Our behaviors are significantly affected by others. We seek advice when choosing a college, taking a job offer and relocating to a new place. We consider ratings when downloading an APP, purchasing a merchant, watching a movie and booking a hotel. The increasing development of social media technology allows individuals to share and exchange their opinions faster than ever, which embeds us into a social network tightly connecting each individual. Consequently, there have been intensive investigations on the social influence in networks, including the dynamics of the information spreading, the effects of network structure and the heterogeneous influence ability and susceptibility among individuals, the prediction and the control of the influence, and so on.
The aim of this satellite is to bring together scientists and researchers in multiple areas to present and discuss their recent works on social influence. The speakers are from network science, computer science and research team in IT company, characterized by very diverse backgrounds. We believe a joint discussion and communication, as planned to be performed at this satellite, can bring more insights into fundamental issues of this topic. More importantly, it will stimulate research on problems in real social systems, bringing solutions and applications to the industry in a near future.
We still have a few slots for a 10-minute presentation with slides. Participants are invited to submit via EasyChair an abstract of maximum 1 page in PDF format, specifying title, author(s), affiliation(s) and e-mail address(es). Contributions are evaluated on a rolling base, until all slots are filled up (but no later than April 30th). Please makes sure to submit your abstract soon, if you plan on attending and contributing to the workshop.
White River 102
|09:00 - 09:05||Opening remarks|
|09:05 - 09:40||Invited talk||Impact of social influence on society stability||Boleslaw K. Szymanski|
|09:40 - 10:15||Invited talk||The model of social influence and its applications on WeChat||Peng He|
|10:15 - 10:45||Coffee break|
|10:45 - 11:20||Invited talk||Sequential seeding: a meta-method boosting information diffusion coverage||Przemysław Kazienko|
|11:20 - 11:35||Contributed talk||Network structure in artistic influence||Michael Kitromilidis and Tim S. Evans|
|11:35 - 11:50||Contributed talk||Articulation points in complex networks||Liang Tian, Amir Bashan, Daning Shi and Yang-Yu Liu|
|11:50 - 12:05||Contributed talk||Multinomial matrix factorization: |
joint inference of attachment function and node fitnesses in dynamic networks
|Thong Pham, Paul Sheridan and Hidetoshi Shimodaira|
|12:05 - 14:00||Lunch break|
|14:00 - 14:35||Invited talk||Searching for Influencers via Optimal Percolation: from Twitter to the Brain||Flaviano Morone|
|14:35 - 15:10||Invited talk||Visual Analysis of WeChat Public Official Articles’ Propagation and Popularity Prediction||Quan Li|
|15:10 - 15:45||Invited talk||Finding Fake News||Giovanni Luca Ciampaglia|
|15:45 - 16:15||Coffee break|
|16:15 - 16:50||Invited talk||Learning Concise Representations of Users’ Influences through Online Behaviors||Shenghua Liu|
|16:50 - 17:25||Invited talk||Maximum entropy sampling in complex networks||Filippo Radicchi|
|17:25 - 17:35||Closing remarks|
Title: Impact of Social Influence on Society Stability
Abstract: The naming game has become an archetype for linguistic evolution and mathematical social behavioral analysis. Here, we discuss what naming game with committed minorities reveals about diversity of opinion in a society and its evolution. Clearly, without ability to accept new ideas, the social system stagnates in the status quo. In contrast, easy adoption of ideas makes the state in which it is possible inherently unstable and as a result it may reach the states that may be undesirable. In the talk we present results that summarize several publications on this topic. First, in naming game without committed agents, similarly to voter model, a new opinion needs to become a majority to become full dominant, which leads to stagnation of opinions. Introduction of a single committed minority changes the situation; the committed minority needs to achieve the tipping point, which for fully connected graph is close to 10% to rapidly spread over the entire system. That makes change of opinions orderly since the worthy new idea needs to get reasonable minority committed to it to become universally accepted. However, as the number of committed group grows, the tipping point increases with the strength of opposing minorities. At the certain strength of committed opposition, consensus on an opinion becomes impossible with society stagnating in the rivalry of committed minorities. The final case arises when the number of committed minorities is proportional to the size of the society, in which many small groups compete for attention. We will briefly discuss a robust method that handles this case. The initial condition plays a crucial role in the ordering of the system. We find that the system with high Shannon entropy has a higher consensus time and a lower critical fraction of committed agents compared to low entropy states. We also show that the critical number of committed agents decreases with the number of opinions, and grows with the community size for each word. Finally, we give historical examples of opinion evolution exemplifying some of the discussed cases.
Title: Maximum entropy sampling in complex networks
Abstract: Many real-world systems are characterized by stochastic dynamical rules where a complex network of dependencies among individual elements probabilistically determines their state. Even with full knowledge of the network structure and of the stochastic rules of the dynamical process, the ability to predict system configurations is generally characterized by large uncertainty. Sampling a fraction of the nodes and deterministically observing their state may help to reduce the uncertainty about the unobserved nodes. However, choosing these points of observation with the goal of maximizing predictive power is a highly nontrivial task, depending on the nature of the stochastic process and on the structure of the underlying network. In this talk, I will introduce a computationally efficient algorithm to determine quasi-optimal solutions for arbitrary stochastic processes defined on generic sparse topologies. I'll show that the method is effective for various processes on different substrates. I'll further show how the method can be fruitfully used to identify the best nodes to label in semi-supervised probabilistic classification algorithms.
