Date: November 16th, 2023 (12:30 pm – 1:30 pm)

Speaker: Marina Agranov

Paper Title: “Disentangling Suboptimal Updating: Complexity, Structure, and Sequencing”

Abstract: We study underlying reasons for the failure of individuals to adhere to Bayes’ rule and decompose this departure into three elements: (i) task complexity, (ii) information structure, and (iii) timing of information release. In a series of controlled experiments, we systematically alter all three elements and quantify their magnitude. We address task complexity by introducing a novel connection to nonlinear calculations required for Bayesian updating. We experimentally explore this link and find empirical support for it.

Bio: Marina Agranov is an experimental economist specializing in theory-based experiments. Marina’s recent work includes bargaining games, social learning environments, games on networks, agency problems, auctions, and individual decision-making under risk. Marina is a Full Professor of Economics at Caltech, the Research Associate at the National Bureau of Economic Research (NBER), and the Director of the Center for Theoretical and Experimental Social Sciences at Caltech (CTESS).

Date: November 2nd, 2023 (12:30 pm – 1:30 pm)

Speaker: Jimena Galindo

Paper Title: “Learning with Misspecified Models: The Case of Overestimation“

Abstract: I design a framework and a laboratory experiment that allow for the comparison of multiple theories of misspecified learning. I focus on a framework with endogenous information and a data-generating process ruled by two fundamentals: an ego-relevant parameter and a state. Within this framework, I study three forces that can lead to misspecified beliefs: initial misspecifications, learning traps, and biased updating. I find that biased updating is the main driver of misspecified beliefs in the lab. In addition, I vary the degree of ego relevance of the parameter by introducing a stereotype treatment. The data are consistent with biased updating in both cases but for different reasons: when learning about themselves, subjects attribute successes to their own ability and failures to luck. Instead, in the stereotype treatment, they compensate for initial negative biases by over-attributing positive signals to the ability of others. This tendency translates into similar observed choices but different dynamics in beliefs. 

Date: October 19th, 2023 (12:30 pm – 1:30 pm)

Speaker: WonSeok Yoo

Paper Title: “Belief Updating Under Cognitive Dissonance: An Experimental Study”

Abstract: Cognitive dissonance is the psychological tension that arises from holding two conflicting beliefs, views, or actions simultaneously. We investigate the dynamics of belief updating in the face of such cognitive dissonance, particularly within financial decision making contexts. By introducing transaction costs in a controlled laboratory setting, we are able to simulate conditions that naturally induce cognitive dissonance among participants. Within our experimental design, subjects are initially endowed with one of two potential assets at random. They are then asked to report their beliefs regarding their assigned asset after a series of signals. For particular cases, they subsequently face a decision: to either switch or keep their initial asset. Our findings present evidence that the imposition of transaction costs deters individuals from “switching” compared to Bayesian benchmark. Moreover, individuals distort their beliefs to better justify their sub-optimal decision to retain their original asset in an effort to align their beliefs with their actions and alleviate cognitive dissonance.e a causal relationship between judicial institutions and economic development.

Date: October 5th, 2023 (12:30 pm – 1:30 pm)

Speaker: Daniel Chen

Paper Title: “Data Science for Justice: Evidence from a Nationwide Randomized Experiment in Kenya”

Abstract: Can data science improve the functioning of courts, and unlock the positive effects of institutions on development? In a nationwide experiment in Kenya, we use algorithms to identify the greatest sources of court delay for each court and recommend actions. We randomly assign courts to receive no information, information, or an information and accountability intervention. Information and accountability reduces case duration by 22%. Using continuous household surveys, we find that in regions with treated courts, workers were more likely to have formal contracts and higher wages, especially in contract-intensive industries. These results demonstrate a causal relationship between judicial institutions and economic development.

Bio:

Daniel Li Chen is Director of Research at the CNRS and Professor at the Toulouse School of Economics. He is also a Senior Fellow at the IAST and the founder of oTree Open Source Research Foundation and Data Science Justice Collaboratory. Chen was previously Chair of Law and Economics and co-founder of Law and Economics Center at ETH; he was a tenure-track assistant professor in Law (primary), Economics, and Public Policy at Duke University. He received his BA (Summa Cum Laude, Phi Beta Kappa) and MS from Harvard University in Applied Mathematics and Economics; completed his Economics PhD from MIT; and obtained a JD from Harvard Law School. 

 Chen uses his extensive empirical training to tackle long standing legal questions previously difficult to empirically analyze. He has attained prominence through the development of open source tools to study human behavior and through large-scale empirical studies – data science, artificial intelligence, and machine learning – on the relationship between law, social norms and the enforcement of legal norms, and on judicial systems.

Date: September 21st, 2023 (12:30 pm – 1:30 pm)

Speaker: Alex Imas

Paper Title: “Over- and Underreaction to Information”

Abstract: This paper explores how properties of the learning environment determine how people react to information. We develop a two-stage model of belief formation where people first reduce complexity by channeling attention to a subset of states that are representative of the observed information, then evaluate this information using Bayes’ rule subject to cognitive imprecision. The model predicts overreaction when environments are complex, signals are noisy, or priors are concentrated on intermediate states; it predicts underreaction when environments are simple, signals are precise, or priors concentrated on more extreme states. Results from a series of pre-registered experiments provide direct support for these predictions, as well as the proposed attentional mechanism. We show that the two-stage model is highly complete in capturing explainable variation in belief-updating; in particular, the interaction between the two psychological mechanisms is critical to explaining belief-formation in more complex settings. These results connect disparate findings in prior work: underreaction is typically found in laboratory studies, which feature simple learning settings, while overreaction is prevalent in financial markets, which feature more complex environments.

