MSNs is to maintain relationships with familiar persons (e.g., in a family/company), with special security and privacy policies. For instance, WeChat has a function that allows users to join a private group with friends nearby. PERSONAL CONTEXT AND COMMUNITY CONTEXT IEEE Communications Magazine • October 2017 with the development of social computing, the context from social aspects drew the COMMUNITY CONTEXT TAXONOMY 169 SOCKER serves to Distinctive personal context Distinctive community context Reference create short-/long-term Foursquare Personal location Social status foursquare.com short-range MSNs. The Meetup Personal preferences Membership, community preferences www.meetup.com Instagram Personal activity Relationship, interaction instagram.com WeChat Personal identity, personal activity Community location, community motivation, interaction www.wechat.com SOCKER Trajectory, personal preference Encounter, user popularity, inter-user closeness, user effectiveness, community intent, community size [9] CDSP Expertise, available time, home location Acquaintanceship, physical proximity, interest consistency, interaction, social status, community intent [4] TLI Social relationship, home location, personal preference Overlapping influence, skill coverage, community intent, community size, activity location [10] three metrics, that is, user popularity, inter-user closeness, and user effectiveness, are used primarily for personal or community context. All can be estimated from users’ historical trajectories and interactions. Table 2. Context supported in CA-MSN products and prototypes. LIFE CYCLE MANAGEMENT IN CA-MSNS LIFE CYCLE OF MSNS MSNs involve the management of communities (i.e., a group of people communicating and interacting in a physical and/or virtual space for a common purpose [11]) and supporting resources (e.g., devices, networks, services). Inspired by the community management phases proposed in the EU FP7 SOCIETIES project [12], we divide the life cycle of MSNs into four phases/steps: discover, connect, interact, and organize. Discover: discovering users, resources, services, devices, and networks for creating new communities, or discovering already existing communities for joining, merging, and splitting Connect: connecting users to support interactions, connecting communities, or connecting members/communities to their owned devices, Location Discovery (what) Connection (who) Interaction (why) Organization (how) Profile Preference Activity Intent ••• ••• Right person Right place Suitable device Suitable service Similar interest Common goal ••• ••• networks, resources and services Interact: direct interacting via instant messaging, group chatting, and so on; indirect interacting via social media (tagging the same photos, commenting on the same videos, visiting the same places) Organize: adding users to or removing members from communities; creating, merging, splitting, and terminating communities; managing community hierarchies, coordinating interactions among members, maintaining infrastructures of a community CONTEXT-AWARE DISCOVERY, CONNECTION, INTERACTION, AND ORGANIZATION Based on the concept of context and life cycle, the methodology can be described as: in each life cycle phase, diverse context should be exploited to make MSNs more intelligent, as shown in Fig. 1. Context-Aware Discovery: In order to create a new community or identify an existing community to join, the first step is to discover the related people and resources crossing the boundary between the physical and virtual space. We note that current systems allow to some extent the discovery of people and devices in the physical environment via the Internet of Things (IoT, e.g., RFID tags) or Situation event Same location Right time Similar preference Common idea/goal Common background Community ty, such as community location (e.g., proximity), intra/inter-community interaction, community activity (i.e., the abstraction of a series of interactions among community members), and community intent (i.e., the short-term common goal based on each member’s requirements). In Table 2 we report the context features supported in existing MSN products and research prototypes. Relationship Community dynamics Right parties Updated community ••• ••• ••• ••• ••• ••• Input: context Output Figure 1. Context-aware discovery, connection, interaction, and organization. 170 IEEE Communications Magazine • October 2017 in cyberspace via searching on the Internet. However, these systems do not thoroughly exploit the variety of context for socializing purposes. As a an MSN aims to pull people and resources together, the most important context should include personal/community location, preference, intent, activity, and so on. Context-Aware Connection: People can possibly choose a communication channel from a wide range of methods (e.g., Internet or ad hoc network, text, or video). Personal and community context can ease a series of issues like connection establishment and switching, and can help to choose the most efficient connection from a quality and cost perspective. For example, a group chat is taking place locally, and the members are sharing a video via a mobile ad hoc network; then a remote friend wants to join the chat, and the ad hoc network can connect with the Internet automatically to enable the remote friend to receive the video. Context-Aware Interaction: Both the community context and members’ personal context play a vital role in enabling humanistic social interaction. A major challenge here is to identify the events and situations that should trigger the interactions. Such events and situations might vary greatly. A