Social Computing, Behavioral-Cultural Modeling and Prediction

In the context of modernization and development, the complex adaptive systems framework can help address the coupling of macro social constraint and opportunity with individual agency. Combining system dynamics and agent based modeling, we formalize the Human Development (HD) perspective with a system of asymmetric, coupled nonlinear equations empirically validated from World Values Survey (WVS) data, capturing the core qualitative logic of HD theory. Using a simple evolutionary game approach, we fuse endogenously derived individual socio-economic attribute changes with Prisoner’s Dilemma spatial intra-societal economic transactions. We then explore a new human development dynamics (HDD) model behavior via quasi-global simulation methods to explore economic development, cultural plasticity, social and political change.

In this paper we try to understand how racial segregation of the geographic spaces of three major US cities (New York, Los Angeles and Chicago) affect the mobility patterns of people living in them. Collecting over 75 million geo-tagged tweets from these cities during a period of one year beginning October 2012 we identified home locations for over 30,000 distinct users, and prepared models of travel patterns for each of them. Dividing the cities’ geographic boundary into census tracts and grouping them according to racial segregation information we try to understand how the mobility of users living within an area of a particular predominant race correlate to those living in areas of similar race, and to those of a different race. While these cities still remain to be vastly segregated in the 2010 census data, we observe a compelling amount of deviation in travel patterns when compared to artificially generated ideal mobility. A common trend for all races is to visit areas populated by similar race more often. Also, blacks, Asians and Hispanics tend to travel less often to predominantly white census tracts, and similarly predominantly black tracts are less visited by other races.

Most leading computational theories of decision-making under risk do not have mechanisms to account for the incorporation of cultural factors. Therefore, they are of limited utility to scholars and practitioners who wish to model, and predict, how culture influences decision outcomes. Fuzzy Trace Theory (FTT) posits that people encode risk information at multiple levels of representation – namely, gist, which captures the culturally contingent meaning, or interpretation, of a stimulus, and verbatim, which is a detailed symbolic representation of the stimulus. Decision-makers prefer to rely on gist representations, although conflicts between gist and verbatim can attenuate this reliance. In this paper, we present a computational model of Fuzzy Trace Theory, which is able to successfully predict 14 experimental effects using a small number of assumptions. This technique may ultimately form the basis for an agent-based model, whose rule sets incorporate cultural and other psychosocial factors.

We consider the problem of labeling actors in social networks where the labels correspond to membership in specific interest groups, or other attributes of the actors. Actors in a social network are linked to not only other actors but also items (e.g., video and photo) which in turn can be linked to other items or actors. Given a social network in which only some of the actors are labeled, our goal is to predict the labels of the remaining actors. We introduce a variant of the random walk graph kernel to deal with the heterogeneous nature of the network (i.e., presence of a large number of node and link types). We show that the resulting heterogeneous graph kernel (HGK) can be used to build accurate classifiers for labeling actors in social networks. Specifically, we describe results of experiments on two real-world data sets that show HGK classifiers often significantly outperform or are competitive with the state-of-the-art methods for labeling actors in social networks.

Stochastic cusp catastrophe model has been utilized extensively to model the nonlinear social and behavioral outcomes to detect the exisitance of cusp catastrophe. However the foundamental question on sample size needed to detect the cusp catastrophe from the study design point of view has never been investigated. This is probably due to the complexity of the cusp model. This paper is aimed at filling the gap. In this paper, we propose a novel Monte-Carlo simulation-based approach to calculate the statistical power for stochastic cusp catastrophe model so the sample size can be determined. With this approach, a power curve can be produced to depict the relationship between its statistical power and samples size under different specifications. With this power curve, researchers can estimate sample size required for specified power in design and analysis data from stochastic cusp catastrophe model. The implementation of this novel approach is illustrated with data from Zeeman’s cusp machine.

When a novel research topic emerges, we are interested in discovering how the topic will propagate over the bibliography network,

i.e.,

which author will research and publish about this topic. Inferring the underlying influence network among authors is the basis of predicting such topic adoption. Existing works infer the influence network based on

past

adoption cascades, which is limited by the amount and relevance of cascades collected. This work hypothesizes that the influence network structure and probabilities are the results of many factors including the social relationships and topic popularity. These heterogeneous information shall be optimized to learn the parameters that define the homogeneous influence network that can be used to predict

future

cascade. Experiments using DBLP data demonstrate that the proposed method outperforms the algorithm based on traditional cascade network inference in predicting novel topic adoption.

