Elsevier

Social Networks

Volume 52, January 2018, Pages 282-293
Social Networks

Core-periphery or decentralized? Topological shifts of specialized information on Twitter

https://doi.org/10.1016/j.socnet.2017.09.006Get rights and content

Highlights

  • Twitter agriculture social web is modelled with the core-periphery profile approach.

  • Network centralization increases when Twitter users share specialized information.

  • Network shifts from centralized to decentralized as conversations turn more generic.

  • Results identify when Twitter is an information diffusion system or a social network.

  • The agriculture social web replicates the top-down model from government to growers.

Abstract

In this paper we investigate shifts in Twitter network topology resulting from the type of information being shared. We identified communities matching areas of agricultural expertise and measured the core-periphery centralization of network formations resulting from users sharing generic versus specialized information. We found that centralization increases when specialized information is shared and that the network adopts decentralized formations as conversations become more generic. The results are consistent with classical diffusion models positing that specialized information comes with greater centralization, but they also show that users favor decentralized formations, which can foster community cohesion, when spreading specialized information is secondary.

Introduction

In this paper we investigate how Twitter networks can shift from a centralized topology, characterized by a high core-periphery profile, to a decentralized topology characterized by low core-periphery estimates. Classical diffusion models (Rogers, 2010, Schon, 1971) posit that centralized networks are more efficient in spreading specialized information to specific communities of interest. On the other, recent studies have foregrounded the role of decentralized networks in disseminating behavior and facilitating the development of social norms that reinforce learning in local networks (Centola, 2010, Centola and Baronchelli, 2015). Centralized networks are particularly salient in sectors relying on a small number of specialists who engage a highly diverse and continuously expanding body of potential stakeholders, a diffusion system in which experts constitute the network core feeding information to the peripheral audience. Decentralized systems, on the other hand, facilitate the emergence of new ideas growing out of practical experience. Such systems lack a clear core or periphery as the information is more widely sourced and shared by all members of the network.

Twitter is an atypical social network in which the topological characteristics of both centralized and decentralized diffusion systems are present (Gabielkov et al., 2014, Kwak et al., 2010). The basic proposition of this study is that communities of interest assume different network formations that optimize the information diffusion from an active core to a relatively passive periphery; or inversely, allow the horizontal sharing of information that can be tailored to fit with users’ needs where individual decisions on which source to seek information from are relatively free, thus facilitating adaptation and implementation by local users. We explore this proposition by isolating subsets of generic and specialized tweets posted by several communities of users involved in agriculture and subsequently measuring the core-periphery profile of their multiple, comparable subgraphs. For the purposes of this study, we refer to subgraphs as a defined set of nodes and arcs of the original Twitter graph selected on the basis of specific characteristics of the message.

Agriculture and the more specialized field of sustainable agriculture are an important and useful setting in which to study the diffusion of specialized information. Modern agricultural systems are experiencing a revolution in how knowledge is disseminated and exchanged among networks of outreach professionals, farmers, consumers, and community stakeholders. The traditional approach to agricultural extension is highly centralized and relies on a top-down, continuum model going from university researchers to cooperative extension agents and finally to farmers (Rogers, 2010, Van den Ban and Hawkins, 1996). With internet penetration rates growing in rural communities (USDA, 2015), stakeholders are increasingly adopting social media and other online forms of communication to share agricultural information across local, national, and global networks. Notwithstanding these major developments, the impact of network technologies to the diffusion of specialized information remains relatively uncharted, with only a handful studies exploring the use of social media within the agriculture and food sectors (Chowdhury et al., 2013, Rhoades and Aue, 2010).

Although agricultural extension services in the United States are historically associated with centralization (Rogers, 2010), sustainable agriculture comprises a subset of agricultural extension that can benefit from decentralized diffusion systems, with stakeholders increasingly adopting digital strategies to complement more traditional outreach systems (Lubell et al., 2014). Agricultural extension and outreach remains rooted in specialized information about agricultural practices, economic conditions, and other relevant decision-making parameters. This specialized information must be applicable at the local level to individual farms and agricultural communities, but more general ideas need to be developed at the global level by upscaling multiple local experiences and then downscaling information to catalyze local learning. Thus, the diffusion of specialized information about sustainable agriculture requires a capacity to continuously facilitate a recursive flow of local and global information, a dynamic that can benefit from both centralized and decentralized diffusion systems (Valente and Rogers, 1995).