simple event may involve two members being available for a chat, while a complex situation could be friends negotiating a local tour based on their interests and free time. In general, the relationship and commonality between interaction parties should be monitored. Context-Aware Organization: This task includes introducing new members to an established community through additional discovery/ connection steps, and removing members who are not relevant to the community anymore. A major challenge here is to also detect events and situations that would trigger community evolution, such as joining, leaving, splitting, merging, and so on. Community dynamics such as change of location, interaction, activity, and intent can be exploited to support adaptive membership management. OUR PRACTICES ON CA-MSNS SOCKER: SOCIALLY AWARE BROKER-BASED COMMUNITY CREATION MECHANISM Various methods can rally people for a local activity, for example, posting a public announcement saying that there will be a weekend party at a nearby bar. However, some extra concerns include: • The number of attendees should be controlled precisely, but not in a competitive way. • The information of the activity, its attendees, and the sifting process should be kept private. • The initiator of the activity has certain social expectations, for example, to make new friends or to entertain old friends. Overlooking these concerns would disappoint rejected individuals. Our work, SOCKER [9], a socially aware broker-based community creation mechanism, aims to get together like-minded persons for a particular face-to-face social activity, with consideration of the above concerns. IEEE Communications Magazine • October 2017 Initiator as the first broker Initiator Y: fail Y: succeed Enough invitees? N Deadline expired? Encounter N N Present broker Y N Better broker? Inviting Y New broker Broker handover Figure 2. Community creation procedure of SOCKER. SOCKER regards the community creation problem as a broker-based information dissemination task. Three metrics (detailed below) are measured to judge whether a person is a proper broker. Figure 2 illustrates the procedure of SOCKER. Concretely, the initiator serves as the first broker when he plans to create a new community. As the broker moves in the physical world, he/she will encounter other persons opportunistically. For each person u i she/he meets, the broker decides whether to invite ui to join the community (according to social expectations of the initiator and matchmaking between the activity type and ui’s preference) and updates progress records such as current invitees and met-but-uninvited user list. Meanwhile, SOCKER decides whether to hand over the broker role to u i. If the broker handover condition is satisfied according to the three predefined metrics, the current broker will send the progress records to the new broker, then stop acting as a broker. The new broker will continue the community creation task just as her/his predecessor was doing it. The task is completed successfully if the required community size is reached before the deadline. If so, the last broker will report on the task accomplishment to all the invitees and the activity initiator. Otherwise, when the deadline expires but the current community size is still smaller than needed, the last broker will notify the activity initiator saying that the community creation task has failed. SOCKER serves to create short-/long-term short-range MSNs. The three metrics, that is, user popularity, inter-user closeness, and user effectiveness, are used primarily for personal or community context. All can be estimated from users’ historical trajectories and interactions. For a specific user, the user popularity is defined as the number of different persons she/he will encounter in a forthcoming period (e.g., a week). Intuitively, a user with higher popularity tends to meet more people, and thereby community creation can be accelerated. For two users, their inter-user closeness is defined as the number of encounters 171 Request creation Request resolution Request notification Answer vote Request processor On-site On-site users users Request manager Relevant off-site user selection Request database Query interface (a) Instant message service User profile database Off-site user grouping User profile extractor Off-site Off-site users users Off-site user manager LOG (b) Web user interfaces Manual input Social networks work Historical usage records Figure 3. System overview of CDSP. in a forthcoming period. The inter-user closeness of two successive brokers should be higher than a threshold when the activity initiator wants to play with old friends, while vice versa to make new friends. User effectiveness considers the progress records obtained during the community growing process to avoid broker handover in this case: although the candidate new broker would encounter many users in the future, perhaps previous brokers have already encountered most of them. It is defined as the number of unrecorded users one will encounter in a forthcoming period. Our experience shows that more communities can be created successfully with lower costs when with the help of selected context. User popularity is helpful to meet enough users rapidly, inter-user closeness can increase the chance of encountering the right persons, and user effectiveness is able to minimize the risk of appointing inappropriate brokers. CROWDSOURCING DISASTER SUPPORT PLATFORM When a disaster (e.g., an earthquake) occurs, we need a support system to help rescue teams in saving lives, property, and the environment. On-site rescue teams may meet with many problems such as