Repository for the Winter 2024 Computational Social Science Workshop
Time: 11:00 AM to 12:20 PM, Thursdays Location: 1155 E. 60th Street, Chicago IL 60637; Room 295
Sign up to meet over lunch, dinner, or small group settings with our speakers here
2/29 Jeremy Birnholtz, is a Professor of Communication Studies at Northwestern with research focused on the mechanisms and dynamics of attention and self-presentation in online environments. Affiliated with the Social Media Lab, HCI+D(esign) Center, and the Institute for Sexual and Gender Minority Health and Wellbeing, he has published widely on subjects that include LGBTQ+ experiences online, young people’s attention to instant messaging, and deception in text messaging. Jeremy has served as Visiting Professor at Facebook.
Self-Presentation in Sociotechnical Life: How We Present Ourselves to Each Other in a World of Digital Platforms. Self-presentation, rooted in Goffman’s classic work, is the fundamental social process by which people shape their public personas and play the social roles (e.g., teacher, student, lesbian, doctor, etc.) that structure our everyday interactions. Today’s social platforms and communication technologies, however, complicate this process in ways that Goffman could never have anticipated. Specifically, the “physics” of how information moves in the environment have changed and can vary widely from platform to platform. And we lack a systematic framework for discussing these differences and how people cope with them (and their consequences). In this talk, I will discuss a book manuscript Michael Ann DeVito and I are working on to address this gap. I will give an overview of the framework, our focus in particular on LGBTQ+ populations, and how we can use this work to better understand and describe important social behavior in a range of online contexts. Paper available here.
Pose your questions here!
2/22 Yingdan Lu, is an Assistant Professor in the Department of Communication Studies at Northwestern University. Her research centers on digital technology, political communication, and authoritarian politics. In particular, her research explores: How do authoritarian governments (like China) use digital media and AI strategically to maintain their rule and what are the consequences? How do individual experiences vary with digital technology in different media environments and what are their impacts?
Narratives of Foreign Media Ecosystems in Chinese Social Media Discussions of the Russo-Ukrainian War How do audiences living in countries with strong government censorship learn about foreign news? The prevailing expectation is that these governments act as the sole gatekeepers of information about foreign affairs. This talk presents an alternative perspective – the construction of foreign political news results from complex, transnational assemblages of online and traditional media, even in seemingly “closed” information systems such as China. However, under conditions of stringent censorship, transnational assemblages reflect pre-existing geopolitical alignments. By comparing narratives about the 2022 Russo-Ukrainian War circulating on Chinese social media with narratives found in over 24 million articles from 10,000 Chinese, Russian, Ukrainian, and U.S.-based news websites, we find, in alignment with our theoretical expectation, that Russian news websites are the largest originators and influencers of narratives related to the Russo-Ukrainian War found on Weibo, followed closely by Ukrainian news websites and more distantly by Chinese and U.S. news websites. These results show that transnational assemblages, rather than the Chinese government alone, shapes the inflow of information and construction of foreign news in a system of stringent censorship. This paper provides content-based computational frameworks for identifying multilingual, cross-country, and cross-platform digital communication, shedding light on the consequences of information control to the global information ecosystem. Paper in invitation email (currently under review).
Pose your questions here!
2/15 Sandra Gonzáles-Bailón, Carolyn Marvin Professor of Communication and Sociology, Annenberg School of Communication, University of Pennsylvania; Director, Center for Information Networks and Democracy, where she teaches and researches about communication networks and how they shape exposure to information through the consumption of news and political content and the coordination of information campaigns.
The Diffusion and Reach of Information on Social Media. Social media create the possibility for rapid, viral spread of content. We analyze the virality of and exposure to information on Facebook during the US 2020 Presidential election by examining the diffusion trees of the approximately 1B posts that were reshared at least once by US-based adults (from July 1, 2020 to February 1, 2021). Only N ∼ 12.1 million posts (1.2%) were reshared more than 100 times, involving N ∼ 114 million adult U.S. users and accumulating ∼ 55% of all views. We differentiate broadcasting versus peer-to-peer diffusion to show that: (1) Facebook is predominantly a broadcasting (rather than viral) medium of exposure; (2) Pages (not Groups) are the key engine for high-reach broadcasting; (3) misinformation (as identified by Meta’s Third Party Fact Checkers) reverses these trends: this type of content relies on viral spread through long, narrow, and slower chains of resharing activity; and (4) a very small minority of users (older and more conservative) power the spread of misinformation, triggering very deep cascades that accumulate large numbers of views. The preprint paper will be distributed by email and is not to be shared online.
Pose your questions here!
2/8 Hyejin Youn, Associate Professor of Management and Organizations at Northwestern University, where she teaches about social, technological, and organizational innovation.
