Author: William Kennedy

The Research Colloquium on Computational Social Science/Data Sciences speaker for Friday, November 01, 2019, will be Robert Axtell, Professor, Department of Computational and Data Sciences. Dr. Axtell’s talk entitled “Working with Heavy-Tailed Data: A Tutorial” will begin at 3:00 in the Center for Social Complexity Suite located on the 3rd floor of Research Hall.  The talk will be followed by a Q&A session along with light refreshments. 

This session will be live-streamed on the YouTube channel:  https://www.youtube.com/channel/UC7YCR-pBTZ_9865orDNVHNA

For announcements regarding this and future streams, please join the CSS/CDS student and alumni Facebook group: https://www.facebook.com/groups/257383120973297/

 For a list of upcoming and previous seminars, please visit:  https://cos.gmu.edu/cds/calendar/

Abstract:  This will be a hands-on talk in which Dr. Axtell will illustrate challenges and pitfalls of manipulating and statistically describing data which are extremely skew and possibly large in quantity. Motivated mainly by models from economics (e.g., firms and cities) and finance, datasets with millions of observations, both continuous and discrete, will be analyzed and plotted, with and without binning. There will be some discussion of parameter estimation but this will not be the primary focus of the talk. Heavy-tailed size distributions, the corresponding weighted-distributions, and the related ideas of moment distributions and size-biased sampling will all be discussed. Terminology such as ‘scaling’ and ‘scale-free’ will be unpacked and illustrated. The relation of Zipf-style rank-size plots to probability distributions will be developed. The competition between power law and lognormal distributions to represent heavy-tailed data will be addressed, including the ability of the lognormal to mimic a power law. Finite size effects and truncated distributions will also be discussed.

will be the presenter at this Friday’s Research Colloquium on Computational Social Science/Data Science. Dr. Anderson is a visiting assistant professor at the University of Pittsburgh’s Department of Informatics and Networked Systems in the School of Computing and Information. She uses the tools of network analysis and computational modeling to look at how skills and ideas interact in collaborative environments.

Her talk entitled “Skill Networks and Measures of Complex Human Capital” will begin at 3:00 in the Center for Social Complexity Suite located on the 3rd floor of Research Hall.  The talk will be followed by a Q&A session along with light refreshments. 

For a list of upcoming and previous seminars, please visit:  https://cos.gmu.edu/cds/calendar/

 Abstract:  The relationship between worker skills and wages is a problem of tremendous economic interest, making it critical to have effective measures of the skills, knowledge, and experience that a worker brings to production: a bundle of worker characteristics that economists refer to as human capital. Traditional models of human capital measures either divide workers into broad categories (e.g., laborers and management) or treat skills as a uni-dimensional measure of speed, education, or experience. However, in knowledge based production, the value a worker brings to production depends on both her individual skills and the interaction between them. Here, I present a network-based method for characterizing worker skills. I construct a human capital network, in which nodes are skills and two skills are connected if a worker has both or both are required for the same job. A worker’s human capital can be measured according to the position of her skills on the network. I illustrate this method using a novel dataset, gathered from an online freelance labor market. I show that workers with diverse skills earn higher wages than their peers with more specialized skills, and that those who use their diverse skills in combination earn the highest wages of all. I also show that network-based measures of human capital capture variation in wages beyond that captured by the skills individually. Finally, I will show how these same techniques can be used outside of the economic context, to quantify the relationship between the skills of collaborators.

The Research Colloquium on Computational Social Science/Data Sciences speakers for Friday, October 18, 2019, will be Fahad Aloraini, Computational Social Science PhD student.  Fahad’s talk entitled “Modeling Solar-Panel Technology Adoption in Austin:  A Test of the Power of Integrating GIS and Cognitive Modeling” will begin at 3:00 in the Center for Social Complexity Suite located on the 3rd floor of Research Hall. The talk will be followed by a Q&A session along with light refreshments. 

This session will be live-streamed on the YouTube channel: https://www.youtube.com/channel/UC7YCR-pBTZ_9865orDNVHNA

For announcements regarding this and future streams, please join the CSS/CDS student and alumni Facebook group: https://www.facebook.com/groups/257383120973297/

For a list of upcoming and previous seminars, please visit: https://cos.gmu.edu/cds/calendar/

Abstract:  Agent-based models are being created by researchers from increasingly diverse fields. While this wealth of different perspectives enriches the subject of these models, it nevertheless poses problems and challenges such as the depth of knowledge required to construct these models — an aspect where diversity is an issue is agent-decision-making. While there are cognitive architectures that have been validated in lab conditions against human data, they are too computationally expensive to be used for agent-based models with even a moderate number of agents. Moreover, these cognitive architectures do not specify the way to construct the environment and do not leverage the power of agent-based models. As part of my master’s project I have come up with a framework that solves these issues:  The Objects Memory Dynamics and Actions framework. The framework moves away from the use of unreliable subjective data (surveys and self-reports) to more objective forms of data (rate of exposure to objects). The framework integrates two cognitive architectures (ACT-R and Fast and Frugal) to include a computationally efficient agent.  It specifies the objective data needed for agent learning and decision-making and how to estimate the unique rate at which an agent is exposed to an object in the environment. To test the framework, I model Solar-Panel Technology adoption of households in Austin, Texas, and compare it to work done in the field.

