Brandon Derr

Publications

Interstate Competition and Transnational Threats (2025)

Kinne, B., Tellez, J., Stewart, A., Iliev, I., Derr, B., Murthy, S., & Bernhard, P.

Under review at International Organization.

“Gone with the Wind?” Hardly: The Persistence of Climate Beliefs amid Extreme Events (2025)

Stan, K., Hill, J. B., Bernstein, D., Tucker, C. S., Iliev, I. R., Cochran, D. M., Jr., Derr, B., Acarlar, M., & Guerra, D.

Submitted for review.

Terrorist Responses to Verbal Attacks: Strategic Logic Towards Whom? (2025)

Iliev, I. & Derr, B.

Working paper.

Projects

"Polarized Without the Other Side: Measurements from Congress on Bluesky"

Masters Thesis

Political Science @ USM | 2025 | PDF

• Gathered and analyzed 80,000+ Congressional posts from Bluesky's API to measure polarization and engagement.
• Constructed and presented project proposal with a timeline to supervisors.(Slides here)
• Built an end-to-end pipeline (scraping → cleaning → sentiment analysis → negative binomial regression) to quantify how rhetoric shapes user engagement.
• Found that highly polarized posts were associated with 91% more likes and 62% more replies than neutral posts, highlighting incentives for extreme messaging.
• Presented methodology and findings to committee members, defending modeling choices, robustness checks, and limitations.(Slides here)

ICEWSnet R Package

R package for researchers to transform raw ICEWS data into aggregations for network analysis.

Research Assistant @ USM | 2024

• Developed a core function to convert ICEWS data into date-indexed interstate networks.
• Optimized code from other developers and documentation for users.
• Prepared the package for CRAN publication, ensuring compliance with standards and testing.
• Code is in a private repo. Ask me about it!

Obsidian Daily Notes Rainmeter

A plugin for Rainmeter that displays the current daily note from your Obsidian folder on the desktop.

Personal | 2024 | GitHub Repository

• Developed a Rainmeter skin that dynamically displays Markdown-formatted daily notes directly on desktop.
• Published plugin online; 350 downloads in its first year.

"Large Language Models for Sentiment Analysis on Political Tweets: A Case Study on Environmental Issues"

Senior Capstone Paper.

Political Science @ USM | 2024 | PDF

• Identified problems with traditional social science sentiment analysis methods.
• Developed an automated zero-shot topic classifier in Python using APIs from Anthropic, OpenAI, and Mistral. Tested on sample of 2,000 Tweets.
• Reduced processing time from several days to a few hours, increased flexibility, and provided more detailed results compared to a traditional sentiment analysis method.
• Evaluated AI model accuracy, speed, cost, and inter-coder reliability.

"Climate Change Policy Preferences and Senatorial Tweets"

Undergraduate Thesis.

Honors College @ USM | 2023-2024

• Investigated the difference in climate change opinion between elected representatives, scientists, and the public.
• Collected more than 200,000 Tweets from Congress members using a Python web scraper.
• Trained machine learning algorithms with cloud Jupyter notebooks to automatically identify Tweets related to renewable energy, climate change concern, and climate events.
• Modeled public opinion using the frequency of these topics. Presented results at 4 academic conferences.
• Identified processing-time roadblocks that I addressed in my senior capstone research.

© Brandon Derr. All rights reserved.