Determining the emergent thought leader of an ad hoc organization would have enormous benefit in a variety of domains including emergency response. To determine if such leaders could be identified automatically, we compared an observer-based assessment to an automated approach for detecting leaders of 3-person ad hoc teams performing a logistics planning task. The automated coding used a combination of indicator phrases indicative of reasoning and uncertainty. The member of the team with the most reasoning and least uncertainty matched the observer-based leader in two thirds of the teams. This determination could be combined with other analyses of the topics of discussion to determine emergent thought leaders in different domains. As an example, a real-time user interface providing this information is shown which highlights communications by others that are relevant to the automatically detected, topic-specific, emergent thought leader.

Two features are commonly observed in large and complex systems. First, a system is made up of multiple subsystems. Second there exists fragmented data. A methodological challenge is to reconcile the potential parametric inconsistency across individually calibrated subsystems. This study aims to explore a novel approach, called system-subsystem dependency network, which is capable of integrating subsystems that have been individually calibrated using separate data sets. In this paper we compare several techniques for solving the methodological challenge. Additionally, we use data from a large-scale epidemiologic study as well as a large clinical trial to illustrate the solution to inconsistency of overlapping subsystems and the integration of data sets.

Individuals make decisions every day. How they come up with estimates to guide their decisions could be a result of a combination of different information sources such as individual beliefs and previous knowledge, random guesses, and social cues. This study aims to sort out individual estimate assessments over multiple times with the main focus on how individuals weigh their own beliefs vs. those of others in forming their future estimates. Using dynamics modeling, we build on data from an experiment conducted by Lorenz et al. [1] where 144 subjects made five estimates for six factual questions in an isolated manner (no interaction allowed between subjects). We model the dynamic mechanisms of changing estimates for two different scenarios: 1) when individuals are not exposed to any information and 2) when they are under social influence.

We compare samples of tweets from the Twitter Streaming API constructed from different connections that tracked the same popular keywords at the same time. We find that on average, over 96% of the tweets seen in one sample are seen in all others. Those tweets found only in a subset of samples do not significantly differ from tweets found in all samples in terms of user popularity or tweet structure. We conclude they are likely the result of a technical artifact rather than any systematic bias.

Practically, our results show that an infinite number of Streaming API samples are necessary to collect “most” of the tweets containing a popular keyword, and that findings from one sample from the Streaming API are likely to hold for all samples that could have been taken. Methodologically, our approach is extendible to other types of social media data beyond Twitter.

The ability to learn network structure characteristics and disease dynamic parameters improves the predictive power of epidemic models, the understanding of disease propagation processes and the development of efficient curing and vaccination policies. This paper presents a parameter estimation method that learns network characteristics and disease dynamics from our estimated infection curve. We apply the method to data collected during the 2009 H1N1 epidemic and show that the best-fit model, among a family of graphs, admits a scale-free network. This finding implies that random vaccination alone will not efficiently halt the spread of influenza, and instead vaccination and contact-reduction programs should exploit the special network structure.

Social media has allowed researchers to induce large social networks from easily accessible online data. However, relationships inferred from social media data may not always reflect the true underlying relationship. The main question of this work is: How does the public social network reflect the private social network? We begin to address this question by studying interactions between players in a Massively Multiplayer Online Game. We trained a number of classifiers to predict the social ties between players using data on public forum posts, private messages exchanged between players, and their relationship information. Results show that using public interaction knowledge significantly improves the prediction of social ties between two players and including a richer set of information on their relationship further improves this prediction.

We propose a novel yet simple method for creating a stock market trading strategy by following successful stock market expert in social media. The problem of “how and where to invest” is translated into “who to follow in my investment”. In other words, looking for stock market investment strategy is converted into stock market expert search. Fortunately, many stock market experts are active in social media and openly express their opinions about market. By analyzing their behavior, and mining their opinions and suggested actions in Twitter, and simulating their recommendations, we are able to score each expert based on his/her performance. Using this scoring system, experts with most successful trading are recommended. The main objective in this research is to identify traders that outperform market historically, and aggregate the opinions from such traders to recommend trades.

We present a method to quantify the

political legitimacy

of a populace using public Twitter data. First, we represent the notion of legitimacy with respect to

k

-dimensional probabilistic topics, automatically culled from the politically oriented corpus. The short tweets are then converted to a feature vector in

k

-dimensional topic space. Leveraging sentiment analysis, we also consider the polarity of each tweet. Finally, we aggregate a large number of tweets into a final legitimacy score (i.e., L-score) for a populace. To validate our proposal, we conduct an empirical analysis on eight sample countries using related public tweets, and find that some of our proposed methods yield L-scores strongly correlated with those reported by political scientists.

Many organisms, human and otherwise, engage in path following in physical environments across a wide variety of contexts. Inspired by evidence that spatial search and information search share cognitive underpinnings, we explored whether path information could also be useful in a Web search context. We developed a prototype interface for presenting a user with the “search path” (sequence of clicks and queries) of another user, and ran a user study in which participants performed a series of search tasks while having access to search path information. Results suggest that path information can be a useful search aid, but that better path representations are needed. This application highlights the benefits of a cognitive science-based search perspective for the design of Web search systems and the need for further work on aggregating and presenting search trajectories in a Web search context.

