Plan: automated email outreach to authors of "data available on request" papers

[larnsce]

Goal

Email the corresponding authors of every paper in our datasets whose data availability statement (DAS) says the data are available on request, from an ETH Zurich institutional mailbox, with a personal greeting and a link to the paper in question — asking them to share the data and offering openwashdata publication (open license, DOI, full author credit).

What the data supports today

Measured on inst/extdata/ (2026-07-26):

Dataset	On-request papers	With contact email
washdev	128	127
uncnewsletter	44	43 (26 via first-author fallback)
ploswater	1	1
Total	173	170 → 153 unique addresses

datapapers is excluded (data papers publish their data by definition). 3 papers have no usable address. Selection matches das_type containing "request" (case-insensitive), which also catches the handful of rows carrying the verbatim statement instead of the coded category.

Phases

1. Outreach 1/5: ethics, data protection, and mail-policy checks (gate before sending) #42 — Ethics, data protection & mail-policy checks — gate for everything else; nothing is sent before this is resolved.
2. Outreach 2/5: build and manually review the contact list #43 — Contact list: build & manual review — selection rules, email fallback, greeting heuristic, known address-quality problems.
3. Outreach 3/5: email template and sender identity #44 — Email template & sender identity — wording, subject, signature, opt-out sentence.
4. Outreach 4/5: sending mechanism from the ETH mailbox #45 — Sending mechanism (ETH mailbox) — SMTP via mail.ethz.ch, dry-run/test modes, batching, sent log.
5. Outreach 5/5: response tracking and follow-up policy #46 — Response tracking & follow-up policy — replies, bounces, opt-outs, received datasets, reminder policy.

Sequencing: #43–#45 can be prepared in parallel, but the first real send waits on #42, and #46 starts with the first batch.

Prototype note

A working prototype of phases 2–4 exists on branch claude/automated-author-email-outreach-amipwb (data-raw/outreach/: list builder, template, blastula send script, generated contact CSV). Treat it as a draft implementation of this plan, not a decision — it is unmerged, nothing has been sent, and it gets revised or discarded according to the outcomes of the sub-issues.