Week 6 — Reading the Frontier II upcoming¶
July 27 – July 31, 2026 · Chapters 6.1 – 6.5 · Lab 6 · Mentor Session 6 · Problem Set 6
Text-as-data and machine-learning-flavored empirical finance, plus a full module on using LLMs responsibly as a research co-pilot.
By Friday your capstone charter is locked. Anything not in the signed charter on Friday at 5 pm is not in your symposium poster on August 12.
Daily schedule¶
| Block | What |
|---|---|
| 9:00 – 12:00 | Read Ch 6.1 — Kogan, Papanikolaou, Seru & Stoffman (2017) Reader's Guide (market-based patent value). Work nb6.1 (patent-value event study). |
| 1:00 – 4:00 | Capstone team standups (15 min each). |
| 4:00 – 5:00 | Office hours |
| Block | What |
|---|---|
| 9:00 – 12:00 | Read Ch 6.2 — Hoberg & Phillips (2016). TNIC; 10-K cosine similarity. Work nb6.2 (sklearn vectorization). |
| 1:00 – 4:00 | Read Ch 6.3 — Loughran & McDonald (2011). Finance sentiment dictionaries. Work nb6.3. |
| 4:00 – 5:00 | Office hours |
| Block | What |
|---|---|
| 9:00 – 12:00 | Read Ch 6.4 — Bartlett et al. (2022) + Bhutta–Hizmo–Ringo Reader's Guide (fair lending in the algorithmic era; ties to Sun & Gao 2019). Work nb6.4. |
| 2:00 – 3:30 | Mentor Session 6 — "Responsible AI in empirical research" |
| 4:00 – 5:00 | Office hours |
| Block | What |
|---|---|
| 9:00 – 12:00 | Read Ch 6.5 — The AI Co-Pilot for Research (LLM-in-the-Loop). Prompt patterns, RAG over 10-Ks, OOS-validated classification, env-var key discipline. |
| 1:00 – 4:00 | Lab 6 — RAG over 10-Ks (offline-safe; live Anthropic / Azure cells are env-gated and optional). |
| 4:00 – 5:00 | Office hours |
| Block | What |
|---|---|
| 9:00 – 12:00 | Wrap PS 6.1 – 6.5. |
| 1:00 – 3:00 | Capstone charter — signed off. Each team presents 5 minutes; Prof. Gao signs (or sends back for revision). |
| 3:00 – 4:00 | Demo + Week-7 preview |
| 5:00 pm | Deliverable cutoff |
What to read¶
- Ch 6.1 — KPSS 2017 Reader's Guide
- Ch 6.2 — Hoberg & Phillips 2016 Reader's Guide
- Ch 6.3 — Loughran & McDonald 2011 Reader's Guide
- Ch 6.4 — Bartlett et al. 2022 + Bhutta–Hizmo–Ringo
- Ch 6.5 — The AI Co-Pilot for Research
Notebooks: notebooks/week-06/ — including nb6.5, the AI co-pilot lab (offline-safe, env-gated live cells).
Mentor reading: Mentor 6 — "Responsible AI in empirical research".
Secrets discipline (mandatory)
nb6.5 and any code that touches an LLM API loads keys only from environment variables — never from a literal string in a notebook. Do not commit .env. See Resources → Data access.
What to submit (due Fri Jul 31, 5:00 pm ET)¶
- Problem Set 6 — PS 6.1 through 6.5.
- Lab 6 notebook with a RAG-augmented classification + an OOS-validated accuracy score.
- Capstone charter — signed-off version in
projects/<team-name>/charter.md. - Journal entry —
journal/week-06.md. - Pull request —
week-06 — <your name>.
Data we touch this week¶
- EDGAR 10-K / 10-Q — for Hoberg–Phillips and LM.
- USPTO PatentsView + NBER patent linking table — for KPSS.
- Loughran–McDonald dictionary.
- HMDA — for the fair-lending paper.
Next week preview¶
Week 7 — Capstone build →. From research question to pre-analysis plan to cleaned dataset, with the identification memo locked.