Social network analysis is a rich field with many practical applications like community formation and hub detection. Traditionally, we assume that edges in the network have homogeneous semantics, for instance, indicating friend relationships. However, we increasingly deal with networks for which we can define multiple heterogeneous types of connections between users; we refer to these distinct groups of edges as layers. Naïvely, we could perform standard network analyses on each layer independently, but this approach may fail to identify interesting signals that are apparent only when viewing all of the layers at once. Instead, we propose to analyze a multi-layered network as a single entity, potentially yielding a richer set of results that better reflect the underlying data. We apply the framework of multi-objective optimization and specifically the concept of Pareto optimality, which has been used in many contexts in engineering and science to deliver solutions that offer tradeoffs between various objective functions. We show that this approach can be well-suited to multi-layer network analysis, as we will encounter situations in which we wish to optimize contrasting quantities. As a case study, we utilize the Pareto framework to show how to bisect the network into equal parts in a way that attempts to minimize the cut-size on each layer. This type of procedure might be useful in determining differences in structure between layers, and in cases where there is an underlying true bisection over multiple layers, this procedure could give a more accurate cut.

We use a synthetic population model of Washington DC, including residents and transients such as tourists and business travelers, to simulate epidemics of influenza-like illnesses. Assuming that the population is vaccinated at the compliance levels reported by the CDC, we show that additionally implementing a policy that encourages healthy behaviors (such as covering your cough and using hand sanitizers) at four major museum locations around the National Mall can lead to very significant reductions in the epidemic. These locations are chosen because there is a high level of mixing between residents and transients. We show that this localized healthy behavior intervention is approximately equivalent to a 46.14% increase in vaccination compliance levels.

Infectious disease spread is difficult to accurately measure and model. Even for well-studied pathogens, uncertainties remain regarding dynamics of mixing behavior and how to balance simulation-generated estimates with empirical data. While Markov Chain Monte Carlo approaches sample posteriors given empirical data, health applications of such methods have not considered dynamics associated with model error. We present here an Extended Kalman Filter (EKF) approach for recurrent simulation regrounding as empirical data arrives throughout outbreaks. The approach simultaneously considers empirical data accuracy, growing simulation error between measurements, and supports estimation of changing model parameters. We evaluate our approach using a two-level system, with “ground truth” generated by an agent-based model simulating epidemics over empirical microcontact networks, and noisy measurements fed into an EKF corrected aggregate model. We find that the EKF solution improves outbreak peak estimation and can compensate for inaccuracies in model structure and parameter estimates.

In recent times, social media sites such as Twitter have been extensively used for debating politics and public policies. These debates span millions of tweets and numerous topics of public importance. Thus, it is imperative that this vast trove of data is tapped in order to gain insights into public opinion especially on hotly contested issues such as abortion, gun reforms etc. Thus, in our work, we aim to gauge users’ stance on such topics in Twitter. We propose ReLP, a semi-supervised framework using a retweet-based label propagation algorithm coupled with a supervised classifier to identify users with differing opinions. In particular, our framework is designed such that it can be easily adopted to different domains with little human supervision while still producing excellent accuracy.

We propose a new item-ranking method that is reliable and can efficiently identify high-quality items from among a set of items in a given category using their review-scores which were rated and posted by users. Typical ranking methods rely only on either the number of reviews or the average review score. Some of them discount outdated ratings by using a temporal-decay function to make a fair comparison between old and new items. The proposed method reflects trust levels by incorporating a trust discount factor into a temporal-decay function. We first define the

MTDF (Multinomial with Trust Discount Factor) model

for the review-score distribution of each item built from the observed review data. We then bring in the notion of z-score to accommodate the trust variance that comes from the number of reviews available, and propose a z-score version of MTDF model. Finally we demonstrate the effectiveness of the proposed method using the MovieLens dataset, showing that the proposed ranking method can derive more reasonable and trustable rankings, compared to two naive ranking methods and the pure z-score based ranking method.

Large-scale cyber attack traffic can present challenges to identify which packets are relevant and what attack behaviors are present. Existing works on Host or Flow Clustering attempt to group similar behaviors to expedite analysis, often phrasing the problem as offline unsupervised machine learning. This work proposes online processing to simultaneously segment traffic observables and generate attack behavior models that are relevant to a target. The goal is not just to aggregate similar attack behaviors, but to provide situational awareness by grouping relevant traffic that exhibits one or more behaviors around each asset. The seemingly clustering problem is recast as a supervised learning problem: classifying received traffic to the most likely attack model, and iteratively introducing new models to explain received traffic. A graph-based prior is defined to extract the macroscopic attack structure, which complements security-based features for classification. Malicious traffic captures from CAIDA are used to demonstrate the capability of the proposed attack segmentation and model generation (ASMG) process.

