Video now available: https://www.youtube.com/watch?v=LtZhzoA8B0E
We are delighted to invite Professor Moses Boudourides to present “The Temporal Hypergraph of The Andy Warhol’s Diaries” today, 4pm, at FSB 1-483.
In Pat Hackett’s volume “The Andy Warhol Diaries,” we were able to identify automatically 2278 names of persons (celebrities, artists etc.), who happen to be mentioned in 2024 diaries during a period of 12 years, from November 24, 1976, to February 17, 1987. This was done using the standard Natural Languages Processing technique of POS tagging for the detection of 2-grams or 3-grams corresponding to persons’ names (first-last name, possibly including a middle name or a composite surname). Thus, each diary written in a particular day corresponds to a set of persons, who co-occur in the textual contents of the diary.
From the point of view of the theory of hypergraphs, the collection of diaries generates a (simple) undirected hypergraph G=(V,E), where each element of the set X, called vertex, is a person and each element of the set E, called hyperedge, is a set of (distinct) persons. In other words, each hyperedge is interpreted as a diary at a certain date involving a set of persons who happen to co-occur in that diary (persons counted as singletons even if they can be mentioned more than once in the same diary). Apparently, under this co-occurrence relation, The Andy Warhol’s Diaries give rise to a temporal (longitudinal) hypergraph, in which every hyperedge (diary) has a distinct time stamp (date).
Equivalently, from the network-analytic view, a single diary at a certain date corresponds to a clique of persons co-occurring in this diary. Moreover, by aggregating diaries written in a certain time period, one obtains a weighted undirected graph of persons, in which two persons are adjacent as far as there exists at least one date in this period, during which these persons co-occur in the corresponding diary (the number of such dates/diaries being the adjacency weight). Furthermore, in the usual way, one can define egocentric subgraphs during a certain period of dates. Notice that now k (larger than 2) egocentric subgraphs may define a multi-layer substructure in the hypergraph of diaries in that period of dates. Here, we are going to give an account of a number of different temporal-hypergraph-analytic computations on The Andy Warhol Diaries.
Moses Boudourides Ph.D. is Visiting Professor of Mathematics at NYU Abu Dhabi (https://nyuad.nyu.edu/en/academics/divisions/science/faculty/moses-boudourides.html). He is also in the Faculty of Northwestern University School of Professional Studies Data Science Program (https://sps.northwestern.edu/masters/data-science/faculty.php) and Affiliated Faculty at the Science of Networks in Communities (SONIC) at Northwestern University (http://sonic.northwestern.edu/people/affiliated-faculty/moses-boudourides/).
His research work is on social network analysis, computational social science and digital humanities. In particular, his work in digital humanities is focused on networks of literary text and data analysis of scientometric datasets with emphasis on the temporal assortativity of various attributes in co-authorship and co-publication networks (as authors’ gender, publication keywords and journals’ Open Access type).
Early in 2019, he was awarded a Robert K. Merton Visiting Research Fellowship from the Institute for Analytical Sociology (IAS) at Linköping University in Sweden.
We are delighted to invite Professor David Krackhardt to present a webcast on An Introduction to the Power of Networks in Organizations tomorrow (Friday), Nov 22nd, 2-3pm during livestream: https://bluejeans.com/425928897
David Krackhardt is Professor of Organizations at the Heinz School of Public Policy and Management and the Tepper School of Business, Carnegie Mellon University. Prior appointments include faculty positions at Cornell’s Graduate School of Management, the University of Chicago’s Graduate School of Business, INSEAD (France) and the Harvard Business School. He received a BS degree from the Massachusetts Institute of Technology and a PhD from the University of California, Irvine. Over the past 15 years, his research has focused on how the theoretical insights and methodological innovations of network analysis can enhance our understanding of how organizations function. He pioneered the concept of “cognitive social structures”, wherein individuals provide their perceptions of the entire network in which they are embedded. He empirically has related these perceived structures to turnover, reputations and power in organizations. Another interest of his has been in developing methodologies for better understanding networks and their implications. His contributions in this arena include adapting the quadratic assignment procedure to multiple regression analyses of network data. In addition, he has developed methods drawing from graph theory for studying the shape and structure of organizations as a whole.
His published works have appeared in a variety of journals in the fields of psychology, sociology, anthropology and management. His current research agenda includes developing models of diffusion of controversial innovations, exploring and testing visual representations of networks, identifying effective leverage points for organizational change, and exploring the roles of Simmelian (super-strong) ties in organizations. Prof. Krackhardt was born in 1950 in Massachusetts. He is married with three children.
Video now available on YouTube: https://youtu.be/F6uAm7Myx34
SONIC was delighted to welcome Dr. Laurence Lock Lee on November 11th to share his insights into using computational social science for business analytics.
