نقش شبکه فردی (Ego networks) در شبکه اجتماعی آنلاین و نشر اطلاعات

ABSTRACT Online Social Networks and information diffusion: The role of ego networks

Online Social Networks (OSNs) such as Facebook and Twitter have introduced radically novel means of interactions among people, which quickly became extremely popular. Complementing more traditional ‘offline’ means of communication (such as faceto-face communication and phone calls), ‘Online’ Social Networks are creating a complete virtual social environment, which supports many actions involving social interaction, from extremely simple ones such as “liking” other users’ content, up to very complex ones such as looking for a job, advertising products and organizing events. This is one of the most impressive cases of cyber-physical convergence, i.e., the process whereby the physical world around us and the virtual world of the Internet are deeply interwoven, constantly interacting with, and dependent upon, each other [1]. In this context, the analysis of the human social behavior in online environments is particularly exciting, as it allows us, on the one hand, to understand features of the human behavior based on huge amounts of data and, on the other hand, to design services and applications exploiting this knowledge. This paper presents a comprehensive overview of a body of work that we have carried out in this field, with the objective of highlighting key structural properties of human personal networks in OSNs, how they are related to known structures of offline human social networks, and how they impact on the patterns of information diffusion that have been observed in these environments.

Undoubtedly, the advent of OSNs also led us to a significant advancement in the study of human online behavior. In fact, big data coming from OSNs represent an invaluable source of information to describe the dynamics of complex social phenomena (e.g., the diffusion of information in the network, the formation of social relationships and communities), which are very difficult to analyze with traditional research methods typically based on surveys and interviews. A lot of effort has been put in the last years to characterize OSNs by studying their graphs, since this is the natural way to study structural properties of human social relationships in OSNs. Most of the literature has focused on macroscopic properties of OSN graphs, i.e., the structural characteristics of the global network formed by all users and their connections. A comparatively less explored but equally important subject of investigation are the microscopic properties of OSNs, and primarily the structural characteristics of our personal social networks, also called ego networks. In the anthropology literature it is well known (e.g., [2]) that the characteristics of ego networks are fundamental to determine key facets of human behavior, such as trust, sharing of resources, and formation of communities.

From reference works in psychology and anthropology (e.g., [3–6]), we know that the properties of offline ego networks are constrained by a series of cognitive and time limits, which bound the amount of relationships that each individual can actively maintain due to their intrinsic cost in terms of ‘computational resources’ for the brain. Specifically, cognitive constraints limit the total number of active relationships humans can maintain at a non-negligible level of intimacy. This limit is on average around 150 relationships, which is known as the Dunbar’s number [3]. The same constraints also dictate specific structures according to which social relationships are organized inside the ego network, as explained in detail in Section 2.

Recent analyses of the structural properties of popular OSNs (i.e., Facebook and Twitter) revealed that online ego networks have the same properties of offline ego networks, with similar size and the same hierarchical structure [7–9]. This confirms that egonetwork properties depend primarily on cognitive constraints of the human brain, and are not influenced by the use of specific communication mechanisms, such as mobile phones [10] and also Online Social Networks. In this regard, Facebook and Twitter do not seem to improve human social capacity, but they simply represent additional social channel we can use. Moreover, in addition to confirming that well-known features of human ego networks also manifest in OSNs, these studies have revealed additional properties [7], which had been hypothesized [11], but never observed due to lack of big data sources. This demonstrates that OSNs can be used also as a ‘social microscope’ to investigate novel key properties of our social behavior.

Recent results presented, e.g., in [12–14] show that the patterns of information diffusion we observe in OSNs strongly depend on the structure of the users’ ego networks. Therefore, by understanding the latter it is also possible to design OSN services where the features of ego network structures (and the users’ behavior they determine) are exploited to optimize data management. For example, as demonstrated by Lerman [15], models of information dissemination in OSNs that consider the constrained nature of human online social behavior overcome some intrinsic limitations of previous state-of-the-art models [16]. Moreover, results presented in [12] have shown that, under the assumption that the ego network structure also defines the level of trust between ego and alters, the length of the shortest path along which trusted information flows between two users in the network can be significantly longer than the few hops (compatible with the well-known 6-degrees of separation concept) previously highlighted in the literature (e.g., in [17]).

In this paper, we first present the main results we have obtained from the analysis of the effect of human cognitive limits on the structural properties of ego networks in OSNs. Then, we present results showing the impact of these ego network structures on macroscopic phenomena, with particular attention to the diffusion of information. In addition, we discuss promising directions to the design of information-centric systems exploiting ego network structures, such as data replication strategies based on ego-network concepts for Distributed Online Social Networks (DOSN), and information dissemination protocols for opportunistic networks based on ego-network cognitive heuristics.

This paper starts from the analysis of human cognitive limits in online environments presented in the book by Arnaboldi et al. [18]. However, while the book focuses on the structural analysis of ego networks in OSNs, this paper focuses on the application of models and analyses based on human cognitive limits to user-centric services and applications, in particular to the diffusion of information.

The paper is organized as follows. In Section 2, we provide the main definitions about Online Social Networks and ego networks, and we present an overview of the most important results of ego network analysis in OSNs. In Section 3, we present the background work in the field of information diffusion modeling, we introduce the most important information diffusion models in OSNs and we discuss their limitations. Then, in Section 4, we present analyses and models of information diffusion in OSNs based on the structural properties of ego networks. In Section 5, we present some research directions where knowledge about social network structures is used to design data-centric services both for OSNs and mobile networks, and we draw the main conclusions of the paper.