Many proven clinical interventions that have been tested in carefully controlled field settings have not been widely adopted. We study an agent-based model of innovation adoption. Traditional statistical models average out individual variation in a population. In contrast, agent-based models focus on individual behavior. Because of this difference in perspective, an agent based model can yield insight into emergent system behavior that would not otherwise be visible. We begin with a traditional logic of innovation, and cast it in an agent-based form. The model shows behavior that is relevant to successful implementation, but that is not predictable using the traditional perspective. In particular, users move continuously in a space defined by degree of adoption and confidence. High adopters bifurcate between high and low confidence in the innovation, and move between these groups over time without converging. Based on these observations, we suggest a research agenda to integrate this approach into traditional evaluation methods.

Surveys are an important tool for measuring public opinion, but they take time to field, and thus may quickly become out of date. Social and news media can provide a more real-time measure of opinions but may not be representative. We describe a method for combining the precision of surveys with the timeliness of media using an agent-based simulation model to improve real-time opinion tracking and forecasting. Events extracted through text analytics of Afghan media sources were used to perturb a simulation of representative agents that were initialized using a population survey taken in 2005. We examine opinions toward the U.S., the Afghan government, Hamid Karzai, and the Taliban, and evaluate the model’s performance using a second survey conducted in 2006. The simulation results demonstrated significant improvement over relying on the 2005 survey alone, and performed well in capturing the actual changes in opinion found in the 2006 data.

Mobile devices collect a variety of information about their environments, recording “digital footprints” about the locations and activities of their human owners. These footprints come from physical sensors such as GPS, WiFi, and Bluetooth, as well as social behavior logs like phone calls, application usage, etc. Existing studies analyze mobile device footprints to infer daily activities like driving/running/walking, etc. and social contexts such as personality traits and emotional states. In this paper, we propose a different approach that uses multimodal mobile sensor and log data to build a novel user modeling framework called mFingerprint that can effectively and uniquely depict users. mFingerprint does not expose raw sensitive information from the mobile device, e.g., the exact location, WiFi access points, or apps installed, but computes privacy-preserving statistical features to model the user. These descriptive features obscure sensitive information, and thus can be shared, transmitted, and reused with fewer privacy concerns. By testing on 22 users’ mobile phone data collected over 2 months, we demonstrate the effectiveness of mFingerprint in user modeling and identification, with our proposed statistics achieving 81% accuracy across 22 users over 10-day intervals.

Online social networks such as Twitter have become a major type of information sources in recent years. However, this new public social media provides new gateways for malicious users to achieve various malicious purposes. In this paper, we introduce an extended trust model for detecting malicious activities in online social networks. The major insight is to conduct a trust propagation process over a novel heterogeneous social graph which is able to model different social activities. We develop two trustworthiness measures and evaluate their performance of detecting malicious activities using a real Twitter data set. The results revealed that the F-1 measure of detecting malicious activities in Twitter can achieve higher than 0.9 using our proposed method.

Massively multiplayer online games (MMOGs) offer a unique laboratory for examining large-scale patterns of human behavior. In particular, the study of guilds in MMOGs has yielded insights about the forces driving the formation of human groups. In this paper, we present a computational model for predicting guild membership in MMOGs and evaluate the relative contribution of 1) social ties, 2) attribute homophily, and 3) existing guild membership toward the accuracy of the predictive model. Our results indicate that existing guild membership is the best predictor of future membership; moreover knowing the identity of a few influential members, as measured by network centrality, is a more powerful predictor than a larger number of less influential members. Based on these results, we propose that community detection algorithms for virtual worlds should exploit publicly available knowledge of guild membership from sources such as profiles, bulletin boards, and chat groups.

Most of the recent research on online text communication has been conducted in social contexts with diverse groups of users. Here we examine a stable group of adult scientists as they chat about their work. Some scientists communicated in their first language (L1) and others communicated either in their L1 or in a second (L2) language. We analyze the production in English of emoticons and of lines of text and compare measures in L1 and L2 speakers. L1 and L2 speakers differed significantly along multiple measures. English L1 speakers used more lines of text per message. English L2 (French L1) speakers used more emoticons per message. Patterns suggest compensatory emoticon/text productivity. In future analyses we will undertake a more fine-grained analysis of how emoticon use varies across social and linguistic settings. Computer-mediated communication is often viewed as impoverished, but even our initial research provides hints that users repurpose the technology according to social dynamics previously associated only with face-to-face communication.

In this paper we semi-automatically construct a multilingual lexicon for Maslow’s seven categories of needs. We then use the semi-automatically constructed lexicons and a metaphor recognition system to analyze the change in rate of the expression of needs in the presence of metaphor. We examine four languages, English, Farsi, Russian, and Spanish, and focus on metaphors whose target concept is related to poverty or taxation.