Video now availabe on Youtube: https://youtu.be/I9XBbr2ctpM
SONIC is proud to welcome Dr. Hank Green of Indiana University. He will speak on Wednesday, Nov 6th at 10:30 am in Frances Searle Building, Room 3-417. Please contact Carmen Chan with any questions.
Department of Applied Health Science, Indiana University School of Public Health
Indiana University Network Science Institute
Network-based Strategies for Improving PrEP Availability Among those at High Risk forHIV
Pharmaceutical Pre-Exposure Prophylaxis (PrEP) has been proven effective against HIV infection, but it is underutilized as an HIV prevention strategy in part due to a lack of familiarity with PrEP treatment regimes among healthcare providers that serve the most at risk communities. Lack of healthcare provider knowledge has been identified as a substantial contributor to the continuing increase of HIV infections among sexual and gender minority patients. An Institute of Medicine report states, “few physicians are knowledgeable about or sensitive to lesbian, gay, bisexual, and transgender (LGBT) health risks or needs.” Nevertheless, physicians often serve as key points of access for this preventive treatment among minority patients who may not know the full range of prevention options. Thus, physicians’ lack of knowledge is likely to limit the highest risk patients’ awareness of, screening for, and access to PrEP. In this study, we propose to develop network-based strategies that help identify physicians who serve populations at high risk for HIV but who prescribe PrEP infrequently and those physicians likely to influence the prescribing behavior of their peers, facilitating the dissemination of information about PrEP via strategically targeted information campaigns.
Physician networks inferred from medical claims data protect patient identities but enable insights into the numbers and kinds of patients shared by physicians across the country and approximate the structure of physicians’ professional communication networks. Prior research has shown that physician networks inferred from shared patients are strongly predictive of the diffusion of shared medical practices. Medical knowledge is exchanged among physicians who share patients with one another because they often go to each other for advice about new treatments, difficult cases, etc. Medical claims data can provide more reliable information than data provided by physician surveys because the latter typically exhibit low response rates and a high likelihood of nonresponse bias. Additionally, acquiring a complete national set of medical claims is affordable whereas acquiring a set of complete national physician surveys would be prohibitively expensive.
Using national-scale claims data and the network generated from them, we tested hypotheses that ask whether PrEP prescribing behavior in physician sub-communities varies based on the whether a physician 1) is or has close ties to an infectious disease specialist, 2) shares patients with PrEP prescribing physicians, 3) shares patients with a doctor that treats a high proportion of HIV patients, 4) treats patients covered by specific insurance providers. While testing these and other hypotheses, we will account for contextual and geographic factors including location of providers in states with high incidence of HIV or in regions with high stigma against MSM. Understanding which of these network features may be associated with knowledge of and prescription of PreP will lead to new strategies that identify doctors who may be most influenced to begin or increase their recommendation of PrEP or who are positioned to shape the PrEP prescribing behavior of their peers, ultimately increasing its utilization among those at risk for HIV. The strategy, while focused here on PrEP, is broadly generalizable for any innovation in medical care that would benefit from a more targeted approach. This study has 3 aims:
Aim 1. Construct a physician-to-physician network relevant to the population indicated for PrEP. We used aggregated medical claims from both private and public payers to create a physician-to-physician network using standard network approaches. The data included all claims for patients being treated with drugs uniquely prescribed for STIs such as HIV and genital herpes, establishing a baseline of shared patients representative of the population indicated for PrEP.
Aim 2. Test alternative hypotheses for processes that may drive PrEP prescription rates. We used information about the distribution of PrEP prescribing in this network to assess the relative contribution of 1) social influence 2) differential awareness of PrEP, 3) geographic variation related to features of the HIV epidemic (such as stigma), and 4) other network based features such as clustering to prescription rates while controlling for factors such as the rural or urban location of a doctor.
Aim 3. We will use the structure of the physician network developed in Aim 1 and the results of our analyses in Aim 2 to identify targets for a strategic information campaign. We will generate a series of rules that identify physician characteristics associated with being an appropriate target and an appropriate peer change agent. Then, based on those rules we will simulate the diffusion of PrEP information and PrEP prescribing behavior in the network we generated in Aim 1. We will also contact a small set of targets and change agents to assess the accuracy of our rules and determine the opinions of these physicans regarding the broader utility of such targeted campaigns for increasing PrEP prescribing, fundamentally affecting the PrEP care cascade.
To learn more about the speaker, please visit: https://publichealth.indiana.edu/research/faculty-directory/profile.html?user=hdgreen
The SONIC Speaker Series presents
University of Lausanne (UNIL), Switzerland
A Multi-Level Network Perspective on Complex Urban Systems
SONIC Lab is proud to welcome Dr. Céline Rozenblat of the University of Lausanne. She will speak on Tuesday, May 21st at 10 am in Frances Searle Building, Room 1-483. Please contact Brent Hoagland with any questions.