finding a passable road or recognizing a nameplate written in an unknown language. Crowdsourcing is a feasible way to subcontract these tasks/requests to a large number of off-site volunteers. Therefore, CDSP [4] was developed in the SOCIETIES project. With CDSP, volunteers from all over the world can share the burden of large tasks, take turns responding to requests immediately, and interact with each other to make the answers more credible. In order to avoid presenting a long request 172 list to bother off-site users (i.e., volunteers), we introduce the skill-matching mechanism. On one hand, performing a task may involve some kinds of expertise, which is specified by the request creator (e.g., an on-site user). For example, damage assessment from satellite images needs the skill of image analysis. On the other hand, CDSP can discover users’ expertise from various sources, such as social networks, historical usage records of the platform, and manual input. The matching mechanism makes sure that the tasks are assigned to those who can handle them, and irrelevant requests are screened. For a large task, users first divide it into sub-tasks, and then work on these sub-tasks in parallel, and finally merge and output answers. Furthermore, they can have conversations through an instant messaging interface. Figure 3 shows the system overview of CDSP. The “off-site user grouping” component retrieves a group of users whose expertise is demanded by the large task, then connects them with a short-term long-range MSN, whose goal is to facilitate information sharing or collaborative work to accomplish a disaster relief task. To manage the MSN life cycle, both personal and community context are useful. Besides expertise, other personal context considered includes available time and home location. For example, a plenary meeting needs all members to be available at the same time, while monitoring a webcam incessantly needs their available times should not be necessary. Some tasks are more favorable to users from a particular regional background (i.e., home location). For instance, nameplate recognition is easier for users from regions near the disaster site. The above is for the discovery and connection phase. For the interaction and organization phase, com- IEEE Communications Magazine • October 2017 Restaurant “Tasting night” call Lessons from this prac- LBSNs tice include: personal context is useful for specifying and testing the condition to join a community; members’ interaction can help to complete a large task more effectively and ce er pr efe re n ati on Us sh me loc l re cia So Ho lat ion le Se rve di tem s ca Ev en ts Team (event participants) Ev en t lo ca tio n ip efficiently; and we can rely on community context from multiple sources to boost the interaction. Marketing effect maximization Marketing effect quantification Figure 4. Team formation procedure of TLI. munity context is newly gathered or inferred from personal context. Acquaintanceship (e.g., linkage on social networks), physical proximity (e.g., from the same region), and interest consistency (e.g., with many overlapped skills) can be leveraged to warm up and boost the interaction among community members. Typical interactions include brainstorming and voting, which can improve the answer’s credibility. When a task is finished, the community intent does not exist, and all members go back to the pool of volunteers, waiting/looking for the next task. Lessons from this practice include: personal context is useful for specifying and testing the condition to join a community; members’ interaction can help to complete a large task more effectively and efficiently; and we can rely on community context from multiple sources to boost the interaction. TLI: TEAM FORMATION WITH LOCAL INFLUENCE FOR OFFLINE EVENT MARKETING Offline events are favorable ways for business owners to transform current customers into brand advocates and reach potential customers. For example, a restaurant hosts a tasting night, invites a certain number of participants, and hopes that after the event these participants can attract more users (by their social influence) to visit this restaurant. Who should be invited? This is a problem of influence maximization with limited budget. However, several factors should be considered in this specific application. Unlike existing works that simply count how many users a single celebrity influences, we depict influence by estimating pi, which denotes the possibility of a user to visit a venue (viz. where the business is located, e.g., a restaurant). First, pi depends on how many influencers (i.e., event participants) have exerted influence on the user. A team’s influence on a user is not straightforwardly the sum of individual members’ influence, but follows the law of diminishing IEEE Communications Magazine • October 2017 marginal utility. We name it the overlapping factor. Second, pi depends on the distance between the user and the venue, the so called distance factor. Third, pi depends on whether the influencer and the user like the brand, specifically, the products or services offered by the business, named the coverage factor. In order to tackle these factors, we propose TLI [10], a team formation approach with local influence. We first build a marketing effect quantitative model with considerations of the factors of overlapping, distance, and coverage. Then a combinatorial optimization problem is formulized for the marketing effect maximization, which is approximately solved with a heuristic algorithm. The participant team can be determined accordingly. Figure 4 shows the team formation procedure of TLI. Apparently, team