CCI Seminar Series, 2023-24

Abraham Bernstein, University of Zürich, Escaping the Echo Chamber: The Quest for the Normative News Recommender System and a new notion of Computer Science / Information Systems Research
June 25, 2024, 12:00-1:00, MIT Building E62 Room 687
Zoom recording: https://mitsloan.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=d205066c-947a-4d89-8edb-b19a010caf6b

Recommender systems and social networks are often faulted to be the cause for creating Echo Chambers – environments where people mostly encounter news that match their previous choices or those that are popular among similar users, resulting in their isolation inside familiar but insulated information silos. Echo chambers, in turn, have been attributed to be one cause for the polarization of society, which leads to the increased difficulty to promote tolerance, build consensus, and forge compromises. To escape these echo chambers, we propose to change the focus of recommender systems from optimizing prediction accuracy only to considering measures for social cohesion. This proposition raises questions in three spheres: In the technical sphere, we need to investigate how to build “socially considerate” recommender systems. To that end, we develop a novel recommendation framework with the goal of improving information diversity using a modified random walk exploration of the user-item graph. In the social sphere, we need to investigate if the adapted recommender systems have the desired effect. To that end, we present an empirical pilot study that exposed users to various sets (some diverse) of news with surprising results. Finally, in the normative sphere, these studies raise the question what kind of diversity is desirable for the functioning of democracy. Reflecting the consequences of these findings for our discipline, this talk highlights that the computer science and information systems disciplines increasingly need to engage with social and normative challenges as well as technical ones, possibly producing a new understanding of our disciplines. It proposes similar consequences for other disciplines in that they increasingly need to embrace all three spheres.

Abraham Bernstein is a Full Professor at the Department of Informatics (Institut für Informatik) at the University of Zurich. He conducts research on the Semantic Web and Knowledge Discovery. His work draws from social science (organizational psychology/sociology) and technical (computer science, artificial intelligence) foundations.

For more information, contact Rina Thanasi at ethanasi@mit.edu

Brian Uzzi, Northwestern University, Human and Machine Conditions Favoring Innovation in Science
May 2, 2024, 1:00-2:00, MIT Building E62, Room 446

Innovation involves generative and communicative processes. Generative processes involve the setup conditions associated with innovativeness such as teams, networks, and funding. While we know much about generative processes, we know little about how new AI tools can benefit innovation. In study 1, I use artificial intelligence to predict when innovative ideas fail. We estimate a paper’s replicability using ground truth data, and then test the model on extensive out-of-sample studies.The model predicts replicability better than reviewer base rates and on par with prediction markets. We then conduct a discipline-wide census of replicability in psychology over the last 20 years. We find that a study’s replicability is predicted by its authors’ characteristics, subfield, methods, and media coverage. In Study 2, I address the communication problem, the neglected sibling of generative processes, which focuses on how to effectively communicate the merits of innovative scientific ideas. We examine links between “scientific promotional language” and an innovative idea’s funding potential, innovativeness, and impact. We analyze over 12K funded and unfunded grants from three prominent funding agencies worldwide. Our analyses demonstrate a robust relationship between promotional language and the communication of scientific innovation. Promotional language predicts the funding decision with up to a four-fold increase in the probability of receiving funding and predicts a grant’s inherent innovativeness and citation impact. Computer experiments suggest that promotional words can change and impact a reviewers’ cognitive activation and impressions of a proposal’s quality.

Wu, Youyou, Yang Yang and Brian Uzzi. A discipline-wide investigation of the replicability of Psychology papers over the past two decades, PNAS, 120 (6) e2208863120, January 30, 2023.

Millar et al. Trends in the use of Promotional Language. JAMA, August 25, 2022.

Brian Uzzi is the Richard L. Thomas Professor of Leadership at the Kellogg School of Management, Northwestern University. He also co-directs the Northwestern Institute on Complex Systems (NICO) and the Ryan Institute on Complexity, and has professorships in Sociology and the McCormick School of Engineering. His research uses social network science and computational methods to explain outstanding human achievement and has received 32 teaching and research prizes in social, ecological, physical, and computer sciences. https://www.kellogg.northwestern.edu/faculty/uzzi/htm/

Anita Woolley, Carnegie Mellon University, Teaching Algorithms to Facilitate Collective Intelligence
April 17, 2024, 12:00-1:00, MIT Building E62, Room 687
Zoom recording here

Organizations are operating in increasingly complex environments, and the traditional ways of collaborating need to be adapted. Research over the past century on intelligence has identified the memory, attention and reasoning functions that enable a system to adapt and solve problems in a wide range of environments. My colleagues and I have built on this work by identifying how collective memory, attention and reasoning develop to enable collective intelligence in human systems, and I will present our recent work focused on developing AI to diagnose and strengthen these functions to enhance collective intelligence.

Anita Williams Woolley is the Associate Dean of Research and a Professor of Organizational Behavior and Theory at Carnegie Mellon University’s Tepper School of Business. Dr. Woolley received her doctorate in organizational behavior from Harvard University, and she is currently the PI on a number of projects funded by DARPA, NSF, ARO and AFRL on the development of collective intelligence in human and human-machine collaboration, and the ways that AI can enhance it. She is a Senior Editor at Organization Science and a founding Associate Editor of Collective Intelligence. For more, see https://sites.google.com/view/anitawoolley-com/home.