Bio: Alex Imas is a Professor of Behavioral Science and Economics at the University of Chicago Booth School of Business. He studies behavioral economics with a focus on dynamic decision-making. His research explores topics related to choice under uncertainty, discrimination, mental representation, and how people learn from information. Alex Imas’ work utilizes a variety of methods, including lab experiments, field experiments, analysis of observational data and theoretical modeling.

Date: May 4th, 2023 (12:30 pm – 1:30 pm)

Speaker: Daniel Martin

Paper Title: “Labeling and Training with Elicited Beliefs”

Abstract: We introduce the use of incentive-compatible belief elicitation for labeling data and training machine learning models.  Eliciting beliefs truthfully through proper scoring rules is now standard in experiments and surveys, but has not yet been applied to labeling or training.  We conduct an online experiment in which participants were incentivized to truthfully report their belief that a white blood cell was cancerous for a series of cell images and propose methods for labeling each image based on participant reports.  We evaluate these methods by training a convolutional neural net on the labels they generate and find that they outperform standard labeling methods in terms of both accuracy and calibration.

Bio: Daniel Martin is a behavioral, cognitive, and experimental economist who studies attention and perception (how information is processed) and information disclosure (how information is communicated). His current research explores many ways in which human and AI interactions are impacted by attention, perception, and information disclosure. He is currently the Wilcox Family Chair in Entrepreneurial Economics at UCSB, and before receiving a PhD in Economics from NYU, he was the co-founder of a small business that is now one of the leading providers of IT services to small and medium-sized businesses in the Carolinas. At UCSB he teaches undergraduate courses in entrepreneurship and PhD courses on attention and perception.

Date: April 27th, 2023 (12:30 pm – 1:30 pm)

Speaker: Annie Liang

Paper Title: “The Transfer Performance of Economic Models” (joint with Isaiah Andrews, Drew Fudenberg, Lihua Lei, and Chaofeng Wu)

Abstract: Economists often estimate models using data from a particular setting, e.g. estimating risk preferences in a specific subject pool. Whether a model’s predictions extrapolate well across settings depends on whether the estimated model has captured generalizable structure. We provide a tractable formulation for this out-of-domain prediction problem, and define the transfer error of a model to be its performance on data from a new domain. We derive finite-sample forecast intervals that are guaranteed to cover realized transfer errors with a user-selected probability when domains are iid, and use these intervals to compare the transferability of economic models and black box algorithms for predicting certainty equivalents.  We find that in this application, black box algorithms outperform the economic models when estimated and tested on different data from the same domain,  but models motivated by economic theory generalize across domains better than the  black-box algorithms do. 

Bio: Annie Liang is an assistant professor of economics and of computer science (by courtesy) at Northwestern University. Her research is in economic theory—in particular, learning and information—and the application of machine learning methods for model building and evaluation. Prior to joining Northwestern, she was an assistant professor of economics at the University of Pennsylvania and a post-doctoral researcher at Microsoft Research.

Date: April 20th, 2023 (12:30 pm – 1:30 pm)

Speaker: Joshua Schwartzstein

Paper Title: “Model Persuasion: Theory and Experimental Evidence”

Abstract: To understand new information, we exchange models or interpretations with others. This talk provides a framework for thinking about such social exchanges of models and presents recent experimental results that test key assumptions and predictions of the framework.

Bio: Joshua Schwartzstein is the Jakurski Family Associate Professor of Business Administration in the Negotiation, Organizations & Markets Unit at Harvard Business School. His research incorporates psychologically realistic assumptions about human cognition into formal economic models and uses these models to re-think questions in domains ranging from persuasion to health-care markets.  

Date: April 13th, 2023 (12:30 pm – 1:30 pm)

Speaker: Severine Toussaert

Paper Title: “Stochastic dominance and preference for randomization”

Abstract: Decision theorists usually take a normative view on stochastic dominance: a decision maker who chooses a lottery that puts more weight on options he likes less must be making a mistake. In this project I argue that stochastic dominance violations may naturally occur in situations where anticipatory utility is high, such as going on a holiday trip. In such a situation, the decision maker may trade the certainty of going to his favorite destination for the excitement of not knowing where he will go. To document this phenomenon, I conduct an experiment in which participants make a series of binary choices between a sure destination and a lottery over holiday trips. The outcome of the lottery is revealed close to the date of travel. I vary the characteristics of the lotteries to understand when violations of stochastic dominance are most likely to occur and analyze their properties. I discuss the implications for the modelling of anticipatory utility.  

Bio: Séverine Toussaert is an Associate Professor at the Department of Economics of the University of Oxford. She obtained her PhD from NYU in 2016. Her research combines theory, lab and field data to shed light on the non-material costs and benefits that may enter a decision maker’s utility function such as the self-control cost of avoiding temptation. Several applications she has worked on cover health topics, including smoking cessation and diet choices.