Home
Monitors
The Weather Report · harm · this week The Fault Line Report · polarization · this week Combined dashboard Narrative lineage Blog · all articles Religious hate, Jan–Jul 2026 Hate at the World Cup The AI fault line, Jan–Jul 2026 The immigration fault line, both ends Past weeks · every edition Methodology Hate Polarization Reliability & corrections

Digital Civility Research · longitudinal instrument

Narrative lineage: the chain record

Each row is one narrative tracked as a chain: joined week to week by a ruled key, never by text similarity, so a rename cannot break the record and a coincidence cannot fake one. Node size is the week's tier. Everything below is the real record, W29 through this week, with next week's open column on the right.

12live chains entering W36, across both instruments, each carrying a ruled history
34 · 11lineage rulings so far: 34 joins made, 11 refused on human review
5 wkslongest chain so far, renamed 4 times while its key never changed
harm instrument polarization instrument ruled join not carded, key retained ×join refused node size = tier that week
  1. 1
    The key survives renames. The Odyssey casting chain was carded under four different names in four weeks. The record held, because the join is the ruled key pol:2026-W29:pol-odyssey-casting, not the title.
  2. 2
    Change is only measurable on a chain. The UK far-right label fight rose a tier and had its logic verdict flip in the same week. Without the ruled join, both facts would be invisible: two unrelated snapshots instead of one narrative moving.
  3. 3
    Spikes die visibly. The Ceuta invasion script was the largest script we had measured, 43 items and 27 million impressions, and the chain record is what shows it vanishing a week later while the border news kept running.
  4. 4
    Refusals keep the record honest. The Michigan Senate arena produced two narratives in two weeks, corporate capture then credential fraud, and we refused to join them: different claims about the same race are different chains, and two chains record a strategy change that one chain would erase. The same test refused two hate joins in W33 on the different-predicate rule that began with the Glasgow patrols refusal. A chain is a claim genealogy, not a topic feed.
  5. 5
    The first ruled tier move on the harm instrument. Great-replacement fell P1 to P2 as its single-event carriage receded. A tier move only exists on a ruled join; without one there are just two unrelated scores.
  6. 6
    Turnover cuts both ways. W33 recorded this instrument’s first polarization continuations: two chains ruled joined and volume-measured against their own prior week for the first time, both under the five-post floor the measurement requires. Twelve of W32’s fourteen narratives did not recur. Whether that is real churn or our candidate matching failing to surface a continuation stays an open question, and the chain record is what lets us ask it.