Repository for the Winter 2025 Computational Social Science Workshop
Time: 11:00 AM to 12:20 PM, Thursdays Location: 1155 E. 60th Street, Chicago IL 60637; Room 295
Benjamin Golub is an American economist who is a professor of economics and computer science at Northwestern University. His research focuses on the economics of networks. He was named the winner of the 2020 biannual Calvó-Armengol International Prize, which recognizes a “top researcher in [e]conomics or social sciences younger than 40 years old for contributions to the theory and comprehension of the mechanisms of social interaction.”
Incentive design with spillovers. How should contracts be designed to motivate a group of agents to work efficiently toward a risky outcome, such as a scientific advance? This is a fundamental question about optimal contracting for teamwork, and the framework to formalize it is nearly 50 years old. Nevertheless, economics has made little progress on this problem for general team production technologies. We solve the problem by combining insights from contract theory and the theory of network games. The talk will also discuss using recent advances in large-matrix statistics to address realistic cases in which the network of collaborations and spillovers is not known to the designer, and which offer a variety of open avenues for both theoretical and empirical research. The talk will be accessible to an interdisciplinary social science audience.
Reading List
Patrick Park is a computational social scientist whose research delves into the structure and evolution of large-scale social networks. His work examines how people form and sustain social ties and how broader social, economic, and environmental contexts influence this process. His research spans several thematic intersections, including social contagion, economic sociology, social psychology, the diffusion of innovation, and social movements, all supported by empirical data capturing population-scale online social interactions. Patrick actively seeks interdisciplinary collaboration with researchers from diverse fields, including computer science, applied mathematics, social sciences, and management science. His research has been published in journals such as Science, Social Networks, PLoS One, Lecture Notes in Computer Science, and Big Data and Society.
Currently, Patrick is a tenure-track assistant professor in the Software and Societal Systems Department (S3D) at Carnegie Mellon University's School of Computer Science. Prior to joining CMU, he was a postdoctoral fellow at the University of Michigan and Northwestern University. He earned his doctoral degree in sociology from Cornell University.
The Centre Cannot Hold: Online Polarization and the Paradox of Network Diversity. An influential network-based explanation for the growing political polarization observed in social media platforms focuses on the self-reinforcing opinion dynamics where social influence processes lead to the formation of two opposing ideological camps. This explanation overly focuses on the tail ends of the ideological spectrum, thereby overlooking the social processes that describe the moderate majority who are not necessarily a susceptible herd of politically apathetic, uninformed bystanders. How, then, does online polarization intensify despite the existence of this moderate majority? In this study, I develop an explanation that focuses on the amplified relational constraints experienced by users with diverse social networks in highly visible, open communication environments that strip away contextual information. Using Twitter communication data of 26M U.S. Twitter users, I show that users with highly diverse communication networks tend to delete tweets with higher frequency, more so when their network neighbors exhibit wider cognitive distance. These findings lend support to the idea that observed opinion polarization in social media could result from the relative invisibility of balanced opinions that relationally constrained network brokers are positioned to form and disseminate.
Reading List
- Park, P. S. (n.d.). The Centre Cannot Hold: Online Polarization and the Paradox of Network Diversity. Carnegie Mellon University. (Unpublished Manuscript, see email attachment).
- Park, P. S., Blumenstock, J. E., & Macy, M. W. (2018). The strength of long-range ties in population-scale social networks. Science, 362(6421), 1410-1413. https://doi.org/10.1126/science.aau9735
Dr. Hirokazu Shirado is an Assistant Professor of the Human-Computer Interaction Institute in the School of Computer Science at Carnegie Mellon University. His research focuses on hybrid systems of humans and machines, particularly how machine intelligence can help people address collective action challenges. He leads the Computation & Collective Lab (C&C Lab) at the Human-Computer Interaction Institute, Carnegie Mellon University. C&C Lab aims to explore how machine intelligence can aid human cooperation and coordination in hybrid systems of humans and machines. They also strive to expand the boundaries of computation social science and social computing by unlocking the potential of computational social experiments.
He obtained his doctorate in Sociology at Yale University in 2019, where he was a member of the Human Nature Lab at the Yale Institute for Network Science. Before Yale, he spent eight years as an engineering researcher at Sony Corporation.