formation is a special type of community creation. The team members should meet certain constraints (e.g., skill coverage) while maximizing another metric (e.g., working achievement). This makes a team not as flexible or dynamic as an ordinary community, that is, members cannot join or leave a team at any time. In order to form a (near) optimal team, several kinds of personal and community context are collected and analyzed comprehensively, including social relationship, home location, and user preference, from location-based social networks. Users’ social relationships are used to eliminate overlapping social influence. Users’ home locations are used for discovering the relationship between the visiting probabilities and the distances. Since previous works prove that it follows the power law, we train the parameters of a power function from check-in records. Users’ preferences are extracted from visiting histories and joined together to ensure that the participant team covers all served products. Based on the above context awareness, we build an accurate and fine-grained influence model for a team. 173 Until now, MSNs have revolutionized the style people interact and communicate. The next generation of MSNs is expected to not only facilitate interaction and communication among people with better effectiveness, but also match the demand and the supply (in terms of information, services and goods) among people in a more intelligent manner. 174 We learned that when simple context cannot meet the requirements of an application, we need to design even richer and higher-semantic-level context, which can be inferred from low-level context with the help of technologies including community modeling, context mining, and so on. CONCLUSION AND FUTURE DIRECTIONS Until now, MSNs have revolutionized the style in which people interact and communicate. The next generation of MSNs is expected to not only facilitate interaction and communication among people with better effectiveness, but also match the demand and supply (in terms of information, services, and goods) among people in a more intelligent manner [13]. This is particularly pertinent to the many businesses that desire a strong online social presence, but remain challenged by this concept [14]. The CA-MSN is a promising evolution direction of MSNs; they could be enhanced from the following perspectives. Extending the Sensing Capability of MSNs with IoT and Mobile Crowd Sensing: By leveraging the sensors embedded in mobile devices, sensor networks installed in our surroundings, and the human digital footprints recorded by the IoT, a huge amount of context about users and their interactions in MSNs can be acquired. With these context data and corresponding big data technologies, more intelligent matchmaking and interaction among mobile users and resources can be supported. For example, by collecting bus/metro card records for a period, the system can learn members’ daily movement patterns, and then the community’s planning the location and time of a gathering can be more convenient. Mobile crowd sensing utilizes citizens’ off-the-shelf smartphones to capture social and urban dynamics [15]. By leveraging human power in the loop of the sensing and computing process, MSNs have the most favorable position to gain the advantages of the crowdsourced context. Extending the Communication Capability of MSNs by Bridging Mobile Ad Hoc Networks to Infrastructure-Based Networks Seamlessly: Nowadays, most existing online social networking services lack effective support for face-to-face interaction in the physical world, especially when/ where no infrastructure is available. This fact calls for research on the creation, organization, and migration of offline social networks to online social networks, and seamless transition between online and offline social networks. That is to say, future MSNs should be infrastructure-independent and capable of supporting both long-term relationships and spontaneous social interactions. For example, a user meeting others opportunistically in a coffee shop can create a local community via a mobile ad hoc network. When she leaves the coffee shop, this community can still be maintained online for further interactions. Extending the Service Platform of MSNs with Context Awareness Features: Generally speaking, MSNs can be seen as a service platform to ease information sharing, user interactions, service discovery/consumption, and users’ personal demand/satisfaction. To provide a convenient and effective service platform for mobile users, it is advisable to enhance key features of the platform with context awareness. Last but not least, the MSN should be an open platform that enables developers to freely create new context-aware applications for specific needs. ACKNOWLEDGMENT This work was partially supported by the National Natural Science Foundation of China (nos. 61772136, 61772428, 61402369, 61373119, 61672159), the Fujian Collaborative Innovation Center for Big Data Application in Governments, and the Technology Innovation Platform Project of Fujian Province (no. 2014H2005). REFERENCES [1] N. Vastardis and Y. Kun, “Mobile Social Networks: Architectures, Social Properties, and Key Research Challenges,” IEEE Commun. Surveys & Tutorials, vol. 15, no. 3, 2013, pp. 1355–71. [2] D. Zhang et al., “The Emergence of Social and Community Intelligence,” IEEE Computer, vol. 44, no. 7, 2010, pp. 21–28. [3] H. Chen et al., “A Generic Framework for Constraint-Driven Data Selection in Mobile Crowd Photographing,” IEEE Internet of Things J., vol. 4, no. 1, 2017, pp. 284–96. [4] D. Yang et al., “Providing Real-Time Assistance in Disaster