Potential Research Projects
Please see below for a sample list of research projects. Please note this is not a complete list. We will continue to add opportunities and will have additional projects in the Humanities and Social Sciences.
The Center for Decision Research at the University of Chicago Booth School of Business seeks conscientious research assistants to conduct behavioral science and social psychology experiments through a variety of settings through the PIMCO Decision Research Laboratories during Summer 2023. Undergraduate students studying social psychology, cognitive psychology, and behavioral economics are encouraged to apply.
Each research assistant will have an opportunity to facilitate behavioral science experiments led by faculty and graduate student researchers of the Center. These studies involve interfacing with participants, and guiding them through precise study protocols in order to conduct research and learn about how we understand ourselves and each other.
Each research assistant will also have an opportunity to be mentored by members of the CDR’s academic community, including faculty members, postdoctoral researchers, and graduate students. Workshops and presentations of original and emergent research by behavioral scientists around the world are offered by the Center, and research assistants are welcome to attend as they deepen their understanding and interest in the field of behavioral science.
Deep bias in generative models for scientific discovery
Recent advances in deep learning have led to powerful “counterfactual” models that can realistically manipulate input data (such as chest XRays or images of people’s faces). However, the notion of “similarity” implied by such a deep learning model does not always agree with the notion of similarity familiar to most humans. In this project, we wish to identify common dimensions of disagreement between machines and humans. This has important implications for the use of such counterfactual models for chemistry and healthcare, for example. The successful applicant should have some quantitative education and experience using Python.
Generalizability of heart condition predictions
Using machine learning models, it is possible to predict some kinds of adverse health events such as sudden cardiac death (SCD), using only the data collected from an EKG. Because of natural variation in populations (among other things), these predictions can vary from source to source. In this project, we wish to take a model trained to predict an outcome such as SCD, measure its performance on a new data source, and diagnose the drivers of reduced performance. The successful applicant should have some interest in healthcare and experience using Python.