AI-Powered Individualism and Relational AI. Artificial intelligence is increasingly shaping human collective action, reshaping prosocial norms like reciprocity. In this talk, I present findings from a cyber-physical experiment with 300 participants (150 dyads) coordinating robotic vehicles in a game. Our results reveal that active driving assistance shifts users’ focus toward self-interest at the expense of reciprocity—a pattern also observed in other social contexts. To address these challenges, we are developing “relational AI,” a novel concept designed to prioritize relational contexts alongside individual preferences in AI customization. This approach has shown promise in fostering intergroup communication and cooperation, with ongoing efforts to expand its application across diverse social settings.
Reading List
- H. Shirado, S. Kasahara, & N.A. Christakis, Emergence and collapse of reciprocity in semiautomatic driving coordination experiments with humans, Proc. Natl. Acad. Sci. U.S.A. 120 (51) e2307804120, https://doi.org/10.1073/pnas.2307804120 (2023).
Dr. Jindong Wang is a Tenure-Track Assistant Professor at Data Science, William & Mary. Previously, he was a Senior Researcher in Microsoft Research Asia from 2019 to 2024. His research interest includes machine learning, large language and foundation models, and AI for social science. Since 2022, he has been selected by Stanford University as one of the World’s Top 2% Scientists and one of the Most Influential AI Scholars by AMiner. He serves as the associate editor of IEEE Transactions on Neural Networks and Learning Systems (TNNLS), guest editor for ACM Transactions on Intelligent Systems and Technology (TIST), area chair for ICML, NeurIPS, ICLR, KDD, ACMMM, and ACML, SPC of IJCAI and AAAI. He has published over 60 papers with 16000+ citations (H-index 45) at leading conferences and journals such as ICML, ICLR, NeurIPS, TPAMI, IJCV etc. His research is reported by Forbes, MIT Technology Review, and other international media. He received best and outstanding papers awards at several international conferences and workshops. He published a book Introduction to Transfer Learning. He gave tutorials at IJCAI’22, WSDM’23, KDD’23, AAAI’24, and AAAI’25. He leads several impactful open-source projects, including transferlearning, PromptBench, torchSSL, and USB, which received over 16K stars on Github. He obtained his Ph.D from University of Chinese Academy of Sciences in 2019 with the excellent PhD thesis award and a bachelor’s degree from North China University of Technology in 2014.
Does AI Feel, Trade, and Discriminate? Insights in Emotion, Economics, Culture, and Bias. Generative AI has achieved impressive superhuman performance across various tasks, triggering shared interest in both AI and social science domains. From a social science perspective, what role does AI play? How can we leverage AI to accelerate social science research? What are its challenges and limitations? Can AI models truly feel, trade, and discriminate? In this talk, I will share recent investigations into four critical aspects of human society: emotion, economics, culture, and bias. I will demonstrate how we interpret model behaviors, understand their mechanisms, gain insights from their responses, and improve their performance. The ultimate goal of our AI+SocialScience research is to build a platform to comprehensively understand and co-evolve with advanced generative AI models.
Reading List
- Li, C., Teney, D., Yang, L., Wen, Q., Xie, X., & Wang, J. (2024). CulturePark: Boosting Cross-cultural Understanding in Large Language Models. arXiv preprint arXiv:2405.15145.
- Zhao, Q., Wang, J., Zhang, Y., Jin, Y., Zhu, K., Chen, H., & Xie, X. (2023). Competeai: Understanding the competition behaviors in large language model-based agents. arXiv preprint arXiv:2310.17512.
- Li, M., Chen, H., Wang, Y., Zhu, T., Zhang, W., Zhu, K., ... & Wang, J. (2025). Understanding and Mitigating the Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks. arXiv preprint arXiv:2502.04419.
Filippo Menczer is the Luddy Distinguished Professor of Informatics and Computer Science at the Indiana University Luddy School of Informatics, Computing, and Engineering, director of the Observatory on Social Media, and a member (and former director) of the Center for Complex Networks and Systems Research. He also has courtesy appointments in Cognitive Science and Physics and is a Fellow of the Center for Computer-Mediated Communication, a Senior Research Fellow of the Kinsey Institute, a Fellow at the ISI Foundation in Torino, Italy, and a Fellow of the ACM.
Research in my group, NaN, spans computational social science, network science, Web science, and data science, with a focus on analyzing and modeling the spread of information and misinformation in social networks and detecting and countering the manipulation of social media. They also study social computing, Web search and data mining, and science of science.