Despite the potential benefits of investments in water and sanitation, individual level water treatment remains low in many developing countries. This paper explores the dynamic relationship between water transmitted infectious disease and water treatment behavior. Using evolutionary game theory, I endogenize water treatment decisions in a mathematical model of cholera. I calibrate the model for the ’08-’09 cholera outbreak in Zimbabwe. I show that prevalence dependent water treatment behavior is a factor contributing to endemic cholera. Additionally, I find that in absence of WHO interventions in Zimbabwe, the share of the population treating their water would have converged to a level that would have enabled cholera to persist in the population.

Crosstalk plays a critical role in prevention research to promote purposeful behavior change through randomized controlled trials. However, two challenges prevent researchers from assessing crosstalk between subjects in the intervention and the control conditions that may contaminate an intervention trial. First, it is very hard if not impossible to identify who in the intervention group have talked with whom in the control group; therefore the crosstalk effect cannot be statistically evaluated. Second no method is readily available to quantify crosstalk even if we know who has talked with whom. To overcome the challenges, we devised the random-pairing bootstrap (RPB) method based on statistical principles and adapted the mutual information (MI) technique from the information sciences. The established RPB method provides a novel approach for researchers to identify participants in the intervention and the control groups who might have talked with each other; the MI itself is an analytical method capable of quantifying both linear and nonlinear relationships on a variable between two groups of subjects who might have experienced information exchange. An MI measure therefore provides evidence supporting the effect from crosstalk on a target variable with data generated through RPB. To establish the PRB-MI methodology, we first conducted a systematic test with simulated data. We then analyzed empirical data from a randomized controlled trial (n=1360) funded by the National Institute of Health. Analytical results with simulated data indicate that RBP-MI method can effectively detect a known crosstalk effect with different effect sizes. Analytical results with empirical data show that effects from within-group crosstalk are greater than those of between-group crosstalk, which is within our expectation. These findings suggest the validity and utility of the RBP-MI method in behavioral intervention research. Further research is needed to improve the method.

This study examines the application of a human-agent based network to the illegal trade in antiquities. Specifically, this study tests whether the hierarchical pyramidal structure proposed by law enforcement in the case of Giacomo Medici’s trafficking ring is accurate. The results of the analysis reveal discrepancies in perceptions of how antiquities trafficking networks are organized, how they operate, and how cultural patterns and representation of criminal activity influence the perception of such network structures.

Corruption frequently occurs in many aspects of multi-party interaction between private agencies and government employees. Past works studying corruption in a lab context have explicitly included covert or illegal activities in participants’ strategy space or have relied on surveys like the Corruption Perception Index (CPI). This paper studies corruption in ecologically realistic settings in which corruption is not suggested to the players a priori but evolves during repeated interaction. We ran studies involving hundreds of subjects in three countries: China, Israel, and the United States. Subjects interacted using a four-player board game in which three bidders compete to win contracts by submitting bids in repeated auctions, and a single auctioneer determines the winner of each auction. The winning bid was paid to an external “government” entity, and was not distributed among the players. The game logs were analyzed posthoc for cases in which the auctioneer was bribed to choose a bidder who did not submit the highest bid. We found that although China exhibited the highest corruption level of the three countries, there were surprisingly more cases of corruption in the U.S. than in Israel, despite the higher PCI in Israel as compared to the U.S. We also found that bribes in the U.S. were at times excessively high, resulting in bribing players not being able to complete their winning contracts. We were able to predict the occurrence of corruption in the game using machine learning. The significance of this work is in providing a novel paradigm for investigating covert activities in the lab without priming subjects, and it represents a first step in the design of intelligent agents for detecting and reducing corruption activities in such settings.

Compared to desktop interfaces, touch-enabled mobile devices allow richer user interaction with actions such as drag, pinch-in, pinch-out, and swipe. While these actions have been already used to improve the ranking of search results or lists of recommendations, in this paper we focus on understanding how these actions are used in exploration tasks performed over lists of items not sorted by relevance, such as news or social media posts. We conducted a user study on an exploratory task of academic information, and through behavioral analysis we uncovered patterns of actions that reveal user intention to navigate new information, to relocate interesting items already explored, and to analyze details of specific items. With further analysis we found that dragging direction, speed and position all implied users’ judgment on their interests and they offer important signals to eventually learn user preferences.

