Bolashak Fellows at Johns Hopkins SAIS · Two-Day Seminar 2026
Day 2 · AI & policy text

Day 2 · AI & Policy Documents

Turning documents into data with dictionary methods and large language models, applied to real Federal Reserve, IMF, and World Bank text. Two labs, a responsible-use session, and a closing research workshop — all run in claude.ai.

Go to the labs Download the slides
Schedule
9:00–9:40
Day2a

Text as data

The document pack introduced; document-term matrices; what LLMs change; the FOMC deliberation study.

9:45–10:40
Day2b

NLP · Dictionaries

Loughran–McDonald scoring of real Fed language; auditable word lists; the failure gallery (negation, hedging, domain shift); build-your-own dictionary.

10:45–11:00

Break

11:00–12:00
Day2c

NLP · From BERT to LLMs

Embeddings and attention without equations; the Fedspeak evidence (Hansen–Kazinnik; Trillion Dollar Words); live hawk/dove classification; what LLMs still get wrong.

1:00–2:20
Day2d

Lab 3 · Prompting the Fed

Zero-shot vs. rubric prompts; 3-run stability test; benchmark vs. human labels; structured extraction with 'not stated' rules; adversarial framing.

2:20–2:35

Break

2:35–3:40
Day2e

Lab 4 · IMF & World Bank reports

Extraction schema with mandatory verification; the hedging ladder; two-institution comparison; a one-page policymaker brief.

3:45–4:25
Day2f

LLMs in research · Rules of the game

Hallucination, reproducibility, bias, confidentiality; disclosure sentences; NIST / OECD / EU AI Act; research practice as internal control.

4:30–5:00
Day2g

Closing workshop

One-slide study template; cohort-matched project ideas; two generalized starter prompts; the take-home toolkit.

Morning 9:00–12:00 · Afternoon 1:00–5:00 · Two labs plus rules-of-the-game and a closing workshop. Head to the Labs page to run the prompts in Claude.

Slides · download