About the research guide
The research guide is an experimental chat assistant that answers questions about Supreet Kaur's research. This page explains what it is, how it works, and where it falls short.
What it is
The guide answers questions about Supreet Kaur's papers, survey chapters and talks, for students and policymakers, in English, Hindi or Punjabi, with links to the sources it draws on.
It describes her work in the third person. It is not Supreet Kaur, it does not speak for her, and she has not reviewed its answers. Answers are written by Claude, an AI model made by Anthropic, and can be wrong: check the linked papers, and use “Flag this answer” when something doesn't match the work.
This is a basic first version, built in September 2026 with Claude Code. It will change as we learn from how people use it.
What it knows
- Every paper on the Research page, with its abstract and links, and the full text of every paper that has a PDF there (the main text, without appendices).
- Plain-language summaries and fact sheets for twelve papers, prepared for the paper explainer videos and checked against the papers.
- The two survey chapters with Emily Breza: “Psychology and Development” and “Labor Markets in Developing Countries”.
- Transcripts of two talks: “Labour and Migration” at the Nobel Symposium (Stellenbosch, March 2024) and “The Psychology of Poverty” (Berkeley Economics Speaker Series, 2024).
- This website: her CV, research statement, and talks and media.
It uses only her own work. Press coverage is linked on the site, but the guide doesn't draw on it.
How it works
Sending every paper with every question would be slow and expensive. Instead, each question goes to the model with a compact overview: the abstracts and summaries, outlines of the chapters and talks, and an index of longer documents. When a question needs detail, such as a number, a research design, how the authors answer an objection, or what she said in a talk, the guide reads the relevant fact sheet, section of the paper, chapter section or transcript part, and answers from that.
When the site's administrators find an answer that misses the mark, they write a correction, and the guide takes it into account from the next question on.
Spending is capped: there are daily and monthly budgets and a limit on questions per visitor. When a limit is reached, the guide pauses until the next day.
Questions and answers are stored so they can be reviewed and used to improve the guide. IP addresses are not stored.
What's next
The guide will improve as more of her material is added, including talks, columns and material that isn't published elsewhere, and as reviews of stored answers and flags show where it goes wrong.
Doing it yourself
You could do most of this yourself. Download the papers from the Research page, where each title links to its PDF, give them to an AI assistant, and ask your questions. With the full papers in front of it, an assistant can go into more detail than this guide keeps at hand, although each question then costs more and takes longer. The site also has a plain-text overview for AI assistants at /llms.txt.
What the guide adds is preparation: fact sheets checked against the papers, talk transcripts that link to the moment in the video, and answers that point back to the sources.