This paper presents a theoretical model for interpreting the underlying meaning in virtual dialogue from computer-mediated communication (CMC). The objective is to develop a model for processing dialogues and understanding the meaning of online users’ social interactions based on available information behavior. The methodology proposed in this paper is built on a demonstrated observation that humans – in analogy to “sensors” in social networks – can detect unusual or unexpected changes in humans’ trustworthiness based on observed virtual behaviors. Even with limited resources such as email, blogs, online conversations, etc., humans “sensors” can infer meaning based on observed behaviors, and assign attributes to certain words or actions. The idiosyncratic nature of human observations can be arbitrated by an attribution mechanism that provides the basis for a systematic approach to measuring trustworthiness. In this paper, we discuss a particular trust scenario called the Leader’s Dilemma with the objective of identifying how anomalous online behavior can be interpreted as untrustworthy. We adopt the dyadic attribution model to analyze how a human disposition can be systematically uncovered based on words and actions, as evidenced by information behavior. This model is better suited for computational analysis of attribution engines. The novel goal of this research is to design a sensor system with the ability to attribute meaning to virtual interactions as supported by computer-mediated technologies.

This paper describes a computational model of time allocation in a social network. It particularly focuses on the impact of inequalities or so called stratification process on the efficiency of resources allocation. The presented model is a multi-agent system with implemented network evolution mechanism. The stratification is modeled by the order in which the agents’ actions are executed. The model was tested against as well different initial network parameters as a selection of time allocation heuristics. Efficiency of the system is measured using best applicable measures, e.g. agents schedule usage. By comparing the efficiency of the stratified system against baseline egalitarian system in our simulations we look for the answer whether social stratification brings benefits only for those who are on the top of the pyramid but decreases summarized performance of a society (in comparison to non-stratified systems).

Word-of-mouth is being replaced by online reviews on products and services. To identify the most useful reviews, many web sites enable readers to vote on which reviews they find useful. In this work we use three hypotheses to predict which reviews will be voted useful. The first is that useful reviews induce feelings. The second is that useful reviews are both informative and expressive, thus contain less adjectives while being longer. The third hypothesis is that the reviewer’s history can be used as a predictor. We devise impact metrics similar to the scientific metrics for assessing the impact of a scholar, namely

h-index, i

5

-index

. We analyze the performance of our hypotheses over three datasets collected from Yelp and Amazon. Our surprising and robust results show that the only good predictor to the usefulness of a review is the reviewer’s impact metrics score. We further devise a regression model that predicts the usefulness rating of each review. To further understand these results we characterize reviewers with high impact metrics scores and show that they write reviews frequently, and that their impact scores increase with time, on average. We suggest the term

local celebs

for these reviewers, and analyze the conditions for becoming local celebs on sites.

Energy security is placed at risk by exogenous supply shocks, in particular political crises and conflicts that disrupt resource extraction and transportation. In this paper, a computational model of the security of international crude oil supplies is described, and its output analyzed. The model consists of country agents, linked geographically and by a data-derived oil trade network. Countries stochastically experience crises, with probabilities and durations drawn randomly from data-fitted distributions. The effect of these crises on secure oil supplies is measured globally and by country, and the effect of conflict contagion and spare production capacity are also estimated. The model indicates that Russia, Eastern Europe, and much of the Global South are at the greatest risk of supply shocks, while American producers are at greatest risk of demand shocks. It estimates that conflict contagion decreases energy security slightly, while spare capacity has minimal effect.

We describe the theory, implementation and initial validation of a tool to assess interpersonal relationships and team states through non-intrusive discourse analysis. ADMIRE was developed to assess power/leadership relationships through scoring politeness behaviors in textual dialog to compute a graph of power asserted vs. afforded between individuals—an org chart—trackable over time. ADMIRE was formally tested in a military exercise where it proved 100% successful at deriving power relationships in 3 military chat rooms. Subsequent work extended ADMIRE’s core approach with novel linguistic behaviors to identify “Team Comfort/Routine” (C/R) indicating when a team is performing in a well-understood, relaxed task context. Initial validation is provided by discriminating the disastrous Apollo 13 mission from others.

The presence of social networks and mobile technology in form of secondary screens used in conjunction with television plays a significant role in the shift from traditional television to social television (TV). In this research, we investigate user interactions with secondary screens during live and non-live transmission of TV programs. We further explore the role of handheld devices in this second screen interaction. We perform statistical tests on more than 418,000 tweets from second screens for three popular TV shows. The results identify significant differences in the social element of second screen usage whilst the show is on air versus after the live broadcast, with the usage of handheld devices differing significantly in terms of the number of tweets during live telecast of TV shows. Desktop devices as second screens are also significant mediums for communication. This research identifies the change in users’ interaction habits in terms of information sharing and interaction for TV broadcasts via the presence of a computing device as a second screen.

This study evaluates if social network structures in primary care teams are related to biometric outcomes of diabetic alcohol drinkers. The study results show that primary care teams with less hierarchical face-to-face social networks (i.e. more connection density, more 3-tie closures and less network centrality) have better controlled HbA1c, LDL cholesterol and blood pressure among their diabetic alcohol drinking patients. Notably, more interconnected primary care teams, with members who engage others in face-to-face communication about patient care, who feel emotionally supported by their coworkers, and who feel like they work with friends, share the same goals and objectives for patient care and have better patient outcomes, as evidenced by the diabetic biometric measures of their team’s patients.

