How to Use This Site

We track online harm and polarizing narratives in US, Canadian, UK, and English-language European discourse, weekly, with a published method. This page is for the people we built the record for: researchers, journalists, and educators.

We do not score rudeness, volume, protest, or disruption. The harm instrument tests for speech that targets protected groups; the Fault Line instrument measures framing between political camps. Neither codes incivility itself.

The output is a record: which narratives ran in our collection, who carried them, what they scored, and how each claim fared against its own predictions.

What we publish, and when

Two instruments, every week, covering Monday through Sunday, published at the week’s end.

The Weather Report tracks harm to protected groups. We collect first-party material from a standing roster of accounts, plus news, forums, and Bluesky. Every quoted item is verified against a fetchable primary source before coding. Each confirmed item receives one of five harm codes on a closed ladder, and confirmed items cluster into scripts. Each script then receives a separate severity score from 0 to 5 under the published formula.

The Fault Line Report tracks polarization. Camp-versus-camp narratives become cards. Each card states both frames, then runs the claim against its own predictions and records the verdict: supported, mixed, contradicted, or not testable. The verdict describes how that week’s evidence fits the claim’s predictions, not whether the broader political position is right. Each card also carries a mirror check: whether an opposing camp runs a materially comparable move, and whether either camp answers the other. A card can find no counterpart, and a counterpart does not establish equal reach or severity. Charge is scored 0 to 5.

The two scores are separate measures under separate formulas. Neither estimates platform-wide prevalence or audience size.

The combined dashboard puts both instruments on one screen. The narrative lineage page joins narratives week to week by ruled keys, so persistence, tier movement, and disappearance are recorded within the collection rather than remembered as impressions. Special reports go deeper on one fault line.

What a card gives you

A card is not a take. It carries the claim in the claimant’s own words, the cited primary sources with dates, the factor-by-factor basis for the score, the logic test, and the mirror check. Where a reading is ours, it is labeled as ours and the reasoning is on the card. Each report’s Limitations section lists the gaps we know about.

Published editions do not change after they ship. Corrections are logged on the reliability page, not silently edited in. If you cite a week, that week stays citable; check the reliability page for corrections attached to it.

For researchers

The scoring formulas, vocabularies, and coding rules are published on the methodology pages, hate and polarization each. The harm ladder is a closed five-value scale. Coordination verdicts come from a closed set with a documented evidence bar. Country tags come from the trigger event’s arena, not the carrier’s passport.

The lineage record is what makes week-over-week claims checkable. Joins between weeks are ruled on identity, not text similarity, and recorded refusals sit beside the joins. When we say a narrative persisted three weeks and rose a tier, that is a chain of ruled keys you can audit.

What we do not have: platform-representative sampling. The roster is outcome-selected and carried exhaustively. That supports rate and trajectory claims within the roster, not prevalence claims about a platform. The Limitations section of each report states what the collection covered that week.

For journalists and fact-checkers

Every quoted item in either instrument carries a primary-source URL and a confirmed date, and was read in context before coding. Links can die after publication; the quote and the date were verified before it. You can check our work before you repeat it, and you should.

The coding default is that political speech does not harm a protected group. An item receives a harm code only when it clears a documented bar, and the code is an analytical judgment, not a legal one. Officials speaking in official capacity are read as speaking for their camp, which raises reach and responsibility in scoring and never lowers it. Private individuals are not treated as representatives of any group. The record analyzes narratives; it is not a list of people to go after.

If your reporting argues for something, a card gives you a documented starting point: the evidence, the date, and the counter-frame arrive attached to the claim. Verify the primary source before you print it.

For fact-checking, the lineage page answers the question a claim’s age raises: when we first observed a narrative, what it was called then, and what changed since. Our verdicts are instrument-specific readings of a claim against its own predictions. They are inputs to a fact-check, not substitutes for one.

For educators

A narrative card is a worked example of applied source criticism, a few hundred words long: a claim, the test the claim sets for itself, and a verdict. The mirror check runs the same skepticism in both directions, including at the side a student agrees with. The weekly reports work as current-events material with the sources attached and checkable, and the lineage chart shows narratives with lifespans: they appear, move, get renamed, drop out of the record, and sometimes come back.

Students do not need to be told what to think about a topic to use any of this. They can check the sources, the rules, and our judgment calls, and the coding is part of what is open to challenge. The methodology pages can carry a full unit on testing claims.

A word of caution: the Weather Report quotes charged language verbatim, slurs and dehumanizing terms included, because paraphrase would hide what was said. Review a card before assigning it.

For trust-and-safety and civil-society teams

The record can serve escalation tracking and retrospective review: when a narrative first appeared in the collection, how it moved, when its ignition markers accelerated, and what the coordination evidence actually was. The ignition watch and the lineage chains are the parts built for that reading.

The limits are hard ones. A score does not establish a policy violation, intent, imminent risk, or coordination beyond the documented verdict on the card. Nothing on this site arrives ready to act on.

The caveat over all of it: we are strictly non-partisan and non-interventionist. We do not advocate silencing, removing, or deplatforming anyone. We present the data as it was discovered and adjudicate it consistently under a published methodology. It is useful context. It is not a decision.

Sharing a card

Every narrative card has a share control. The link is stable, carries its own preview image, and opens the card inside the full week’s report, so the evidence and the score arrive together. Share the link, not a screenshot: a screenshot detaches the claim from its sources, scores, and any later corrections.

Start here

• The current Weather Report and Fault Line Report, also one click from the gauges on the homepage.
• The combined dashboard.
• The narrative lineage record.
• Methodology: hate and polarization.
• Reliability and corrections.