As a consequence of this duality in communicating specialized agricultural information, the different strategies surrounding agricultural extension and sustainable agriculture outreach offer an ideal case study to investigate the diffusion of specialized information on social media. Sustainable agriculture is a quintessential example of a community where knowledge networks must spread information across specialized sub-communities that are concerned with different aspects of the complex global food system (Klerkx et al., 2015, Klerkx and Proctor, 2013). The overall knowledge network not only has to deal with internal components of the system, for example understanding climate change and water management, but must also link the specialized system to the broader global culture represented by social media platforms like Twitter. Sustainable agriculture is not unique in this way-we expect similar dynamics may apply to other broad epistemic communities, e.g. social media users discussing “energy independence,” “national defense,” and other similarly specialized topics (Lubell et al., 2011, Lubell et al., 2014).

However, sustainable agriculture is a particularly useful domain in which to study the dynamics between network structure and knowledge specialization because there is an important tradition of knowledge extension among the education and outreach professionals involved with agriculture (Clark et al., 2016). The traditional approach to knowledge extension was to deliver research findings from universities to farmers and other interested stakeholders via personal communication and networks of local extension agents. With the advent of new information and communication technology (ICT) and social media, extension professionals involved with agricultural knowledge systems (Hermans et al., 2015) are increasingly experimenting with online forms of communication and continue to contend with general ideas such as network centralization and knowledge specialization that may apply to the specific topics of interest for agriculture.

In the following, we briefly review the literature on diffusion of innovations and detail an approach to core-periphery analysis that returns a continuous measurement of the centralization observed in the network. In the later sections of the paper we present the results of this study and discuss the more general policy implications of our findings.

Classical diffusion models posit that innovation originates from expert sources and then diffuses uniformly to potential adopters who either accept or reject the innovation. The source of information is situated at the center of the communication network and adoption is mostly a passive act of imitation of the source behavior. This classical model was successfully applied to agricultural extension services and the underlying model is derived from Ryan and Gross (1943) seminal study that tracked the diffusion of hybrid corn throughout the Midwest. The original study identified diffusion agencies, commercial channels, and neighbors as key actors that informed farmers of the new seed and affected their rate of adoption. Much agricultural diffusion in the United States emerged from this centralized model, in that key decisions about how to diffuse them, and to whom, were left to a small number of technical experts (Rogers, 2010).

Schon (1971) called into question this seminal model by exploring the reality of emerging diffusion systems and criticizing the classical diffusion theory, which he referred to as the “center-periphery model.” According to Schon (1971), the assumption that innovations originate from a centralized source and then diffuse to users fails to capture the complexity of decentralized diffusion systems in which innovations originate from numerous sources, are shared among individuals, and evolve as they diffuse via horizontal networks. In such decentralized systems, innovations pop up from users at the operational levels (as opposed to the core) and new ideas can spread horizontally via peer networks, with a high degree of re-invention occurring as innovations are modified by users to fit their conditions. The topology of decentralized systems shares a remarkable resemblance with social networks, which allow information diffusion to be widely shared by adopters who also serve as their own change agents (Centola, 2010, Gibbons, 2007).

Diffusion of innovation theories thus comprehend a spectrum from centralized, information diffusion systems to decentralized, horizontal networks. Rogers (2010) argued that centralized diffusion systems were defined by a top-down diffusion from governmental agencies and technical experts to local users and often displayed a low degree of local adaptation and sharing of innovation among adopters, whereas decentralized diffusion systems were characterized by peer diffusion through horizontal networks and a high degree of local adaptation and sharing among adopters. These models of diffusion of innovations were subsequently revised and applied to the diffusion of new communication technologies, with Valente (1996) presenting a threshold concept to provide a social network formulation to the diffusion of innovations and Rice (1987) arguing that computer networks facilitated the diffusion of information to organizations’ environments.