Title: Searching for Influencers via Optimal Percolation: from Twitter to the Brain
Abstract: The whole frame of interconnections in complex networks is hinged on a specific set of nodes, called influencers, much smaller than the total size, which if activated would cause the spread of information to the whole network; or, if immunized, would prevent the diffusion of a large scale epidemic; or efficiently binds the different processing units of the brain. Localizing this optimal, i.e. minimal, set of influencers is an important problem in network science. I illustrate the theoretical framework to identify superspreaders using the concept of optimal percolation. This optimization problem is solved by minimizing the energy of a many-body system, where the form of the interactions is fixed by the non- backtracking matrix of the network. Big data analyses in social networks reveal that the set of superspreaders identified with our theory is much smaller than the one predicted by heuristic centralities. Remarkably, a large number of previously neglected weakly connected nodes emerges among the optimal influencers. In telecommunication networks, we find that the influencers predicted by the theory are highly correlated with personal economic status. To validate this finding, we carry out a marketing campaign that shows a three-fold increase in response rate by targeting individuals identified by our social network metrics as compared to random targeting. In brain networks, the theory identifies the essential nodes for the integration of the memory network. This is confirmed by pharmacogenetic inactivation of the predicted set of essential nodes, which, in fact, prevents the formation of the integrated network.
Title: Finding Fake News
Abstract: Two-thirds of all American adults access the news through social media. But social networks and social media recommendations lead to information bubbles, and personalization and recommendations, by maximizing the click-through rate, lead to ideological polarization. Consequently, rumors, false news, conspiracy theories, and now even fake news sites are an increasingly worrisome phenomena. While media organizations (Snopes.com, PolitiFact, FactCheck.org, et al.) have stepped up their efforts to verify news, political scientists tell us that fact-checking efforts may be ineffective or even counterproductive. To address some of these challenges, researchers at Indiana University are working on an open platform for the automatic tracking of both online fake news and fact-checking on social media. The goal of the platform, named Hoaxy, is to reconstruct the diffusion networks induced by hoaxes and their corrections as they are shared online and spread from person to person
Title: The model of social influence and its applications on WeChat
Abstract: analysis and model has become one of the most important technieques in social media marketing. In this talk, I will first introduce applications of social influence in WeChat soical marketing.Then I will talk about the social influence model we use in WeChat. Specifically, empirical analysis on pairwise and structural influence, their network-embedding based measurement and a data-driven influence diffusion model will be discussed. Finally, I will share a couple of influence related problems and our thinkings.
Title: Sequential seeding: a meta-method boosting information diffusion coverage
Abstract: Information spreading in complex networks is often modeled as diffusing information with certain probability from nodes that possess it to their neighbors that do not. Information cascades are triggered when the activation of a set of initial nodes – seeds – results in diffusion to large number of nodes. Several novel approaches for seed initiation that replace the commonly used activation of all seeds at once with a sequence of initiation stages are introduced. Sequential strategies at later stages avoid seeding highly ranked nodes that are already activated by diffusion active between stages. The gain arises when a saved seed is allocated to a node difficult to reach via diffusion. Sequential seeding and a single stage approach are compared using various seed ranking methods and diffusion parameters on real complex networks. The experimental results indicate that, regardless of the seed ranking method used, sequential seeding strategies deliver better coverage than single stage seeding in about 90% of cases. Longer seeding sequences tend to activate more nodes but they also extend the duration of diffusion. Various variants of sequential seeding resolve the trade-off between the coverage and speed of diffusion differently
Visual Analysis of WeChat Public Official Articles’ Propagation and Popularity Prediction
Abstract: As a new type of social networking services, WeChat has already become ubiquitous in people’s daily mobile communication. The abundant information on social media intensifies the competition of WeChat Public Official Articles (e.g., posts) for gaining user attention due to the zero-sum nature of attention. Therefore, only a small portion of information tends to become extremely popular while the rest remains unnoticed or quickly disappears.In this talk, we first investigate how WeChat Public Official Account Articles propagate in WeChat platform and carry out visual analysis tasks on the propagation network. Then, we adopt a user-centered method to design and develop an interactive visualization system for visually analyzing and reasoning the popularity of WeChat articles based on an adapted popularity prediction model. Our system first assists users in understanding the prediction model and then enables users to infer the underlying reasons which account for the propagation of the WeChat article. The proposed interactive visualization supports users to uncover the working mechanism behind the model and improve the model accordingly based on the insights gained.
Title: Learning Concise Representations of Users’ Influences through Online Behaviors
Abstract: Social network services generally allow users to post, forward, share or "like" a piece of information; order a product or comment on a hotel; "check in" a place of interest. All the above behaviors can be grouped as temporal sequences of users and at what time users act. While it is well known that social network users' behaviors influence each other, a fundamental problem in influence maximization, opinion formation and viral marketing is that users’ influences are difficult to quantify.Previous work has directly defined an independent model parameter to capture the interpersonal influence between each pair of users. However, these models do not consider how influences depend on each other if they originate from the same user or if they act on the same user. However, these models do not consider how influences depend on each other if they originate from the same user or if they act on the same user. To do so, these models need a parameter for each pair of users which results in high dimensional models easily trapped into the overfitting problem. Given these problems, another way of defining the parameters is needed to consider the dependencies. Thus we propose a model that defines parameters on every user with latent influence vector and susceptibility vector. Such low dimensional and distributed representations naturally make the interpersonal influences involving the same user coupled with each other, and in turn, reduce the model complexity. Additionally, the model can easily consider the sentimental polarities of users’ messages and how sentiment affects users’ influences. We conduct extensive experiments on real Microblog data, showing that our model with distributed representations achieves better performances. Furthermore, without parameterization of the influence functions, we propose to use recurrent neural network (RNN) to encode the cascades with attention mechanism. Our pilot shows some improvement results.