How I Do Research: A Living Overview of My Methods

Over the past few years, I’ve worked with a range of research methods across law, social science, and computation. Like many academics, I’m driven most by curiosity.

My training began in law and policy, so I initially approached research with a natural inclination toward policy rationales and solutions. But as I started exploring questions like how people perceive AI and what effects it might have on our collective future, I realized that not all inquiries call for prescriptive answers. It requires a mixture of imaginative thinking and descriptive research.

At the University of Washington and Princeton CITP, I had the opportunity to work alongside researchers in computer science and information science. Without an engineering background, I gradually expanded my toolkit. With it, I have a braoder range of questions I felt empowered to ask. Learning new methods is challenging, but it has also given me some of the most rewarding moments in my research journey.

This page is a living summary of how I do research, a personal reflection of what I’ve learned, enjoyed, and wrestled with.

🏛 Traditional Legal Analysis

Much of legal scholarship still revolves around deep engagement with court decisions, statutes, and administrative rules — often accessed through specialized databases like LexisNexis or Westlaw. This work is notoriously text-heavy and interpretive. There’s no algorithmic way to summarize “what courts say” about an issue because the legal world isn’t built on representative sampling or generalizable claims.

Instead, legal scholars construct arguments by selecting, interpreting, and synthesizing legal materials. Some cite cases that resonate with their argument; others critique cases that diverge from it — but either way, the choices are often subjective and grounded in the scholar’s point of view. Cases are deeply contextual, and the “truth” is rarely statistical.

Learning to read cases and statutes with depth and purpose is not intuitive. It took me time — and I’m still learning. But I’ve come to admire those who do this work with clarity and even aesthetic beauty. The goal isn’t just to catalog what the law says — it’s to make sense of how legal doctrine functions, and to tell a compelling story about it.

📊 Human Subjects Research

I also use both quantitative and qualitative methods involving human participants.

Quantitative

I design and run online experiments and surveys using tools like Qualtrics, Prolific, R, and Python. I used to think survey research was straightforward — you write a questionnaire, collect responses, run a regression. But I’ve since learned how much work goes into survey design, pre-registration, pilots, and iterative testing before you even get to the “real” data.

There’s no single formula. You’re constantly balancing precision and practicality. And to my surprise, I found that I actually enjoy data cleaning and analysis — it reminds me of doing the dishes: mentally light but fully absorbing, full of small, satisfying puzzles. It’s long work, but it gives you little assignments (often debugging) and you never really get bored.

Qualitative

I also conduct semi-structured interviews and expert panels. I love this part of research. These conversations often reveal insights that no survey could capture — the nuance, the emotion, the stuff between the lines. Even in large-scale surveys, I like to include open-ended questions so I can gather more human texture.

Designing and facilitating these sessions is an art: you want participants to find their flow and speak freely, but you also need to stay focused enough to eventually synthesize what you’ve learned. Some things I’ve come to believe:

  • You never know what you’ll learn until you’re in the room.
  • You get better the more you do it.
  • If you enjoy the session, your participants will too — and that makes a difference.

For analysis, I enjoy using the web version of Atlas.ti or Taguette. Taguette is a lovely, light-weight, free alternative to proprietary coding tools. But if I have funding, I still would like to use paid tools because it allows you to label quotes with both codes and participant ID as well as to establish the hierarchy of codes. When producing a codebook based on the agreement between co-authors, Google sheet just works perfectly.

💻 Computational Social Science (CSS)

Recently, I’ve picked up methods from computational social science — especially content analysis at scale. I’m not an engineer by training, but surrounded by brilliant software engineers using these tools masterfully, gave me the courage to try. (Watching them made me think, “maybe I can do this too.”)

I now use a mixed-methods CSS approach that integrates:

  • Automated web scraping to collect large volumes of digital content,
  • Manual qualitative coding of a sample, and
  • Machine-assisted thematic analysis using large language models (LLMs).

This method blends the depth of qualitative research with the speed and scope of automation. It allows me to track sociocultural patterns in digital spaces — without losing the interpretive sensibility that qualitative research brings.