A select collection of personal research projects—including ongoing and past work—and areas I explore on my own time. This section excludes any work I do not directly lead or co-lead and excludes work performed as part of my regular employment. All of these were undertaken at the doctoral level or higher.
Second Raven. (Newsletter).
Dr. Ryan Kennedy, Dr. William Minozzi, Dr. Laura Moses, The Ohio State University
We are a group of researchers affiliated with the Institute for Democratic Engagement and Accountability (IDEA) and the Machine-Assisted Human Decision-making (MAHD) Lab at The Ohio State University. We use artificial intelligence (AI) and unique analysis techniques to reveal the ways public opinion research and reporting warp our views of U.S. society and of each other. I helped create and currently maintain an end-to-end R framework that streamlines data cleaning and analysis, performs quality checks, and generates custom newsletters and PowerPoint presentations using large language models (LLMs) for substack content.
2026-2027. Congressional Conspiracy Rhetoric Index.
Dr. Adam Cayton, University of West Florida, Dr. Joanne Miller, University of Delaware, & Dr. Ryan Kennedy, The Ohio State University
PhD Candidate Collaborator: Jillian Rothschild, University of Maryland
We develop and validate a novel measure of conspiratorial communication among U.S. Congress members using their official e-newsletters from 2009 to 2025. This project leverages a large language model (LLM) to systematically score e-newsletter content, generating both an overall conspiracy score and theory-specific scores for prominent domestic conspiracy narratives. I led the development and validation of this measure, designing the LLM-based scoring framework and overseeing the creation of newsletter-level, legislator-year, and career-level indicators of conspiratorial communication.
citehelpR package for R. (Github Repo).
This package checks and converts references in academic documents. It uses configurable fuzzy-matching to flag potentially missing in-text citations and reference list entries, allowing users to adjust the matching sensitivity to account for minor inconsistencies and human error (e.g., formatting, typos). This package can also convert standalone bibliographies between APA and Chicago citation styles.
dopplR package for R. (GitHub Repo).
Dr. Ryan Kennedy, The Ohio State University
This package generates and evaluates synthetic survey data. It can simulate responses based on real respondents profiles, create synthetic respondents and datasets that mirror the structure and distributions of real data (for cases where original data cannot be publicly shared), and generate multiple synthetic data variations. Additional features include weighting synthetic respondents to match target marginals, summarizing dataframe structures, comparing distributions with Komolgorov-Smirnov (KS) tests, flagging synthetic data that may be too similar to the original, and synthetic data tuning.
2026. AI Detector for Open-Ended Survey Data.
Priorietary model built for Survey 160 that analyzes open-ended survey responses and assigns probability scores indicating the likelihood that a given answer was AI-generated.
2025-2026. Echoes of Conspiracy: Tracing Deep-State Rhetoric in Congressional E-Newsletters.
Dr. Joanne Miller, University of Delaware
We examine how members of Congress amplify or challenge deep-state conspiracy theories in their constituent communications and track how conspiracy-laden narratives spread across Congress. I trained a large language model (LLM) to analyze a comprehensive database of congressional e-newsletters, creating a coding framework to identify conspiracy-related language and test its reliability against human annotations. I then use these classifications to track when and where conspiracy references appear, descriptively compare patterns across parties and individual legislators, and assess how these narratives move through the institution over time.
* Project funded by the Institute for Humane Studies (Grant No. IHS019471). This initial project evolved into a wider group collaboration (see above “Congressoinal Conspiracy Rhetoric Index”).
2023-2025. Doctoral Dissertation. Negative Public Opinion: How Normalized Out-Group Biases Shape Politics.
More information and relevant files and links can be found here: https://stephpedron.com/dissertation/.
Stephanie Pedron 