Reproduce “BotPercent: Estimating Bot Populations in Twitter Communities” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhaoxuan Tan, Shangbin Feng, Melanie Sclar, Herun Wan, Minnan Luo, Yejin Choi, Yulia Tsvetkov
BotPercent introduces a framework for estimating bot populations within Twitter communities, moving beyond individual account classification. The system provides accurate, calibrated estimates across diverse communities, uncovering substantial bot presence in domains like cryptocurrency and partisan news media and correlating activity with major socio-political events.
Key insights
- Understanding bot populations at the community level provides crucial insights into systemic manipulation that individual-level detection cannot fully capture.
- Bot activity is highly targeted and not uniformly distributed, showing significant concentrations in domains such as cryptocurrency, technology, and hyper-partisan news media, and spiking around major events.
- A multi-modal, multi-dataset, and calibrated ensemble approach significantly improves the robustness, accuracy, and generalizability of bot detection for community-level estimation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.