Crowdfunding platforms are becoming more and more popular for fund-raising of entrepreneurial ventures, but the success rate of crowdfunding campaigns is found to be less than 50%. Recent research has shown that, in addition to the quality and representations of project ideas, dynamics of investment during a crowdfunding campaign also play an important role in determining its success. To further understand the role of investment dynamics, we did an exploratory analysis of the time series of money pledges to campaigns in

Kickstarter

to investigate the extent to which simple inflows and first-order derivatives can predict the eventual success of campaigns. Using decision tree models, we found that there were discrete stages in money pledges that predicted the success of crowdfunding campaigns. Specifically, we found that, for the majority of projects that had the default campaign duration of one month in Kickstarter, money pledges inflow occurring in the initial 10% and 40-60%, and the first order derivative of inflow at 95-100% of the duration of the campaigns had the strongest impact on the success of campaigns. In addition, merely utilizing the initial 15% money inflows, which could be regarded as “seed money”, to build a predictor can correctly predict 84% of the success of campaigns. Implication of current results to crowdfunding campaigns is also discussed.

Computational social science methodologies are increasingly being viewed as critical for modeling complex individual and organizational behaviors in dynamic, real world scenarios. However, many challenges for identifying, representing and incorporating appropriate socio-cultural behaviors remain. Social theories provide rules, which have strong theoretic underpinnings and have been empirically validated, for representing and analyzing individual and group interactions. The key insight in this paper is that social theories can be embedded into computational models as functional mappings based on underlying factors, structures and interactions in social systems. We describe a generic framework, called a Culturally Infused Social Network (CISN), which makes such mappings realizable with its abilities to incorporate multi-domain socio-cultural factors, model at multiple scales, and represent dynamic information. We explore the incorporation of different social theories for added rigor to modeling and analysis by analyzing the fall of the Islamic Courts Union (ICU) regime in Somalia during the latter half of 2006. Specifically, we incorporate the concepts of homophily and frustration to examine the strength of the ICU’s alliances during its rise and fall. Additionally, we employ Affect Control Theory (ACT) to improve the resolution and detail of the model, and thus enhance the explanatory power of the CISN framework.

Texts reveal the subjects of interest in research fields, and the values, beliefs, and practices of researchers. In this study, texts are examined through bibliometric mapping and topic modeling to provide a bird’s eye view of the social dynamics associated with the diffusion of research synthesis methods in the contexts of Social Work and Women’s Studies. Research synthesis texts are especially revealing because the methods, which include meta-analysis and systematic review, are reliant on the availability of past research and data, sometimes idealized as objective, egalitarian approaches to research evaluation, fundamentally tied to past research practices, and performed with the goal informing future research and practice. This study highlights the co-influence of past and subsequent research within research fields; illustrates dynamics of the diffusion process; and provides insight into the cultural contexts of research in Social Work and Women’s Studies. This study suggests the potential to further develop bibliometric mapping and topic modeling techniques to inform research problem selection and resource allocation.

Many social, biological, and technological networks display substantial non-trivial topological features. One well-known and much studied feature of such networks is the scale-free power-law distribution of nodes’ degrees.

Several works further suggest models for generating complex networks which comply with one or more of these topological features. For example, the known Barabasi-Albert ”preferential attachment” model tells us how to create scale-free networks.

Since the main focus of these generative models is in capturing one or more of the static topological features of complex networks, they are very limited in capturing the temporal dynamic properties of the networks’ evolvement. Therefore, when studying real-world networks, the following question arises: what is the mechanism that governs changes in the network over time?

In order to shed some light on this topic, we study two years of data that we received from

eToro

: the world’s largest social financial trading company.

We discover three key findings. First, we demonstrate how the network topology may change significantly along time. More specifically, we illustrate how popular nodes may become extremely less popular, and emerging new nodes may become extremely popular, in a very short time. Then, we show that although the network may change significantly over time, the degrees of its nodes obey the power-law model at any given time. Finally, we observe that the magnitude of change between consecutive states of the network also presents a power-law effect.

This paper shows how big data can be experimentally used at large scale for marketing purposes at a mobile network operator. We present results from a large-scale experiment in a MNO in Asia where we use machine learning to segment customers for text-based marketing. This leads to conversion rates far superior to the current best marketing practices within MNOs.

