Research

Publications

Accepted · JEPS

Perceiving Politics: Considering the Political Categorization of Faces

Thomas Bergeron, Blake Lee-Whiting, Natasha Goel

2026 · Political Behavior

Political Party or Policy Position? The Role of Policy Partisanship and Party Cues in Voter Decision Making

Clareta Treger, Thomas Galipeau, Thomas Bergeron, Sarah Lachance, Natasha Goel, Mujahedul Islam, Blake Lee-Whiting, Beatrice Magistro, Peter J. Loewen

2024 · Political Behavior 47(4): 1483–1500

The Nature of Online Talk: Incivility of Opposing Views and Affective Polarization

Natasha Goel and Eric Merkley

2023 · The Canadian Geographer 67(3): 380–393

Residential Segregation and Inequality: Considering Barriers to Choice in Toronto

Natasha Goel

Book Chapters

Under contract · McGill-Queen's University Press

Media Influence on Trust in Online Remote Voting

In Trust and Trustworthiness: Evaluations of Online Voting

Helen Hayes and Natasha Goel

In works · Routledge

The Comparative Correlates and Effects of Anti-intellectualism in Canada and the United States

In The Routledge Handbook of Anti-Intellectualism

Eric Merkley and Natasha Goel

Working Papers

Revise & Resubmit · PNAS Nexus

Artificial Influence: Comparing the Effects of AI and Human Source Cues in Reducing Certainty in False Beliefs

Natasha Goel, Thomas Bergeron, Blake Lee-Whiting, Thomas Galipeau, Danielle Bohonos, Mujahedul Islam, Sonja Savolainen, Clareta Treger, Eric Merkley

A pre-registered survey experiment tested whether conversations with ChatGPT‑4o reduced certainty in false political beliefs, such as the dangers of the COVID-19 vaccination or abortion in the United States (N = 1,730). The treatment manipulated who participants were told they were speaking with, ChatGPT‑4o, a human expert, or a survey respondent who disagrees with them, to test AI's persuasiveness against a low-credibility peer and a high-credibility expert benchmark. Certainty declined significantly across all three conditions, with 29% of participants shifting toward the accurate belief. An AI label showed a modest edge over a disagreeing peer in reducing certainty but was ultimately outperformed by a human expert label. The findings suggest that AI's persuasive power still depends on the credibility of its label relative to more established sources, which may limit its promise as a scalable fact-checking tool.

Under review

The Relationship Between Polarization Misperceptions and Partisan Hostility Is Reciprocal

Eric Merkley and Natasha Goel

Uses longitudinal data and experiments in Canada to evaluate the causal relationship between misperceiving polarization and partisan hostility.

Submitted for peer review

Threat Primes Do Not Reliably Bias Immigration Information Processing

Natasha Goel

This paper asks whether fear and threat appeals, a staple of campaign messaging, actually bias how people process political information. I test this in the context of immigration, using two pre-registered survey experiments in the United States (N = 1,290) and Canada (N = 1,136), where immigration is politicized very differently. Despite producing real emotional reactions, none of the threat primes, cultural, economic, or partisan, reliably produced the bias the theory predicts. Where effects did emerge, they sometimes moved in opposite directions across the two countries. The paper's larger point is that threat-based messaging does not travel on its own. Its effects depend heavily on the political environment amplifying it, not just the content of the message itself.

Working paper

Motivating Demand for Corrective Information

Natasha Goel

This paper asks not whether corrections work, but whether people are likely to encounter them. Even effective corrections matter little if they never reach people in the first place. In a pre-registered survey experiment fielded in the United States (N = 1,401), participants were shown a false claim before being asked to choose from a set of headlines, one of which corrected the claim. A prompt to consider accuracy did not raise selection of the corrective headline on average, calling into question the value of one-size-fits-all nudges. It did, however, work among those least inclined to question their own views, suggesting such primes may induce curiosity precisely where it is otherwise absent. The lack of genuine demand for corrective information remains a significant challenge.

Working paper

Data Types and Government Access in Social Media Privacy Policies

Gabrielle Lim and Natasha Goel

We examine an understudied feature of political communication on social media: data privacy. As more of our online interactions can be collected, tracked, identified, and monetized, most Americans say they are concerned about how their personal data is used by corporations and governments. We use a pre-registered conjoint experiment fielded in the United States to test which features of a social media platform's privacy policy—such as which political actors and countries can access user data—shape whether users say they would continue using the platform. We also examine what trade-offs they are willing to make, for instance, giving up their personal data in exchange for appealing new features.