About the Series
The Quarterly CS+Econ Workshop brings in three or four experts at the interface between computer science and economics to present their perspective and research on a common theme. Chicago area researchers with interest in economics and computer science are invited to attend. The technical program is in the morning and includes coffee and lunch (on your own). The afternoon of the workshop will allow for continued discussion between attendees and the speakers.
The workshop series is organized by Jason Hartline, Benjamin Golub, Annie Liang, Marciano Siniscalchi, and Alireza Tahbaz-Salehi. Funding for the series is provided by the Shaw Family Supporting Organization CS+X Fund.
Synopsis
This edition of this workshop will be on the theme of Machine Learning and Strategic Behavior. This workshop aims to combine perspectives from economics and computer science on the topics of: (a) strategic interactions between agents and machine learning algorithms, and (b) how agents’ decisions might be modeled through the lens of machine learning algorithms. The speakers are Nika Haghtalab, Jawwad Noor, Ran Spiegler, and Moshe Tennenholtz.
Logistics
- Organizers: Jason Hartline, Benjamin Golub, Annie Liang, Marciano Siniscalchi, and Alireza Tahbaz-Salehi
- Date: Monday, March 15, 2021
- Location: Virtual (on Gather.Town and Zoom).
- View the Event in Full here
Schedule
- 11:00-11:30: Ran Spiegler. View the talk here
- 11:30-12:00: Moshe Tennenholtz View the talk here
- 12:00-1:00: Lunch Break and Q&A in GatherTown
- 1:00-1:30: Nika Haghtalab. View the talk here
- 1:30-2:00: Jawwad Noor. View the talk here
- 2:00-3:00: Hangout and Q&A in GatherTown
Titles and Abstracts
Speaker: Ran Spiegler
Title: Cheating with Models
Abstract: Beliefs and decisions are often based on confronting models with data. What is the largest “fake” correlation that a misspecified model can generate, even when it passes an elementary misspecification test? We study an “analyst” who fits a model, represented by a directed acyclic graph, to an objective (multivariate) Gaussian distribution. We characterize the maximal estimated pairwise correlation for generic Gaussian objective distributions, subject to the constraint that the estimated model preserves the marginal distribution of any individual variable. As the number of model variables grows, the estimated correlation can become arbitrarily close to one, regardless of the objective correlation.(joint with Kfir Eliaz and Yair Weiss)
Speaker: Moshe Tennenholtz
Title: Data Science with Game Theory Flavor
Abstract: Design of data science algorithms and techniques, central to the Internet and on-line media, needs to be revolutionized. Current designs ignore participants’ strategic incentives. We are establishing an entirely new repertoire of incentive-compatible data science algorithms and techniques, with major applications in search and information retrieval, recommendation systems, regression, on-line learning, clustering and segmentation, and social networks analysis. In this talk I will briefly introduce our research agenda, and discuss in more detail a couple of concrete contributions.
Title: Learning and Persuading with Anecdotes
Abstract: This talk presents a model of learning and communication between two agents using hard anecdotal evidence. We use this model to shed new light on human communication and justify when and why polarization and biased belief may arise.
This talk is based on a joint work with Nicole Immorlica, Brendan Lucier, Markus Mobius, and Divyarthi Mohan.
Title: Intuitive Beliefs
Abstract: A probability measure over a multi-dimensional state space is an Intuitive Belief if it is an aggregation of pairwise associations which have been formed on the basis of past experience in the environment. Associations are shown to correspond to an analog of pointwise mutual information, and a separability property in beliefs is shown to characterize the model. The formation of associations is modelled as an extension of machine learning. Intuitive Beliefs are shown to exaggerate correlations in low probability states, exhibit the Disposition Effect documented in behavioral finance, and belief patterns observed in the psychology literature on overconfidence.