Michael Bernstein, Stanford University, Generative Agents: Interactive Simulacra of Human Behavior
March 5, 2024, 4:00-5:00, MIT Building 32, Room D463
Co-sponsored by CCI and the MIT Computer Science and Artificial Intelligence Lab (CSAIL)
Zoom recording of this seminar

Believable proxies of human attitudes and behavior can empower applications ranging from immersive environments to social policy interventions. However, the last quarter century has seen a slow recession of human behavioral simulation as a method, in part because traditional simulations have been unable to capture the complexity and contingency of human behavior. I argue that modern artificial intelligence models allow us to re-examine this limitation. I make my case through generative agents: computational software agents that simulate believable human behavior. Generative agents enable us to populate an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty-five agents using natural language. Our generative agent architecture empowers agents to remember, reflect, and plan — enabling them to act in ways reflective of their jobs and personalities, notice and remember each other, and even plan coordinated events. Extending this line of argument, I explore how proxying human behavior and attitudes can help us design more effective online social spaces, understand the societal disagreement underlying modern AI models, and better embed societal values into our algorithms.

Michael Bernstein is an Associate Professor of Computer Science at Stanford University, where he is a Bass University Fellow. His research focuses on human-computer interaction and social computing systems. This research has been reported in venues such as The New York Times, Wired, Science, and Nature, and Michael has been recognized with an Alfred P. Sloan Fellowship, UIST Lasting Impact Award, and the Computer History Museum’s Patrick J. McGovern Tech for Humanity Prize. He holds a bachelor’s degree in Symbolic Systems from Stanford University, as well as a master’s degree and a Ph.D. in Computer Science from MIT. For more, see https://hci.stanford.edu/msb/.

Scott Page, University of Michigan, Institutional Ensembles and Cultural Institutional Capacity
October 13, 2023, 1:00-2:30, MIT Building E62, Room 450
Zoom https://mit.zoom.us/j/98105285349 (Password: SDFA23)

We construct a series of models within a systems framework to analyze the interdependence between a society’s composition of institutions and its cultural-institutional capacity: the knowledge, behaviors, beliefs, norms, and networks that enables institutions to operate. In our models, a society selects a mixture of institutions of various types to allocate resources and take actions. These include markets, hierarchies, democracies, community-based institutions, or even algorithms. These institutional choices contribute to the production of cultural-institutional capacity, and, conversely, cultural-institutional capacity influences how well each institutional type performs. Cultural-institutional capacity building can be self-reinforcing. Markets can produce greater capacity for markets. It can also be generic and improve all institutional types. Neither of these forms of capacity building necessarily produces efficient ensembles of institutions. Paradoxically, systems with both forms can result in the collapse of an institutional type that builds generic capacity.

Scott Page is the John Seely Brown Distinguished University Professor of Complexity, Social Science, and Management at the University of Michigan, and the Williamson family Professor of Business Administration, professor of management and organizations, Stephen M. Ross School of Business. His research focuses on the myriad roles that diversity plays in complex systems. For example, how does diversity arise? Does diversity make a system more productive? How does diversity impact robustness? Does it make a system prone to large events? He has written five books and published papers in a variety of disciplines including economics, political science, computer science, management, physics, public health, geography, urban planning, engineering, and history. See more at https://sites.lsa.umich.edu/scottepage/bio/.

 

Deborah Gordon, Stanford University, The Ecology of Collective Behavior
October 30, 2023, 12:00-1:00, MIT Building E62, Room 446

Many natural systems, from brains to ant colonies, operate without central control, using networks of interactions that in the aggregate allow the system to adjust to the current situation. Examples from two ant species, harvester ants in the desert and turtle ants in the tropical forest, show how the dynamics of collective behavior fit the dynamics of the environment. Harvester ant colonies in the desert, in a stable but harsh environment, regulate foraging activity slowly, using centralized information flow with low modularity, and feedback in which the default is not to forage, and stimulation is needed activate foraging. Turtle ants form trail networks in the canopy of the tropical forest. In an unstable but humid environment, where activity is easy, the trail is regulated locally, at each node in the vegetation, with highly modular search that fits the modular distribution of resources. The feedback regime is set with the default to go unless inhibited. Their collective behavior responds rapidly to frequently changing conditions and resources. The talk will discuss broad analogies with the ways that rates, feedback regimes and modularity of interaction networks are used in other forms of collective behavior.

Deborah M. Gordon received her PhD from Duke University, then did postdoctoral research in the Harvard Society of Fellows, at Oxford University, and at the Centre for Population Biology at the University of London, and joined the faculty at Stanford in 1991. She is the author of three books, Ants at Work (Norton 2000); Ant Encounters: Interaction Networks and Colony Behavior (Primers in Complex Systems, Princeton University Press, 2010), and The Ecology of Collective Behavior (2023, Princeton University Press). Her awards include a Guggenheim Fellowship, fellowships at the Center for Advanced Study in Behavioral Sciences, and the Quest award of the Animal Behavior Society. Her lab studies collective behavior in ants. More broadly, she is interested in bridging insights from different disciplines that study dynamical systems and feedback control circuits, ranging across mathematical modeling, ecology, evolutionary biology, and neuroscience. http://www.stanford.edu/~dmgordon/