Dynamic systems are hard to model. Dynamic systems involving human behavior, especially when interactions aren’t directly observable, are even harder. But the stakes are too high not to try.
At Tavern Research, we're building models to understand and influence how people form opinions online. You’ll take on problems like identifying bot networks, tracking the spread of narratives, modeling how media exposure shapes political attitudes, and helping campaigns test and target their messaging more effectively. Some days, you’ll design experiments. Other days, you’ll parse engagement metrics and raw text to figure out what’s actually working and why.
We believe people’s beliefs are shaped by the information they consume, and that information is increasingly engineered. The internet cracked the door open for opinion manipulation, and language models kicked it wide open. If we care about the future of democracy, we need to understand how influence spreads, mutates, and lands, and help our partners respond with precision and speed.
You won’t be handed a clean dataset or a fully formed research question. You’ll help define the problem, shape the data, and build tools that turn ambiguity into real-world impact.
One week, you might build a discriminator to detect bot networks. Another, you might model the impact of different messages in noisy, fast-moving attention markets. Or, develop a custom embedding model to track evolving narratives.
You’ll work closely with a tight-knit, interdisciplinary team, testing ideas fast, shipping iteratively, and pushing toward real-world impact. The timelines are tight, the inputs are messy, and the problems are wide open. If that sounds thrilling, we encourage you to apply.
We value in-person collaboration and expect employees to work regularly from our Chicago office.
Responsibilities
- Design and analyze experiments to measure treatment effects. You’ll collaborate on experiment design, analyze results independently, and iterate using methods like randomized trials, difference-in-differences, and causal inference.
- Build and maintain statistical and machine learning models that generate political and behavioral insights.
- Work hands-on with messy, real-world data: cleaning, debugging, and engineering features that support effective models.
- Write clean, well-documented Python code using scientific computing libraries.
- Partner with researchers and engineers to turn complex questions into technical solutions and modeling approaches.
- Contribute to model evaluation, diagnostics, and performance monitoring.
- Stay up to date on best practices and emerging tools in machine learning, AI, and causal inference.