Report polish — how we reproduce third-party geo-amplification charts
Status: standing method (2026-09-14)
Why you care: Steve Jurvetson posted DigitalBorders-branded charts saying Coxon’s posts were heavily amplified by foreign-located X accounts. Those charts are real publications. We have not yet independently reproduced the raw rows. Until we do, cite them as their charts — not as DSC’s own sample.
Related: Amplification scoreboard · Research map
Plain terms
| Term | Meaning |
|---|---|
| DigitalBorders | Digital Borders Initiative — research site that studies where engagers’ accounts appear based |
| Geo / engager-origin chart | Breakdown of reposters into US / foreign / unresolved using public location signals |
| Independent reproduction | We pull our own sample of reposters and tally them — not just screenshot their graphic |
| OPS_GAP | Receipt climb unfinished (API cost, egress, deletes) — ≠ “the chart is fake” |
What DigitalBorders is (and isn’t)
| Thing | Reality |
|---|---|
| digitalborders.com | Free public research site (Engager-Origin studies, method writeups) |
| Paid Coxon CSV product | Not found — we did not “fail to buy” a checkout that doesn’t exist |
| Jurvetson “Source” link | Resolves to the homepage, not a case-file download |
| Coxon / Thayer geo numbers | Published as free charts on the X post (DigitalBorders-branded) |
| Independent reproduction | Uses X’s public location labels + sampling; scale costs X API / browser time, not a DigitalBorders fee |
Standing process (when the figure will appear in a DSC report)
- Archive their publication — chart images, post URL, capture date; transcribe sample size, % foreign/US/unresolved, top countries, method disclaimer.
- Classify epistemic status — CONFIRMED as published chart vs independently reproduced. Never collapse “no site CSV” into “numbers unsupported” if charts exist.
- Reproduce — target status IDs; sample N reposts (match their N when possible, e.g. 3,500); record account + location signals; bucket US/foreign/unresolved; write
raw/rows + our tally + method note. - Compare — MATCH / DIRECTIONAL / DIVERGES / OPS_GAP. Keep their disclaimer: location ≠ bots / coordination / state unless separately proven.
- File — evidence under the dig; update the scoreboard.
Anti-patterns
- Saying “UNSUPPORTED — no DigitalBorders export” when charts are on the citing tweet.
- Treating location geo% as proof of state influence ops or paid bots.
- Skipping reproduction because “it’s free online so we’re done” when the figure will appear in a DSC report.
First return task
Reproduce Jurvetson/DigitalBorders Coxon and Thayer charts + deposit raw sample data. Still open on the board; method lives here so we don’t invent a new process next time.