CCI Seminar Series, 2024-25

Panos Iperiotis, New York University, Algorithmic Hiring and Diversity: Reducing Human-Algorithm Correlation for Better Outcomes
Monday, May 5, 2025, 12:00pm-1:00pm, MIT Building E62, Room 446
 
Algorithmic tools are increasingly used in hiring to improve fairness and diversity, often by enforcing constraints such as gender-balanced candidate shortlists. However, we show theoretically and empirically that enforcing equal representation at the shortlist stage does not necessarily translate into more diverse final hires. We identify a crucial factor influencing this outcome: the correlation between the algorithm’s screening criteria and the human hiring manager’s evaluation criteria. Using a large-scale empirical analysis of nearly 800,000 job applications across multiple technology firms, we find that enforcing equal shortlists yields limited improvements in gender diversity when the algorithmic screening closely mirrors the hiring manager’s preferences. We propose a complementary algorithmic approach designed explicitly to diversify shortlists by selecting candidates likely to be overlooked by managers, yet competitive by managers’ criteria. Empirical simulations show that this approach significantly enhances gender diversity in final hires without compromising candidate quality. These findings highlight the importance of algorithmic design choices in achieving organizational diversity goals and provide actionable guidance for practitioners implementing fairness-oriented hiring algorithms.
 
Panos Ipeirotis is a Professor of Data Science and Information Systems and the David Margolis Teaching Excellence Faculty Fellow in the Department of Technology, Operations, and Statistics at the Leonard N. Stern School of Business at New York University. He received the Lagrange Award in 2015 for his contributions to the fields of crowdsourcing and social media analysis. He is also the recipient of the test-of-time award from KDD in 2020 for his work in crowdsourcing. He has received more than ten best paper awards and is also a recipient of a CAREER award from the National Science Foundation. He has been profiled twice in BusinessWeek, first as an innovator in crowdsourcing, then as the “data dude” in a business school. He was a research scientist at Google and worked with Facebook on their content moderation systems in the early 2010s. He has also started and participated in multiple startups, including Integral Ad Science, Tagasauris, UpWork, and others. His latest startup, Detectica, which built analytics and AI solutions for electronic communication monitoring, was acquired by Compass; while at Compass, the Detectica team created a recommendation system that generated more than $500M for Compass over the last three years. Currently, Panos is a Visiting Research Scientist at Meta Reality Labs, where he is working on transitioning various machine learning algorithms for the Reality Labs devices from the lab to “in the wild” settings.

Steven Dow, UC San Diego, Cultivating Ecosystems of Intelligence for Community-Driven Creativity
Tuesday, April 22, 2025, 12:00 PM-1:00 PM, MIT Building E62, Room 446
Zoom: https://mit.zoom.us/j/93838521414?pwd=5m5ob5lbXwaYMeYuOJSQT5jqYbkncI.1

Humanity’s symbiotic relationship with technology has provided unprecedented access to collective wisdom. Through our devices, the Internet, and now AI, humans have never been more capable. However, smart individuals alone are not enough to solve the wicked problems of the world; we need collective creativity and decision-making. This talk explores new opportunities for AI systems to support the social dynamics of problem-solving — to cultivate ecosystems of intelligence. I will share our work from the Protolab research group on how human-AI interactions can be designed to make individuals, teams, and communities more adaptive, empowered, and wiser.

Steven Dow is a Professor in the Cognitive Science Department and the Design Lab at UC San Diego, Director of the ProtoLab research group, and Co-Founder of the Design for San Diego initiative. His research on human-computer interaction, creativity, social computing and collective intelligence seeks to engage diverse teams and communities to co-create better, more inclusive, and more sustainable outcomes. Prof. Dow received the NSF CAREER Award for research on “advancing collective innovation.” His research has been funded by multiple National Science Foundation grants, a Google Faculty Grant, a Yankelovich Center for Social Science Research Award, Stanford’s Postdoctoral Research Award, and a Hasso Plattner Design Thinking Research Grant. He holds an MS-HCI and PhD in Human-Centered Computing from the Georgia Institute of Technology and a BS in Industrial Engineering from the University of Iowa.

Vicky Yang, MIT, How social influence shapes collective intelligence (or stupidity): From labs to the real world
March 17, 2025, 12:00 PM-1:00 PM, MIT Building E62, Room 446

Understanding when groups make good or bad collective decisions is a key societal concern, with social influence playing a crucial role. Yet, its effects remain debated in experiments—some studies find social influence improves accuracy, others show it harms, and some suggest it depends on context. We reconcile these conflicting findings in binary choice tasks with a simple mathematical model of individuals integrating independent judgment and social information. The model predicts a bifurcation, where group decisions can diverge into two distinct outcomes. Analyzing four published experiments, we show this bifurcation is already present in prior results. We show disparate experimental findings in the literature can be reproduced by the same model, and identify conditions where social influence is expected to help or hinder collective accuracy. I will also discuss preliminary work extending this model to help understand opinion dynamics applicable to the real world, particularly in the formation of partisan attitudes during crises such as COVID-19 and climate change.

Vicky Chuqiao Yang is the Richard S. Leghorn (1939) Career Development Assistant Professor in Management of Technological Innovation and an Assistant Professor of System Dynamics at the MIT Sloan School of Management. She was most recently an Omidyar Fellow at the Santa Fe Institute. Yang uses quantitative behavioral models, assisted by the analysis of data, to study collective human behavior on a broad range of organization levels, from teams to cities. Recent research topics include collective decision-making, political polarization, scaling laws in cities, and bureaucracy in human organizations. She received a PhD in Applied Mathematics from Northwestern University.

Amir Goldberg, Stanford, Not All Distances Were Created Equal, Or Why Culture Is Not Exactly Geometric
February 24, 2025, 12:00-1:00, MIT Building E62 Room 446

Social scientists often use spatial metaphors to describe social reality. The distances between coordinates—representing people, ideas, organizations—in such spatial apparatuses presumably correspond to socially meaningful differences between them. The advent of computational text methods, specifically word embedding models, catalyzed analyses of culture as a spatial geometry. Yet, the distances between words inferred from computational methods do not correspond to how people perceive their semantic properties. We analyze heterogeneity in people’s interpretations of a variety of cultural objects, and distinguish between different interpretations that are inconsistent with one another (which we term oppositional interpretations) and those that are divergent but not inconsistent (which we term orthogonal interpretations). Using a combination of surveys, free text storytelling, and computational analyses, we demonstrate that geometric representations of cultural space overestimate orthogonal interpretative distance, and underestimate oppositional interpretative distance. Moreover, we show that oppositionality and orthogonality have different sociological implications. People perceive others with opposed interpretations more negatively than those whose interpretations are orthogonal to theirs, and are less inclined to engage with them socially.

Amir Goldberg is a Professor of Organizational Behavior and (by courtesy) Sociology at the Stanford Graduate School of Business, where he is the founding co-director of the Computational Culture Lab. His research uses computational methods, specifically machine learning and large language models, to measure and model cultural dynamics in organizations, as well as in broader social contexts. His work has been published in leading journals such as the American Journal of Sociology, the American Sociological Review, and Administrative Science Quarterly. He currently serves as Department Editor for the Organizations Department in Management Science.