AI and Social Media Manipulation: The Good, the Bad and the Ugly. Disinformation can be very harmful. AI provides us with tools to mitigate these harms, such as methods to detect inauthentic accounts and coordinated information operations. Large language models (LLMs) can also help rate the reliability of information sources and aid in fact-checking tasks. However, attempts to scale up these AI-supported interventions must account for unintended consequences when people interact with AI. In some cases, fact-checking information generated by an LLM can actually decrease news discernment. In the hands of bad actors, AI can become a dangerous weapon. Aside from much-discussed deepfakes, generative AI can be used to easily and cheaply create fake but credible profiles and content at scale. These capabilities enable the infiltration and manipulation of vulnerable online communities. Given the near-impossibility of detecting AI-generated content, research is needed to develop new ways of challenging the provenance of content before wide exposure through distribution channels like social media and search engines.
Reading List (Students can decide which one or two to read):
- B.T, Truong., O.M., Allen., & F, Menczer. Account credibility inference based on news-sharing networks. https://doi.org/10.1140/epjds/s13688-024-00450-9
- M.R, DeVerna., H.Y, Yan., K, Yang, & F, Menczer. Fact-checking information from large language models can decrease headline discernment. https://doi.org/10.1073/pnas.2322823121
- B.T, Truong., X, Lou., A, Flammini., & F, Menczer. Quantifying the vulnerabilities of the online public square to adversarial manipulation tactics. https://doi.org/10.1093/pnasnexus/pgae258
- K, Yang, D, Singh., & F, Menczer. Characteristics and Prevalence of Fake Social Media Profiles with AI-generated Faces. https://doi.org/10.54501/jots.v2i4.197
- K, Yang., & F, Menczer. Anatomy of an AI-powered malicious social botnet. https://doi.org/10.51685/jqd.2024.icwsm.7
Steven Rathje, an NSF SBE postdoctoral fellow at New York University's Social Identity and Morality Lab, specializes in studying intergroup conflict, misinformation, and their intersection with digital technologies. A Gates Cambridge Scholar with a PhD from the University of Cambridge and prior studies at Stanford University, his research has been published in prestigious journals including PNAS, Nature Human Behavior, and Science Advances. His work has garnered widespread media attention from outlets like the New York Times, BBC, and CBS Sixty Minutes.
A recipient of over $2.5 million in grants from various institutions and recently named an APS "Rising Star" and Forbes 30 under 30 member, Rathje is also dedicated to science communication. He contributes to major publications like the Washington Post and the Guardian, while maintaining a significant social media presence with over one million followers on TikTok as @stevepsychology, where he shares insights about psychological science.
The Psychology of Virality. Throughout history, technologies—such as the printing press, television, social media, and AI—have transformed how information is created, consumed, and shared. In this talk, I will present a variety of interrelated studies that explore why information spreads online and offline, the consequences of our information diets, and how these phenomena interact with digital technologies such as social media and AI. I will begin by discussing a big data analysis exploring what goes “viral” on social media and why widely shared information is often not widely liked (a phenomenon I call the “paradox of virality”). Then, I will present results from digital field experiments that demonstrate that unfollowing just a few hyperpartisan social media influencers can reduce out-party animosity and improve online behavior—with effects lasting for at least six months. Afterwards, I will discuss a series of online experiments showing that incentivizing accuracy can improve people’s ability to correctly discern between true and false news—yet making partisan identity motivations salient can reduce accuracy and increase intentions to share false information. I will then discuss how I am exploring these questions on a global scale, using a 23-country social media field experiment and a 56-country analysis of social media data using large-language models. Finally, I will outline my future directions, which explore how information spreads offline, how it spread historically, how technology impacts collective behavior, and how accuracy and identity motivations influence how people interact with AI.
Reading List (Students can decide which one or two to read):
- S. Rathje, J.J. Van Bavel, S. van der Linden, Out-group animosity drives engagement on social media, Proc. Natl. Acad. Sci. U.S.A. 118 (26) e2024292118, https://doi.org/10.1073/pnas.2024292118 (2021).