Using metadata and social network analysis, we created new metrics to identify customers that are the most likely to convert into mobile internet users. These metrics falls into three categories: discretionary income, timing, and social learning. Using historical data, a machine learning prediction model is then trained, validated, and used to select a treatment group. Experimental results with 250 000 customers show a 13 times better conversion-rate compared to the control group. The control group is selected using the current best practice marketing. The model also shows very good properties in the longer term, as 98% of the converted customers in the treatment group renew their mobile internet packages after the campaign, compared to 37% in the control group. These results show that data-driven marketing can significantly improve conversion rates over current best-practice marketing strategies.

Surveys of rapidly-developing countries have shown that huge increases in average personal wealth are frequently accompanied by little or no increase in average (self-reported) happiness. We propose a simple agent-based model that may help to explain this phenomenon. The model shows that under certain conditions, the cumulative effect of individuals’ free choices of employment that maximizes their (self-perceived) personal well-being may actually produce a continuing decrease in the population’s average well-being. Like the proverbial “frog in a pot”, the eventual effect is worse when the onset of the decrease is more gradual. More generally, the model indicates that there is a natural tendency in free-market societies for well-being to become defined in increasingly materialistic terms. We discuss the implications of our model on the issue of incentive pay for teachers, and argue that our model may also provide insight into other situations where individuals’ free-market choices lead to progressive worsening of the population’s average well-being.

Agent-based modeling (ABM) is a powerful tool for the study of complex systems; but agent-based models are notoriously difficult to create, modify, and reason about, especially in contrast to system dynamics models. We argue that these difficulties are strongly related to the choice of specification language, and that they can be mitigated by using functional reactive programming (FRP), a paradigm for describing dynamic systems. We describe Frabjous, a new language for agent-based modeling based on FRP, and discuss its software engineering benefits and their broader implications for language choice in ABM.

In the space exploration domain, limitations in the Deep Space Network and the lack of real-time communication capabilities will impact various aspects of future long duration exploration such as a multi-year mission to Mars. One dimension of interest is the connection between flight crews and their Earth-based social support system, their family, friends, and colleagues. Studies in ground-based analogs of Isolated and Confined Environments (ICE) such as Antarctica have identified sensory deprivation and social monotony as threats to crew psychological well-being. Given the importance of behavioral health to mission success and the extreme conditions of space travel, new methods of maintaining psycho-social health and social connections to support systems are critical. In this paper we explore the use of Virtual Environments (VEs) and Virtual Agents (VAs) as tools to facilitate asynchronous human-human communication, and counteract behavioral health challenges associated with prolonged isolation and deep space exploration.

This work focuses on detecting emerging civil unrest events by analyzing massive micro-blogging streams. Specifically, we propose an early detection system consisting of a novel cascade of text-based filters to identify civil unrest event posts based on their topics, times and locations. In contrast to the model-based prediction approaches, our method is purely extractive as it detects relevant posts from massive volumes of data directly. We design and implement such a system in a distributed framework for scalable processing of real world data streams. Subsequently, a large-scale experiment is carried out on our system with the entire dataset from Tumblr for three consecutive months. Experimental result indicates that the simple filter-based method provides an efficient and effective way to identify posts related to real world civil unrest events. While similar tasks have been investigated in different social media platforms (e.g., Twitter), little work has been done for Tumblr despite its popularity. Our analysis on the data also shed light on the collective micr-oblogging patterns of Tumblr.

Social media have played a substantial role in supporting collective actions. Reports state that protesters use blogs, Facebook, Twitter, YouTube and other online communication media and environments to mobilize and spread awareness. In this research, we focus on studying the process of formation of online collective action (OCA) by analyzing the diffusion of hashtags. We examine the recently organized Saudi Arabian women’s right to drive campaign, called ‘Oct26Driving’ and collected the Twitter data, starting from September 25, 2013 to the present. Given the definitive nature of hashtags, we investigate the co-evolution of hashtag usage and the campaign network. The study considers the dominant hashtags dedicated to the Oct26Driving campaign, viz., ‘#oct26driving’ and ‘#قيادة_26اكتوبر’. Morteover, it identifies cross-cultural aspects with individual hashtag networks, with Arabic hashtags relating to local factors and English hashtags contributing to transnational support from other organizations, such as those related to human rights and women’s rights. Despite the wide news media coverage of social movements, there is a lack of systematic methodologies to analytically model such phenomena in complex online environments. The research aims to develop models that help advance the understanding of interconnected collective actions conducted through modern social and information systems.

Without time and geographical limitations, online smoking cessation programs attract a lot of users to help them quit smoking. This study compares two online smoking cessation support groups, QuitNet Forum and QuitNet Facebook, to evaluate the influence of software features that affect communication. We collect data from posts and comments of these two communities respectively, and compare user behavior from aspects of response immediacy and social network analysis. The associations between user behavior and quit stages were investigated. It is found that users of QuitNet Forum participate in communications more actively than users of QuitNet Facebook. The user behavior of QuitNet Facebook has a wider spectrum than that of QuitNet Forum.