Abstract: In the context of knowledge and information societies, new tendencies in the long/medium term evolution of urban systems, together with new data and methods, require that existing theoretical assumptions and conceptualizations be challenged as global urban hierarchies are reconfigured. The connection between urban systems at different levels of organization becomes more and more relevant for understanding urban systems and their transformations. But the current inter-urban perspective is not sufficient to encompass these dynamics. Other cognitive, social, and institutional proximities in the innovation processes combine with spatial proximities. It leads us to consider cities in several dimensions of proximities in a multilayer perspective. The evolution of power distributions inside and between cities reshapes the world organization of central/peripheral cities and the complexity of the global urban system. Actors as multinational firms, or high-level innovation centers, participate actively in these reconfigurations that concentrate wealth, control, innovation, and attractiveness in a few cities. In the complexity of this multi-level system, how is regionalization of the world reshaping in a multipolar urban world? How does the multi-level perspective highlight some resilience properties? The methodologies derived from complex systems sciences bring new forms of intelligibility to worldwide urban dynamics.
Céline Rozenblat, is professor of Urban Geography at the University of Lausanne, Switzerland and president of the urban commission of the International Geographical Union (IGU). She studies systems of cities at European and world scales, multinational firm networks, inter-urban dynamics, comparative urban data, mapping and visualization of networks in geography, and spatial analysis. For several years she has worked on the relations between the evolution of multi-level urban processes and dynamics in city-system networks. To study these topics comparatively, she has built many databases on large European and worldwide cities and the networks they form. In particular, she has dealt since 1990 with databases on multinational firm networks and on city properties and evolution in a multi-dimensional and long temporal approach. Diachronic and dynamic studies supply materials to develop spatial and dynamic models and visualizations. She participated in European projects like ESPON FOCI 2008-2011, FP7 FET Insite (2011-2013) and Multiplex (2012-2016). She also participates to the EuropeAid project with China MEDIUM (2015-2019) on medium size cities in China and launched two projects LOGIICCS (FNS 2015-2018) and MEDIUM (SWISS-Conf. 2016-2018) on modeling Indian and Chinese cities’ integration in global networks of multinational firms and innovation.
The SONIC Speaker Series presents
Survey Research Center at the University of Michigan’s Institute for Social Research
Social Space Diffusion: Applications of a Latent Space Model to Diffusion with Uncertain Ties
SONIC Lab is proud to welcome Dr. Jake Fisher of the University of Michigan. He will speak on Monday, April 15th, 2019 at 10 am in Frances Searle Building, SONIC Conference Room 1-459. Please contact Brent Hoagland with any questions.
Social networks represent two different facets of social life: (1) stable paths for diffusion, or the spread of something through a connected population, and (2) random draws from an underlying social space, which indicate the relative positions of the people in the network to one another. The dual nature of networks creates a challenge: if the observed network ties are a single random draw, is it realistic to expect that diffusion only follows the observed network ties? This study takes a first step toward integrating these two perspectives by introducing a social space diffusion model. In the model, network ties indicate positions in social space, and diffusion occurs proportionally to distance in social space. Practically, the simulation occurs in two parts. First, positions are estimated using a statistical model (in this example, a latent space model). Then, second, the predicted probabilities of a tie from that model—representing the distances in social space—or a series of networks drawn from those probabilities—representing routine churn in the network—are used as weights in a weighted averaging framework. Using longitudinal data from high school friendship networks, the author explores the properties of the model. The author shows that the model produces smoothed diffusion results, which predict attitudes in future waves 10 percent better than a diffusion model using the observed network and up to 5 percent better than diffusion models using alternative, non-model-based smoothing approaches.
The forthcoming paper is available online at: https://doi.org/10.1177%2F0081175018820075
The SONIC Speaker Series presents
Marketing Faculty at INSEAD
Investigating the Multiple-Source Effect in Product-Pitch Videos
SONIC Lab is proud to welcome Prof. Mukherjee of INSEAD. He will speak on Tuesday, April 2nd, 2019 at 11 aM in Frances Searle Building, Room 1-483. Please contact Brent Hoagland with any questions.