- Rathje, S., Pretus, C., He, J. K., Harjani, T., Roozenbeek, J., Gray, K., … Van Bavel, J. J. (2024, October 3). Unfollowing hyperpartisan social media influencers durably reduces out-party animosity. https://doi.org/10.31234/osf.io/acbwg
- Rathje, S., Roozenbeek, J., Van Bavel, J.J. et al. Accuracy and social motivations shape judgements of (mis)information. Nat Hum Behav 7, 892–903 (2023). https://doi.org/10.1038/s41562-023-01540-w
- Rathje, S., Robertson, C., Brady, W. J., & Van Bavel, J. J. (2024). People Think That Social Media Platforms Do (but Should Not) Amplify Divisive Content. Perspectives on Psychological Science, 19(5), 781-795. https://doi-org.proxy.uchicago.edu/10.1177/17456916231190392
William Brady is an Assistant Professor of Management and Organizations. His research examines the dynamics of emotion at the social network level and their consequences for group behavior. His recent work studies how human psychology and AI-mediated social contexts interact to shape our emotions and intergroup attitudes. Combining tools of behavioral science and computational social science, his research aims to develop person-centered and design-centered interventions to improve our digital social interactions.
Professor Brady’s research has been published in leading journals such as PNAS, Nature Human Behaviour, Science Advances, and Perspectives on Psychological Science. His work has also been featured in popular press outlets, including The New York Times, BBC, Wired, and The Wall Street Journal. In recognition of his contributions, he has been selected for the Association for Psychogical Science Rising Star Award, and the SAGE Emerging Scholar award.
The dynamics of moralization on social media. In the attention economy of social media, moralized commentary appears to spread widely, attract user engagement, and be amplified by platform algorithms. This trend—if substantiated—is concerning, as moralization can deepen societal divisions and fuel polarization. In this work, we used natural language processing to investigate whether moralization has increased on social media, tracking the dynamics of moral commentary over time. Our analysis revealed that posts and topics on Twitter (now X) became increasingly semantically aligned with our validated conceptual representation of morality, demonstrating a substantial rise in moralization from 2013 to 2021. To shed light on the processes behind these trends, we examined which patterns of user behavior are associated with rising moralization. Two key patterns emerged: (1) a social learning pattern, where individuals increasingly adopted moral language in their posts and (2) a selection effect pattern, whereby users whose commentary was already moralized became disproportionately active in online discussions. We also identified a strong link between moralization and topic prevalence: as issues became more moralized, they also gained prominence in social discourse. These findings reveal how user dynamics on modern social media can create highly moralized and polarizing public discourse. Understanding this process is essential for addressing the risks of polarization and fostering healthier digital ecosystems.
Reading List:
Sarah Elwood is a Professor in the Department of Geography at the University of Washington, and a faculty affiliate of the UW’s West Coast Poverty Center, Center for Studies in Demography and Ecology, and Certificate in Public Scholarship. With Vicky Lawson, she co-directs the Relational Poverty Network (RPN), a transnational interdisciplinary community of scholars collaborating to develop conceptual frameworks, research methodologies, and pedagogies for the study of relational poverty. Elwood received her Ph.D. in Geography from the University of Minnesota and previously held faculty positions at DePaul University and the University of Arizona. Her research contributes to relational poverty studies, critical GIScience and digital geographies, visual politics and mixed methods, and urban geography. Current activities include research on poverty politics of creative activisms around homelessness, feminist and critical race theorizations of digital geographies, and a collaborative public scholarly project on horizons of critical poverty studies under emerging nationalist populisms.
Stop the Sweeps: Insurgent Knowledge Politics and the Computational City. This paper thinks through algorithmic and data-driven regulation as terrains of struggle over urban space and inhabitations, through a close reading of data praxes and politics around homelessness. With urban management practices and ideologies of good government increasingly prioritizing data-driven algorithmic processes, the computational city runs on ‘data circuits’: Mundane data collection that encodes myriad dimensions of everyday life; administrative codes that enable or prohibit particular urban inhabitations; and the actions (repair, regulation, removal, etc.) that these data and codes make possible. These data circuits function as administrative state violence enacted on unsheltered living in public space in tent encampments – rendering them visible and removable, compelling disclosure of personal data, and forcing compliance as a condition of receiving shelter or other assistance. Yet precarious communities and their allies continue to confront these forces of governance and removal through insurgent knowledge politics that circulate dramatically different possibilities for urban inhabitations calibrated around solidarity, self-determination and staying put. I develop these arguments through a close reading of the social media tactics and away-from-keys activisms of Stop the Sweeps, a loosely connected horizontal network of local collectivities fighting state-sanctioned eviction of encampments of unsheltered people in cities and towns across the US.
