Writings

Short, accessible notes about networks, complexity, and why blending disciplines helps us understand society.

Network Science · 2–3 min read

What is a Network, Really?

A network is a set of nodes (people, cities, proteins) connected by edges (friendships, roads, interactions). This simple representation lets us study how structure shapes behavior: why information goes viral, how traffic jams form, or which hospitals are most critical in an emergency. Two big ideas: degree (how many connections a node has) and centrality (how important a node is to flows through the network). Many social and technological networks are heavy-tailed—a few hubs have many links—so interventions targeted at hubs can be surprisingly effective.

Complex Systems · 3–4 min read

Complex Systems: When the Whole is More than the Parts

In complex systems, many adaptive parts interact, creating patterns that no single part dictates. Think of traffic, economies, or online communities. Feedback loops and nonlinearity mean small changes can have outsized effects. Modeling these systems with agent-based models reveals emergence—collective behavior arising from simple local rules. Rather than predicting exact futures, we map regimes: conditions under which cooperation is stable, cascades are likely, or systems become fragile.

Interdisciplinarity · 2–3 min read

Why Interdisciplinary Work Matters

Real-world problems ignore department boundaries. Pandemics combine biology, behavior, mobility, and policy. By blending tools from computer science, economics, sociology, and public health, we can both interpret data and design interventions. Interdisciplinarity isn’t just mixing jargon: it’s choosing the right level of abstraction for a question, validating models with multiple kinds of evidence, and being honest about uncertainty.

LLM Agents · 2–3 min read

Language Models as Social Agents

Large language models can act as boundedly rational agents in simulations, negotiating, coordinating, or spreading information. Used carefully, they let us test policy ideas and probe mechanisms that are hard to observe in the wild. A key challenge is aligning micro-level behavioral assumptions with macro-level outcomes—so we combine LLM agents with validated data and theory, not as replacements but as complements.