Conventional wisdom suggests that having multiple speakers (“sources”) deliver content enhances persuasion. Laboratory evidence confirms the folk knowledge and terms it the “multiple-source effect.” As the prior evidence was developed in the behavioral laboratory, it derives from relatively simple laboratory stimuli conveying information on far fewer topics than is typical in the real-world marketplace. Our study addresses this limitation. We investigate the multiple-source effect in all (more than 30,000) product-pitch videos in nine product-innovation-related categories on Kickstarter, an online crowdfunding portal, since its inception in April 2009 to mid-February 2017. We use deep-learning models to algorithmically measure the number of speakers, transcribe and analyze the spoken content, and measure other audial and visual control variables. We document a novel boundary condition of the multiple-source effect—the effect depends on the number of topics discussed in the video. In simpler videos discussing fewer topics, which is more similar to stimuli in prior laboratory studies, we corroborate prior findings that having more speakers leads to more funding. However, in more complex videos discussing more topics, we find that having more speakers does not affect funding. The latter is consistent with the literature on information overload. More broadly, our research demonstrates the potential of deep learning to enable the analysis of large-scale audio and video data in order to investigate human behavior in real-world settings.
Prof. Mukherjee is an expert in quantitative and computational marketing methodology. He develops and applies cutting-edge methods to managerially and substantively important marketing phenomena. His work has been published in prestigious journals (such as the Journal of Marketing Research, Journal of Retailing, and Management Science), featured in popular press outlets (including Forbes), and received several awards (including several best paper awards). He has been invited to give research talks at numerous prestigious universities and he consults and teaches for several major companies (such as IBM, LinkedIn, and Sony).
The SONIC Speaker Series presents
George G. Vega Yon
Department of Preventive Medicine at USC’s Keck School of Medicine
Big Problems for Small Networks: Statistical Analysis of Small Networks and Team Performance
SONIC Lab is proud to welcome George G. Vega Yon of USC’s Keck School of Medicine. George will speak on Wednesday, March 20th, 2019 at 2PM in Frances Searle Building, Room 1-483 with a workshop to follow. Please contact Brent Hoagland with any questions.
Small network data such as team, family, or personal networks, is common in many fields that study social networks. Although the analysis of small networks may appear simplistic relative to the difficulties posed by “big” datasets, there are at least two key challenges: (1) fitting statistical models to explain the network structure in small groups, and (2) testing if structural properties of small networks are associated with group-level outcomes; for example, team performance. In this presentation, we introduce two new statistical methods that use a revisited version of Exponential Random Graph Models (ERGMs) in the context of small networks. Using exhaustive enumeration of networks in the support, we are able to calculate exact likelihood functions for ERGMs, which allows us to obtain maximum likelihood estimates directly (without using simulations), avoiding common problems that arise from methods that rely on approximations instead. This is joint work with Prof. Kayla de la Haye (USC).
A workshop on the R packages ergmito and gnet for applying the methods introduced during the talk will be conducted.
George G. Vega Yon is a Biostatistics Ph.D student and Research Programmer in the Department of Preventive Medicine at USC’s Keck School of Medicine. His interests are in computational statistics and scientific software development. Most recently, his research has focused on the development of statistical methods for both phylogenetics and social network analysis. He holds a MS degree in Economics from Caltech, and a MA in Economics and Public Policy from Universidad Adolfo Ibáñez, Chile.
The SONIC Speaker Series presents
Organization and Human Resources, University of Buffalo
The Negative Side of the Social Ledger: A Meta-Analysis
SONIC Lab is proud to welcome Prasad Balkundi of the University of Buffalo. Dr. Balkundi will speak on Wednesday, April 18th, 2017 at 10 AM in Frances Searle Building, Room 1-483. Please contact Dr. Michael Schultz with any questions.
Despite a resurgence of research on negative ties in social networks, a comprehensive understanding of negative and positive has yet to be provided. Incorporating evidence from prior 163 independent samples we examine whether the initiation of positive and negative relationships (i.e., out-degree) or the reception of positive and negative relationships (i.e., in-degree) is more impactful to the focal employee’s effectiveness. Furthermore, to address the negative asymmetry hypothesis in social networks, we compare the relative importance of positive versus negative work relationships while holding the directionality constant. This meta-analytic review makes five contributions to theory on negative and positive social networks by (a) demonstrating the undermining impact of negative ties on performance and job attitudes; (b) providing information on the negative asymmetry hypothesis within social networks to reveal that negative ties occur less frequently than positive ties and that any asymmetry effects depend on the relative number of negative ties to positive ties in the context; (d) distinguishing between haters (senders of negative ties) and jerks (receivers of negative ties) to illustrate that haters have worse job attitudes than jerks, but the two do not differ on performance; and (e) providing positive and negative affect as antecedents to negative ties. Implications of these findings along with study limitations and future research directions are discussed.
Prasad Balkundi is Chair and Associate Professor in the Organization and Human Resources Department at the University of Buffalo School of Management. Prof. Balkundi’s teaching and research interests are in social networks and team processes. He has published in the Academy of Management Journal and Leadership Quarterly. He is active in the Academy of Management National Conference and has presented several papers over the years.
Stream the talk here: