Assistant Professor and Faculty Fellow at the NYU Center for Data Science

Mom, you remember I’m a political scientist? English has an old figure of speech, a person’s politics is their way of looking at things. Well, we are in a brave new world where generative AI redraws the boundaries of fields. I’m taking advantage of this mess to study how one’s political self organizes the way they look at the world, pretty literally, and, the other way around, how you can take a person’s behavior and work backward to recover what their political identity actually is. My broader agenda is to “embed” political identity in computer vision tools themselves, so that identity stops being something bolted on after the fact and becomes part of how the model sees behavior from the start. I believe this matters a great deal for computational political science. I’m on the job market this year, wish me luck!

I see my research as situated primarily within the following fields:

Computational Political Behavior (CPB)
Politics and AI (PAI)

I am on the job market 2026-27

I am currently working on:

Reading Between the Eyes: A Framework for Recovering Political Identity from Eye‑Tracking Weak Signal. (CPB)

Job Market Paper

I build a classifier that uses computer vision and gaze data to learn political identity from a weak behavioral signal of eye tracking.

Human versus Computer Vision. (PAI)

I study what machines learn when they are trained to imitate human behavior, using visual attention as a case where their predictions can be compared directly with people. I show where that imitation fails across people and which of those failures machines can learn to correct.

Unpacking the Eye of the Beholder: Social Location, Identity, and the Moving Target of Political Perspectives. (PAI)

I develop and test a classifier for measuring (political) identity-specific visual sentiment.

Vision and Division: How Partisan Identity Patterns Visual Processing of Political Imagery. (CPB)

In a large-scale eye-tracking study, we show how motivated visual processing works on partisan divides.

Visual Political Polarization. (PAP) (CPB)

Abstract

Theories of political polarization assume that people who see the same content share a common basis for judgment, yet information sampling, the stage at which people select which information to take in, remains the least studied stage of polarization. I develop a theory of visual political polarization and use webcam eye tracking to provide the first quantification of how much the sampling channel contributes to polarization. Pilot evidence suggests that issue positions shape how opponents examine and evaluate visual information. Experimentally activating partisan identity moves gaze yet leaves disagreement at baseline. However, disagreement narrows when an exogenous change in how people view political material changes the information they extract from it, because their evaluations follow that information. The sampling channel therefore offers a previously overlooked opportunity to reduce polarization at the stage of information acquisition, before people form judgments. More importantly, differences in information sampling can recur with every exposure in the wild and compound into far wider disagreement in evaluations than any single viewing produces.

I develop and test a theory of visual political polarization, focusing on how people become polarized through the way they visually perceive the world.

The rest is in my CV.