Reading List:
[1] Glitch epistemologies for computational cities, Agnieszka Leszczynski, and Sarah Elwood, Glitch epistemologies for computational cities, 2022. https://journals.sagepub.com/doi/full/10.1177/20438206221075714
[2] Thinking Geomedia Futures: Indigenous Futurisms, Afrofuturisms, and Counter‐Mediations of Temporality, Spatiality, and Digitality, Sarah Elwood, Media and Communication, 2024. https://www.cogitatiopress.com/mediaandcommunication/article/view/8935
Lexing Xie is a Professor of Computer Science at the Australian National University (ANU), where she leads the ANU Computational Media Lab and directs the ANU-wide Integrated AI Network. Her research spans machine learning, computational social science, and computational economics, with a particular focus on online optimization, neural networks for sequences and networks, and applied problems such as modeling popularity in social media, vision and language, decision-making by humans and machines. Her work has been supported by the Australian Research Council, CSIRO, AFOSR, and the Data to Decisions CRC. Lexing received the 2023 ARC Future Fellowship and the 2018 Chris Wallace Award for Outstanding Research. Her research has garnered seven best paper and best student paper awards at ACM and IEEE conferences between 2002 and 2019. Among her many editorial roles, she served as the inaugural Editor-in-Chief of the AAAI International Conference on Web and Social Media (ICWSM) and will be the Program Co-Chair of ACM Multimedia 2024. Prior to joining ANU, she was a Research Staff Member at the IBM T.J. Watson Research Center in New York. She holds a PhD in Electrical Engineering from Columbia University and a BS in Electrical Engineering from Tsinghua University.
Understanding Online Attention: From Items to Markets. What makes a video popular? What drives collective attention online? What are the similarities and differences between clicks and transactions in a market? This talk aims to address these three questions. First, I will discuss a physics-inspired stochastic time series model that explains and forecasts the seemingly unpredictable patterns of viewership over time. This model provides novel metrics for predicting expected popularity gains per share and assessing sensitivity to promotions. Next, I will describe new measurement studies and machine learning models that analyze how networks of online items influence each other’s attention. Finally, I will introduce a macroscopic view of attention, offering mathematical descriptions of market equilibriums and distributed optimization. These results lay the groundwork for our ongoing research into the computational view of attention markets and potential mechanisms for fostering a healthy online ecosystem. Additionally, I will present an overview of Influence Flower, an interactive web app and arXiv plugin designed for qualitatively visualizing the intellectual influence of academic entities. I suggest that this tool, along with the research on attention, raises many intriguing questions and offers insights into knowledge creation and the dynamics of attention within crowds.
Reading List:
[1] Expecting to be HIP: Hawkes Intensity Processes for Social Media Popularity, Marian-Andrei Rizoiu, Lexing Xie, Scott Sanner, Manuel Cebrian, Honglin Yu, and Pascal Van Hentenryck, World Wide Web 2017, International Conference on, 2017 https://arxiv.org/abs/1602.06033
- Paper [1] sets the problem context and presents the initial version of the model. [1b] connects it to the well-known epidemic model, [1c] presents a strict generalization, [1d] incorporates graph structure around the content.
- [1b] SIR-Hawkes: Linking Epidemic Models and Hawkes Processes to Model Diffusions in Finite Populations, Marian-Andrei Rizoiu, Swapnil Mishra, Quyu Kong, Mark Carman, and Lexing Xie, Proceedings of the 2018 World Wide Web Conference, 2018 https://arxiv.org/abs/1711.01679
- [1c] Interval-censored Hawkes processes, Marian-Andrei Rizoiu, Alexander Soen, Shidi Li, Pio Calderon, Leanne J. Dong, Aditya Krishna Menon, and Lexing Xie, Journal of Machine Learning Research, 2022 https://arxiv.org/abs/2104.07932
- [1d] Radflow: A Recurrent, Aggregated, and Decomposable Model for Networks of Time Series, Alasdair Tran, Alexander Mathews, Cheng Soon Ong, and Lexing Xie, Proceedings of The Web Conference 2021, 2021 https://arxiv.org/abs/2102.07289
[2] Stability and Efficiency of Personalised Cultural Markets, Haiqing Zhu, Yun Kuen Cheung, and Lexing Xie, The Web Conference 2023, 2023 https://dl.acm.org/doi/abs/10.1145/3543507.3583315