The Research Colloquium on Computational Social Science/Data Sciences speaker for Friday, October 11, 2019, will be Neil Johnson, Professor of Physics, George Washington University. Dr. Johnson’s talk entitled “Slaying the Online Hydra of Hate, Distrust and Anti-Science” will begin at 3:00 in the Center for Social Complexity Suite located on the 3rd floor of Research Hall.  The talk will be followed by a Q&A session along with light refreshments. 

This session will be live-streamed on the YouTube channel:  https://www.youtube.com/channel/UC7YCR-pBTZ_9865orDNVHNA

For announcements regarding this and future streams, please join the CSS/CDS student and alumni Facebook group: https://www.facebook.com/groups/257383120973297/

 For a list of upcoming and previous seminars, please visit:  https://cos.gmu.edu/cds/calendar/

 Abstract:  Hate and distrust are thriving on the Internet [1]. In addition to the political world, medicine and science are also under attack. Social media platforms such as Facebook have access to state-of-the-art software tools, yet appear unable to keep it under control. In this talk, I discuss why this might be using a simple theory of the functional properties of networks-of-networks. I also show how the ’sociophysics’ of heterogeneous objects offers a fresh understanding of such systems as interacting gels in a multi-dimensional space. I then discuss how this analysis is connected to the unexpected surge in online pro-ISIS support that arose several years ago [2].
[1] N.F. Johnson et al., Hidden resilience and adaptive dynamics of the global online hate ecology, Nature, Sept. 12, 2019

[2] N.F. Johnson et al., New online ecology of adversarial aggregates: ISIS and beyond, Science 352, 1459 (2016)

Bio:  Neil Johnson is a professor of physics at George Washington University and heads up a new initiative in Complexity and Data Science which combines cross-disciplinary fundamental research with data science to attack complex real-world problems. He is a core member of GW’s new Knight Foundation-funded Institute for Data, Democracy and Politics. His research interests lie in the broad area of Complex Systems and ‘many-body’ out-of-equilibrium systems of collections of objects, ranging from crowds of particles to crowds of people and from environments as distinct as quantum information processing in nanostructures through to the online world of collective behavior on social media. He is a Fellow of the American Physical Society (APS) and is the recipient of the 2018 Burton Award from the APS. He received his BA/MA from St. John’s College, Cambridge, University of Cambridge and his Ph.D. as a Kennedy Scholar from Harvard University. He was a Research Fellow at the University of Cambridge, and later a Professor of Physics at the University of Oxford until 2007, having joined the faculty in 1992. Following a period as Professor of Physics at the University of Miami, he was appointed Professor of Physics at George Washington University in 2018. He presented the Royal Institution Christmas Lectures “Arrows of Time” on BBC TV in 1999. He co-founded and co-directed CABDyN (Complex Agent-Based Dynamical Systems) which is Oxford University’s interdisciplinary research center in Complexity Science, and an Oxford University interdisciplinary research center in financial complexity (OCCF).

The Research Colloquium on Computational Social Science/Data Sciences speakers for Friday, October 04, 2019, will be Amira Al-Khulaidy and Valentin Vergara, CSS PhD students. Their talk, entitled “Corruption and the effects of influence within social networks: An agent-based model of the “Lava Jato” scandal,” will begin at 3:00 in the Center for Social Complexity Suite located on the 3rd floor of Research Hall. The talk will be followed by a Q&A session along with light refreshments.

This session will not be video-streamed.

For announcements regarding future streams, please join the CSS/CDS student and alumni Facebook group: https://www.facebook.com/groups/257383120973297/

For a list of upcoming and previous seminars, please visit: https://cos.gmu.edu/cds/calendar/

We hope to see you on Friday, October 04, 2019 at 3PM.

Abstract: Corruption, and more specifically corruption in Latin America, is a complex phenomenon which is affected by politics, social structures and institutions, as well as individual behaviors. The Lava Jato is a large-scale example of corruption in Brazil. As of early 2019, there is still an ongoing investigation surrounding what has been lauded as the largest corruption scheme in Latin America. Advances in data analysis, computation, and social networks have allowed progress to be made with these types of investigations. The Lava Jato case has been a clear example of how breaking up social networks and understanding the extent of crime and individual corruption has revealed webs of corruption that have influenced politics, as well as hindered economic developed in Brazil. The several layers of interactions between individuals and institutions can be difficult to grasp and understanding the patterns and relationships within complex large-scale phenomena, such as corruption can seem impossible. Agent-based models can help with understanding these complex behaviors and systems. By capturing the patterns and gaining a better understanding of how corruption emerges and is manifested, we can help inform policy, as well as create better tools and methods for crime prevention and detection.

Bios:
Amira Al-Khulaidy – Computational Social Science PhD student, MEd. in Education, University of Virginia.
Valentin Vergara – Computational Social Science PhD student, MSc. Natural Resource Economics, Universidad de Concepción, Chile.