Based on this history, it is apparent that literature exploring the link between network structure and knowledge distribution is centered on the extent to which decentralized networks are more effective at distributing information, specialized or otherwise. The relationship between network structure and task performance was found to be dependent on the type of task performed within organizations (Ahuja and Carley, 1999, Cummings and Cross, 2003), with non-routine tasks performing better in less hierarchical networks compared with more routine or simpler tasks which benefit from hierarchy, in line with the postulates of classical diffusion of information theory. Transposed to our empirical study, we hypothesize that as the proportion of specialized information being shared increases, the more likely Twitter communities will be to display centralized network formations.

Diffusion of innovation theories have offered a fertile ground for the study of “influentials” and the spread of novel information on social media, with a range of studies exploring potential metrics to assess users’ influence and passivity based on their information-forwarding activities (Bakshy et al., 2011, Romero et al., 2011, Wu et al., 2011). These seminal theories also echoed the literature of social networks and the formal definition of small-world networks. Watts and Strogatz (1998) and Newman (2000) designed a mechanism to investigate interpersonal influences through high clustering coefficient and small path length. Such a network topology deviates from centralized networks that are mostly optimized for information diffusion from a clearly-defined core to a large periphery of nodes. Compared with decentralized, well-structured small-world networks, diffusion of innovations was found to be slow in regular networks and fast but sporadic in random networks (Delre et al., 2007).

Decentralized networks proved particularly useful in the propagation of rumor and the spread of diseases (Valente, 1995). Albrecht and Ropp (1984) found that workers were more likely to report talking about new ideas with colleagues with whom they also discussed personal matters, as opposed to following prescribed channels based on hierarchical role relationships. This body of scholarship led to developments such as targeted advertising directed at cohesive subgroups who were next in the line of innovation adoption. It also contributed to the theoretical debate by suggesting that social network characteristics which influence the diffusion of innovations include centrality, density, and particularly reciprocity; a feature markedly absent of Twitter social networks, in which network topology presents characteristics typical both of social networks and of highly centralized diffusion systems (Wu et al., 2011).

Social media literature has long debated whether Twitter is an information diffusion system, characterized by a skewed distribution of links and low rate of reciprocal ties (Bakshy et al., 2011, Wu et al., 2011), or a social network, structured around social relations, with a higher incidence of reciprocal ties and a distribution of outgoing links similar to that of incoming links (Newman and Park, 2003). The debate hinges on the overall network structure observed on Twitter and is relevant to organizations and users seeking to optimize the reach of their message in the social network. If an outreach organization uses Twitter primarily as an information diffusion system, then to be effective, it is imperative for the community to identify “influentials”—i.e., users that belong to a central core and perform the role of hubs relaying information to the periphery of the network. On the other hand, if it is being used primarily as a social network, then outreach strategies should involve the development of many local relationships and dense network structures with reciprocal ties and transitive triangles will be beneficial.

The definition of core-periphery is intuitive and comprehends the union of a dense core with a sparsely connected periphery. More nuanced approaches tend to contrast, at one extreme, one homogeneous group with a large set of undifferentiated actors, and at the other, a two-class partition of nodes with one class being the core and the other being the periphery (Boyd et al., 2006). The core-periphery structure was only relatively recently given a formal definition by Borgatti and Everett (2000) and the bulk of the scholarship remains rooted on economics research, social inequality, and power dynamics between elites and non-elites (Csermely et al., 2013, Holme, 2005, Rombach et al., 2014). Core-periphery analysis has also been applied to the structural patterns of Usenet groups (Choi and Danowski, 2002), dynamics of protest movements (Barberá et al., 2015), the spatial distribution of ties in social networks (Volkovich et al., 2012), and knowledge networks of wine producers (Giuliani, 2005). While these studies rely on blockmodeling and k-shell decomposition (Dorogovtsev et al., 2006, Žnidaršič et al., 2017), we rely on the core-periphery profile approach (Della Rossa et al., 2013) which returns a global network measure of “core-peripheriness”—i.e., an indicator of network centralization.

Drawing from this body of scholarship, we hypothesize that the observed networks will exhibit increasingly higher estimates of core-periphery structure as users relay more specialized information (i.e., agriculture-related information) compared with the baseline of generic, non-agriculture-relevant information shared by the same set of users. In short, we expect the network topology to display structural shifts depending on the type of information that is shared among its users. Specifically, we suggest that the more specialized the information, the more diffusion will rely on core users surrounded by brokers who export information to the broader public. This framework describes a process of information diffusion that often deviates from patterns observed in social networks, as the network behaves much like a broadcast system with pronounced amplification effects for information dissemination (Myers et al., 2014). Compared with decentralized systems, it stresses the diffusion of information from experts and elite users towards ordinary users, foregrounding classical diffusion models and underplaying the potential for horizontal information sharing allowed by social networks.

Lastly, a growing body of scholarship has explored the role of brokers in spreading innovation between Twitter communities (Frahm and Shepelyansky, 2012, Mantzaris, 2014), particularly in reference to nodes that do not belong to any community but that may bridge different groups (Shore et al., 2016, Takaguchi et al., 2014). Within this line of inquiry, Grabowicz et al. (2012) explored the relationship between weak, intermediary, and strong ties on Twitter and found it to be largely structured around groups, with personal interactions more likely to occur on internal links to the groups, and new information more likely to be channeled through links connecting different groups. While these studies have advanced the understanding of information diffusion across Twitter communities, they are largely focused on information diffusion between non-specialized communities. With this new analysis, we seek to understand how communities of interest interact with specialized and generic information. We explore how network formations within these communities are affected by the type of information being shared. As such, we do not consider brokerage across communities or inter-community information diffusion, but only the relationship between information diffusion and the type of content transmitted within communities.

With this study we test the hypothesis that the topology of Twitter network becomes increasingly more centralized as communities share more specialized information. We start by calculating the core-periphery score of the entire network and move forward towards the ten largest communities of interest within this network identified with the Walktrap community detection algorithm implemented in igraph, a max-modularity method based on random walks to find communities of densely connected vertices (Csardi and Nepusz, 2006, Pons and Latapy, 2005). These endogenous communities reflect areas of agricultural expertise with limited overlap across each topical subnet. Each community also presents a common group of users that tweeted both agriculture and non-agriculture-relevant tweets, thus allowing for estimating the core-periphery structure of the network in the absence or presence of specialized information.

Our substantive research questions are informed by the literature on social media and diffusion of innovations and inquire about the process by which specialized information spreads on Twitter. Does it spread from user to user in a decentralized fashion? Does it depend on influential users positioned at the center of the network? To observe these effects, we modelled and estimated the core-periphery profile of multiple, comparable subgraphs of the network that included the same set of users but resulted from messages that were either specialized (i.e., agriculture-relevant) or generic (i.e., non-agriculture-relevant). For the purposes of this study, the variation in core-periphery estimates are only meaningful if they refer to the same set of users but are generated by substantively different sets of information. By calculating the core-periphery profile of multiple subgraphs resulting from the exchange of specialized and generic information, and testing the significance of each test against a set of 100 randomized comparable networks, we test the following hypotheses to advance our understanding of how sharing specialized information on Twitter alters its network topology:

H1

The network structure of the Twitter agricultural social web is centralized and displays strong patterns of core-periphery;

H2

The network structure of Twitter agricultural social web becomes more centralized when users share specialized as opposed to generic information;

H3

Specialized, hashtag-based subnetworks are more centralized than generic, hashtag-based subnetworks.

The goal of our research design was to identify the agricultural network centered on the University of California Division of Agricultural and Natural Resources (UCANR), which is the organizational unit that coordinates Cooperative Extension in California comprising extension faculty, extension specialists, and extension agents dedicated to outreach activities related to agricultural and natural resources. UCANR and the University of California system comprise one of the largest, most sophisticated, and most experienced agricultural outreach systems in the world. On that basis, the UCANR provides an excellent case study to investigate the diffusion of specialized information.

We relied on a census-based approach to collect data, starting with 153 Twitter users identified by UCANR as important sources of information on the topics of agriculture and environment which are central to the mission of the organization.1 This purposeful research design seeks to begin with an exogenously determined community focused on specialized agricultural topics and related subfields of agricultural expertise. We believe this approach is appropriate to isolate an initial segment of the Twitter user base that is of interest for the purposes of this study. This initial set of users formed the seed nodes from which we snowballed data collection to the larger group of users following or being followed by these users, rendering a population of 59,761 Twitter accounts that tweeted a total of 285 M (285,628,862) messages since first joining Twitter.

This sample is intended to capture a specific group focused on California Agriculture where notions of expertise and communities of interest play a pivotal role. From the 59,761 accounts in the population, we managed to retrieve the timelines of 91% of the population (54,422 accounts) and geographic information from 73%. These users denote the networks nodes we explore in this study. The results reported in the next sections are limited to our sampling strategy, focused on the ANR-centered community of 153 users and their immediate network of followers and followees. Although limited to the University of California Division of Agricultural and Natural Resources, we believe this approach provides a defensible and relevant sampling of a knowledge network dedicated to the distribution of specialized information.

From the total of 5352 users left out of the network, we found that 501 accounts have yet to tweet a single message and 4894 accounts were protected or have been deactivated. Finally, 43 accounts were both protected and had not tweeted any message at the time of data collection. In addition to these silent accounts, Twitter Rest API limits access to a maximum of 3200 statuses (tweets) per user. The target population of 59,761 users includes 13,112 accounts that tweeted over this limit. This technical limitation imposes considerable challenges to retrieving complete sets of messages posted within a specified time-frame. As a result, we managed to retrieve only 65 M (65,294,710) tweets from 54,422 users, as opposed to 285 M potential tweets. These limitations are not only to be expected, but also experienced by most Twitter research and our conclusions are conditional on these constraints (Liang and Fu, 2015).

The temporal series is subject to variations as only a portion of the timelines violated Twitter restriction of 3200 tweets per user. We addressed this problem by identifying the average cut-off date for users that posted over 3199 tweets and removed messages posted prior to this date. We resorted to this procedure to filter the 65 M tweets collected from Twitter API and ended up with a total of 43 M messages. We subsequently removed messages that were not retweets or @-mentions—messages from which network edges are later drawn as proxies for information diffusion—and further reduced the dataset to 26 M messages.

Fig. 1 shows a histogram of messages binned by month, with a cut-off date at the end of 2013 (Fig. 1a) that includes the complete set of messages tweeted both by filtered users (>3200 tweets) and unfiltered users (<3200 tweets). Although the period from the end of 2013 to 2015 includes a comprehensive set of tweets posted by this community, we found inconsistencies in the temporal distribution of tweets. Kernel estimation shows that the data drops off artificially at the upper end of the time series. This is likely a result of user’s timelines being collected sequentially and thus at different points in time. In addition to that, we anticipated that older messages would fare relatively better in terms of retweets compared with newer messages, as they would have benefitted from a longer period to spread throughout the network.

We addressed this issue by selecting an intermediate period that was unaffected by variations resulting from data collection (Fig. 1b). This period comprises the entire year of 2014 and includes a total of 3.7 M (3,691,342) tweets. Our resulting dataset includes messages posted in 2014 and the analysis reported in this paper refers to this subset of tweets. The frequency of tweets binned by week is shown in Fig. 1b, with similar frequency distributions across filtered and unfiltered users. We expect these procedures to have addressed the restrictions imposed by Twitter REST API and to have provided a comprehensive set of messages posted by our population in 2014. In summary, our data collection is informed by Liang and Fu (2015) and relies on a purposive sample of Twitter users (egos) extended to accounts listed as their followers and/or followees.

The last step in data collection consists of obtaining the profiles and the timelines of the selected users (egos and alters) and processing the data to generate a network with various edge and node properties. We expect this approach based on sampling of users, rather than sampling of tweets, to provide a more reliable and replicable approach to analyzing individual and community-level social media behavior. We relied on the sampled data to graph a network of @-mentions and retweets connecting users, with A  B when B retweets A and A  B when A mentions B (thus following the directionality of the information flow). In both cases, we draw an @-mention or retweet edge connecting two accounts that have posted at least one message with specialized information at some point in the year of 2014. These sampling processes rendered a network of 4.4 M edges and 32 K nodes. Fig. 2 details the process of data collection, processing, and analysis, and describes the resulting network.

This approach to data collection is thus based on a specifically selected community of users, with the subgraphs generated for hypothesis testing being not only of comparable size, but resulting from patterns of interaction across regular communities of users. Specialized, agriculture related subgraphs are often less dense compared with their generic counterparts, but they result from patterns of information exchanged between the same set of users and observed during the same timeframe. This represents a considerable departure from hashtag-based studies, which necessarily filter information on contextual markers and thus include entirely different populations. In short, we expect this approach to data collection to allow us to better understand how Twitter network structure changes as users discuss different sets of topics. This is only possible once we can draw from the same population, with network nodes remaining unchanged over the multiple iterations of hypothesis testing.

Section snippets

Methods: the core-periphery profile

In this work we rely on the approach introduced by Della Rossa et al. (2013) to estimate the core-periphery structure of a network. By elaborating the dynamics of a random walker, a curve (the core-periphery profile) and a numerical indicator (the core-periphery score) are derived. The approach measures the extent to which the network is organized in core and periphery or, inversely, in an homogeneous structure. Simultaneously, a coreness value is attributed to each node, qualifying its

Results

We started by categorizing the communities identified by the community detection algorithm into ten specialized topical subnetworks. These ten large modules account for 80% of the graph (32,152 users) and the remaining, more sparsely connected nodes, are not considered in the following analyses. Consistent with previous results (Bastos et al., 2013), we found that the communities tweeted dominant hashtags that could be leveraged to distinguish substantive thematic communities. We used our

Discussion

The results reported herein indicate that the entire network shows strong patterns of core-periphery shifts due to a dense, cohesive core and a large sparsely connected periphery (Borgatti and Everett, 2000). The communities providing specialized agricultural information become significantly more star-shaped as users tweet, retweet, or comment on messages relaying specialized information. The shift from more horizontal, decentralized topologies in which users interact and discuss generic topics

Conclusion

In this paper we analyzed the relationship between Twitter network structure and the diffusion of specialized information. We relied on a snowball-based census of the California agriculture social web to identify the boundaries of multiple communities invested in agriculture and retrieved a sample of tweets comprising both specialized and generic information shared by this population. After identifying ten endogenous communities organized around topical themes, we controlled for the type of

Acknowledgments

This work was supported by the University of California Division of Agriculture and Natural Resources and carried out while the first author was a postdoctoral researcher at the University of California, Davis.

Marco Bastos is a lecturer of media and communication at City, University of London where he studies the cross-effects between online and offline social networks.

References (61)

  • M.T. Bastos et al.

    Tweeting across hashtags: overlapping users and the importance of language, topics, and politics

  • D. Centola et al.

    The spontaneous emergence of conventions: an experimental study of cultural evolution

    Proc. Natl. Acad. Sci.

    (2015)
  • D. Centola

    The spread of behavior in an online social network experiment

    Science

    (2010)
  • J.H. Choi et al.

    Making a global community on the net −global village or global metropolis?: A network analysis of usenet newsgroups

    J. Comput. Mediated Commun.

    (2002)
  • A. Chowdhury et al.

    Social media for enhancing innovation in agri-food and rural development: current dynamics in Ontario, Canada

    J. Rural Commun. Dev.

    (2013)
  • W.C. Clark et al.

    Crafting usable knowledge for sustainable development

    Proc. Natl. Acad. Sci.

    (2016)
  • G. Csardi et al.

    The igraph software package for complex network research

    InterJ. Complex Syst.

    (2006)
  • P. Csermely et al.

    Structure and dynamics of core/periphery networks

    J. Complex Netw.

    (2013)
  • F. Della Rossa et al.

    Profiling core-periphery network structure by random walkers

    Sci. Rep.

    (2013)
  • S.A. Delre et al.

    Diffusion dynamics in small-world networks with heterogeneous consumers

    Comput. Math. Org. Theory

    (2007)
  • S.N. Dorogovtsev et al.

    K-core organization of complex networks

    Phys. Rev. Lett.

    (2006)
  • K.M. Frahm et al.

    Google matrix of twitter

    Eur. Phys. J. B

    (2012)
  • M. Gabielkov et al.

    Studying Social Networks at Scale: Macroscopic Anatomy of the Twitter Social Graph

    (2014)
  • D.E. Gibbons

    Interorganizational network structures and diffusion of information through a health system

    Am. J. Public Health

    (2007)
  • E. Giuliani

    The structure of cluster knowledge networks: uneven and selective, not pervasive and collective

  • P.A. Grabowicz et al.

    Social features of online networks: the strength of intermediary ties in online social media

    PLoS One

    (2012)
  • F. Hermans et al.

    Structural conditions for collaboration and learning in innovation networks: using an innovation system performance lens to analyse agricultural knowledge systems

    J. Agric. Educ. Ext.

    (2015)
  • P. Holme

    Core-periphery organization of complex networks

    Phys. Rev. E

    (2005)
  • L. Klerkx et al.

    The emergence and functioning of innovation intermediaries in maturing innovation systems: the case of Chile

    Innov. Dev.

    (2015)
  • H. Kwak et al.

    What is Twitter, a social network or a news media?

  • Cited by (23)

    • Free riding in networks

      2023, European Economic Review
    • Twitter conversations reveal issue salience of aviation in the broader context of climate change

      2022, Journal of Air Transport Management
      Citation Excerpt :

      As in other research, Twitter networks are quite dispersed, but can be characterised by a small number of influential nodes (Park et al., 2015). Often these sub-sets or communities are not connected, but within themselves display a marked core-periphery structure (Bastos et al., 2018). The small and poorly connected network of the economic and industry frame deserves more attention as it might indicate that aviation stakeholders are not well represented in the communication system.

    • Nestedness in complex networks: Observation, emergence, and implications

      2019, Physics Reports
      Citation Excerpt :

      This result suggests that in communication networks, a nested architecture can emerge as a result of the emergence of a consensus on a given topic. A more recent study [141] also focused on temporal variations of the topology of a Twitter communication network, and found that centralization tends to increase when the discussion becomes more specialized. Given an observed structural pattern, it is important to assess how that pattern is related to other system-specific macroscopic properties.

    • A novel negative feedback information dissemination model based on online social network

      2019, Physica A: Statistical Mechanics and its Applications
      Citation Excerpt :

      It reveals the complex interactions and interactions between the user relationship structure, network community groups, and cyberspace information in social networks, and provide new understanding and theoretical support for social network evolution analysis. The effective management of information networks plays an important guiding role through reasonable use of network information, correct guidance of user behavior [16,17]. Based on the traditional SIR model, this paper has carried out in-depth research work in connection with the theory of communication, graph theory, and mean field theory.

    View all citing articles on Scopus

    Marco Bastos is a lecturer of media and communication at City, University of London where he studies the cross-effects between online and offline social networks.

    Carlo Piccardi is a professor of complex systems and networks at Politecnico di Milano where he explores network methods and applications.

    Michael Levy is a PhD Candidate and affiliate with the Data Science Initiative of UC Davis.

    Neil McRoberts is a quantitative biologist and interdisciplinary scientist working in the intersection between biology, economics, and human ecology.

    Mark Lubell is a political scientist and director of the Center for Environmental Policy and Behavior at UC Davis.

    View full text