Relief by Leveraging Crowdsourcing Power,” Personal and Ubiquitous Computing, vol. 18, no. 8, 2014, pp. 2025–34. [5] A. Dey et al., “A Conceptual Framework and a Toolkit for Supporting the Rapid Prototyping of Context-Aware Applications,” HumanComputer Interaction, vol. 16, no. 2, Dec. 2001, pp. 97–166. [6] Z. Yu et al., “Personalized Travel Package with Multi-Pointof-Interest Recommendation Based on Crowdsourced User Footprints,” IEEE Trans. Human-Machine Systems, vol. 46, no. 1, 2016, pp. 151–58. [7] A. Pentland, “Socially Aware, Computation and Communication,” IEEE Computer, vol. 38, no. 3, 2005, pp. 33–40. [8] X. Wang et al., “Ontology-Based Context Modeling and Reasoning Using OWL,” Proc. PerCom Wksp., Orlando, FL, Mar. 2004, pp. 18–22. [9] Z. Wang et al., “SOCKER: Enhancing Face-to-Face Social Interaction Based on Community Creation in Opportunistic Mobile Social Networks,” Wireless Personal Commun., vol. 78, no. 1, 2014, pp. 755–83. [10] Z. Yu et al., “Participant Selection for Offline Event Marketing Leveraging Location-Based Social Networks,” IEEE Trans. Systems, Man, Cybernetics: Systems, vo. 45, no. 6, 2015, pp. 853–64. [11] N. Lane, “Community-Aware Smartphone Sensing Systems,” IEEE Internet Computing, vol. 16, no. 3, 2012, pp. 60–64. [12] I. Roussaki et al., “Context Awareness in Wireless and Mobile Computing Revisited to Embrace Social Networking,” IEEE Commun. Mag., vol. 50, no. 6, June 2012, pp. 74–81. [13] A. Chin, “Ephemeral Social Networks,” Mobile Social Networking: An Innovative Approach, Springer, 2013, pp 25–64. [14] E. Claffey and M. Brady, “A Model of Consumer Engagement in a Virtual Customer Environment,” J. Customer Behaviour, vol. 13, no. 4, 2014, pp. 325–46. [15] B. Guo et al., “ActiveCrowd: A Framework for Optimized Multi-Task Allocation in Mobile Crowdsensing Systems,” IEEE Trans. Human-Machine Systems, vol. 47, no. 3, 2017, pp. 392–403. BIOGRAPHIES ZHIYONG YU ([email protected]) is an associate professor at the College of Mathematics and Computer Science, Fuzhou University, China, also affiliated with the Fujian Provincial Key Laboratory of Network Computing and Intelligent Information Processing and the Key Laboratory of Spatial Data Mining and Information Sharing, Ministry of Education, China. He received his Ph.D. from Northwestern Polytechnical University, China, in 2011. He was a visiting student at Kyoto University, Japan, from 2007 to 2009 and a visiting researcher at Telecom SudParis, France, from 2012 to 2013. His current research interests include pervasive computing, mobile social networks, and crowdsensing. DAQING ZHANG ([email protected]) is a professor with Telecom SudParis and SAMOVAR, CNRS, France. He obtained his Ph.D. from the University of Rome “La Sapienza,” Italy, in 1996. His research interests include context-aware computing, urban computing, mobile computing, and so on. He has served as the General or Program Chair for more than 10 inter- IEEE Communications Magazine • October 2017 national conferences. He is an Associate Editor of ACM Transactions on Intelligent Systems and Technology, IEEE Transactions on Big Data, and other publications. ZHU WANG ([email protected]) is an associate professor at the School of Computer Science, Northwestern Polytechnical University. He received his Ph.D. in computer science and technology from Northwestern Polytechnical University in 2013. from November 2010 to April 2012, he worked as a visiting student at Telecom SudParis. His research interests include pervasive computing, mobile social networking, and healthcare. BIN GUO ([email protected]) is a professor at the School of Computer Science, Northwestern Polytechnical University. He received his Ph.D. from Keio University, Tokyo, Japan, in 2009. During 2009–2011, he was a postdoctoral researcher at Telecom SudParis. His research interests include pervasive computing, social computing, and mobile crowdsensing. He has served as an Editor or Guest Editor for a number of international journals, such as IEEE THMS and ACM TIST. I OANNA R OUSSAKI ([email protected]) received her Diploma in electrical and computer engineering in 1997 from the National Technical University of Athens (NTUA), Greece. In 2003, she received her Ph.D. in the area of telecommunications IEEE Communications Magazine • October 2017 and computer networks. She has participated in many national and international research and development projects. Since 2015, she has been an assistant professor in the NTUA School of Electrical and Computer Engineering. Her research interests include the Internet of Things, context awareness, social computing, and so on. K EVIN D OOLIN ([email protected]) is director of Innovation at Waterford Institute of Technology’s Telecommunications Software and Systems Group (TSSG). His area of expertise is pervasive computing, which is the forerunner to the Internet of Things. He has coordinated a number of key EU projects in this space, including PERSIST (www.ict-persist.eu) and SOCIETIES (www.ict-societies.eu), which closed in April 2014 and received significant praise from expert scientific reviewers, and which focused on the integration of pervasive and social computing. ETHEL CLAFFEY ([email protected]) is a lecturer in marketing in theSchool of Business at Waterford Institute of Technology. Her Ph.D. was awarded by Trinity College Dublin. Her research interests include consumer engagement, contemporary consumer behavior, virtual communities, technology acceptance, and digital marketing. Her work has been published in a variety of conference proceedings and refereed journal articles such as Psychology & Marketing and the Journal of Customer Behaviour. 175