Illustrated portrait of Rojin Ziaei

Rojin Ziaei also Nazdim

Biostatistician II in social epidemiology at Stanford Population Health Sciences, predicting population health behaviour with social simulation 🩺💻

Hi, I'm Rojin!

Some people know me as Nazdim, and either name is great.

I'm a Biostatistician II in social epidemiology at the Stanford Center for Population Health Sciences (PHS), where I work with Dr. David Rehkopf. My work is a collaboration between PHS and the Stanford Center on Poverty and Inequality (CPI), with Dr. David Grusky. I'm also pursuing an MS in Computer Science at Georgetown University.

My work focuses on predicting population health behaviour, with the goal of improving health outcomes. I combine LLM-powered social simulations with classical NLP and machine learning, working with electronic health records (EHR) and long-form narrative data. It builds on my work on the limits of LLM-simulated societies, which showed that alignment can erase real cross-cultural differences.

More broadly, my research sits between computer science and health. I study how large language models behave in medical settings, from cognitive bias to how they respond when patients arrive with an incorrect self-diagnosis, and I've helped build AgentClinic, a benchmark for evaluating AI in simulated clinical environments. I'm also interested in causal inference, speech and language technology for Persian and other low-resource languages, and what LLMs can and can't tell us about human societies.

Before this, I was a software engineer on Capital One's AI Foundations team, and a research assistant at Stanford with Susan Athey, at Johns Hopkins with Anjalie Field, and at the University of Maryland, where I earned my BS in Information Science and was the department's 2023 commencement speaker.

Things I think about

  • Social epidemiology
  • Social simulation
  • Health behaviour prediction
  • Population health
  • Medical AI and bias
  • Large language models
  • Causal inference
  • Low-resource languages
  • Computational social science

✨ News: My paper The Limits of Simulated Societies: How LLM Alignment Erases Cross-Cultural Variance was accepted at AACL! 🎉