Geometrics of the Adjacent Possibles: Harvesting Values at the Curvature. Novelty is not a sufficient condition for innovation. For new ideas and products to succeed, they must be integrated into the shared knowledge and pre-existing systems. Here, we develop a quantitative framework to assess how past innovations curve search trajectories in the space of adjacent possible, illustrating how the past determines the future. When certain technological building blocks begin to coalesce into noticeable clusters upon frequent combinations, a set of these typical combinations acts as nascent stages of domains in information infrastructure, thus wielding power on the next innovation. The power of typically bends the trajectory of exploration towards typical combinations much like a gravitational force on new ideas and actions. We demonstrate these curvatures are not merely abstract concepts but empirically measurable quantities. For instance, Edison’s inventions are seen in areas of high curvature, echoing his well-known design strategy of leveraging institutionalized domains, while Tesla’s inventions are predominantly located in low-curvature areas, indicating exploration of new territories. This curvature represents the power of typicality from accumulated pasts, guiding search toward well-established, paved paths by compressing the exploration space around the accumulated knowledge repertoire of proven solutions, explaining why the most commercially successful inventions often emerge at the fringes of established domains. Our further analysis of the entire U.S. patents reveals that innovations in areas of high curvature are indeed more likely to harvest monetary values. Our framework provides insights into how new ideas interact with and evolve alongside established structures in both institutional and collective understanding, illustrating the complex dialogue between innovation and convention. Here is the complete paper.
Pose your questions here!
2/1 Ágnes Horvát, Professor of Communication and Computer Science at Northwestern University, where she directs the Technology and Social Behavior PhD program.
How science is (mis)communicated in online media. Most academics are promoting their work online. At the same time, the public, journalists, and interested governments increasingly turn to online platforms for scientific information. It thus becomes ever more critical that we better understand how science is disseminated in online news, social media, blogs and knowledge repositories. My talk will summarize our work about (1) how subsequently retracted articles receive outsized attention online, (2) how scientific publications spread on various types of online platforms, losing essential information, and (2) how gender impacts the coverage of scholarship. Our findings highlight detrimental heterogeneities in online science sharing. They inform efforts to curb the online spread of science-related misinformation and close gaps in scholars’ visibility. Consider the following 4 papers: 1, 2, 3, 4.
Pose your questions here!
1/25 Daniel Alabi, Junior Fellow, Simons Society of Fellows Postdoctoral Researcher, Computer Science and Data Science, Columbia University
"Private Linear Regression". Some researchers and practitioners (including social scientists, political scientists, economists, and healthcare scientists) crucially rely on statistical methods to further the study of individuals, society, and human behavior via inferential analysis. Unfortunately, the naive application and release of resulting statistics could reveal sensitive information about the individuals in the dataset. As a result, differential privacy (DP) has been proposed as a strong definition of privacy that can prevent reconstruction, membership, and inference attacks. However, we are still in the nascent stages of understanding the statistical validity of DP methods. One of the most fundamental statistical techniques is linear regression, commonly used for predicting, testing, and augmenting relationships between two or more variables. In this talk, I will discuss my algorithmic and analytic contributions to private linear regression, showcasing the capabilities and limits of such methods in a range of settings.
The following papers are associated with the talk. 1) https://jmlr.org/papers/v24/23-0045.html (JMLR 2023): focuses on the problem of private hypothesis testing with applications to predicting social and economic mobility; 2) https://dash.harvard.edu/handle/1/37373632 (PhD Dissertation, 2022): Chapters 1 and 2 contains basics for differential privacy, which should provide enough background of what differential privacy is and the contexts in which it can be applied; 3) https://petsymposium.org/popets/2022/popets-2022-0041.php (PoPETS 2022): focuses on the problem of private estimation with a focus on the Opportunity Atlas tool.
Pose your questions here!
1/18 Doug Gilbeault, University of California at Berkeley, Haas School of Business, Management of Organizations
"Online Images Amplify Gender Bias". This talk/paper, forthcoming at Nature has the distinction of being significantly co-authored by a MACSS alumnus, Bhargav Srinivasa Desikan! Because its under embargo, I'll send the paper to student via the listserve.
Each year, people spend less time reading and more time viewing images, which are proliferating online. Images from platforms like Google and Wikipedia are downloaded by millions every day, and millions more are interacting via social media like Instagram and TikTok that primarily consist of exchanging visual content. In parallel, news agencies and digital advertisers are increasingly capturing attention online through the use of images, which people process more quickly, implicitly, and memorably than text. Here, we show that the rise of images online significantly exacerbates gender bias, both in its statistical prevalence and its psychological impact. We examine the gender associations of 3,495 social categories (such as nurse or banker) in over one million images from Google, Wikipedia, and IMDb, as well as in billions of words from these platforms. We find that gender bias is consistently more prevalent in images than text for both female- and male-typed categories. We further show that the documented underrepresentation of women online is substantially worse in images compared to not only text, but also public opinion and US census data. Finally, we conducted a nationally representative, pre-registered experiment which shows that googling for images rather than textual descriptions of occupations amplifies gender bias in participants’ beliefs. Addressing the societal impact of this large-scale shift toward visual communication will be essential for developing a fair and inclusive future for the internet.
Pose your questions here!
Future weeks:
[Hyejin Youn], Associate Professor of Management & Organization Department at the Kellogg School of Management, and a core faculty at NICO, the Northwestern Institute on Complex Systems.