Confidence Levels

Every group profile on CLCI Hub carries a confidence rating — High, Medium, or Low — alongside its numeric score. The confidence rating reports how strong the underlying evidence is. This page explains what each tier means and what readers should infer from it.

Why confidence matters as much as the score

A score of 32 with Low confidence is meaningfully different from a score of 32 with High confidence. The first says “based on currently available public material, this group's operational pattern appears to fit the destructive-control band, but the evidence is thin and a substantive new primary source could move the score in either direction.” The second says “based on multiple corroborating court records, peer-reviewed work, and substantial journalistic investigation, this group's pattern is documented to fit the destructive-control band; a single new source is unlikely to materially change the rating.”

Confidence is a separate dimension because it answers a separate question: not how controlling is the operation but how well do we know what the operation actually does. Some of the most controlling groups in the world are also the most documented (the public-record trail is substantial); some are the least (small, closed, or jurisdictionally protected). The score reports the rating; the confidence reports the evidence base.

High confidence

High confidence applies where the public record contains multiple substantial primary sources plus corroborating secondary work. A typical High-confidence profile carries:

Court records, government investigation reports, or regulator findings on at least one substantive claim;
Peer-reviewed academic work analysing the group's practices;
At least one long-form investigative journalism piece from a major outlet with editorial standards;
Named, on-the-record ex-member testimony corroborating operational patterns;
The organisation's own published doctrine on relevant control mechanisms (which can be quoted directly).

High confidence does not require every BITE-axis claim to be court-attested. It requires that the overall operational pattern is sufficiently documented across multiple independent sources that a new single source is unlikely to materially shift the rating. The Church of Scientology, Jehovah's Witnesses, FLDS, NXIVM, and Aum Shinrikyo are illustrative examples of High confidence: court findings, government investigations, academic studies, and substantial investigative reporting all exist in the public record.

Medium confidence

Medium confidence applies where credible secondary work documents the operational pattern but the primary-source record is incomplete or partially contested. A typical Medium-confidence profile carries:

Substantive investigative journalism from established outlets;
Books or documentaries from major publishers;
Named ex-member testimony, sometimes including court declarations on adjacent matters;
Limited or no peer-reviewed academic study;
Court records on some claims (e.g. civil settlements, restraining orders) but no criminal verdicts or government inquiries.

Medium confidence is the modal rating across the dataset. It honestly reflects the state of public-record evidence for many high-control groups: there is enough credible secondary documentation to score the operational pattern, but not enough primary-source weight to call the rating settled.

Low confidence

Low confidence applies where the available evidence is fragmentary, anecdotal, single-source, or actively contested. A typical Low-confidence profile carries:

Limited or no peer-reviewed academic study;
Sparse mainstream journalism;
Primarily ex-member testimony, often anonymous or single-source;
The group operates in a jurisdiction or language community where the public-record trail is harder to access;
The group is small, closed, or recently emerged.

Low-confidence ratings are explicitly provisional. They represent our current best summary of the documented pattern given the evidence we have been able to surface. A substantive new primary source — a court filing, a government inquiry, a major investigative piece — can move a Low-confidence rating up or down significantly. We mark Low-confidence profiles on the header and invite correction submissions from anyone with access to additional public-record material.

What confidence is not

Confidence is not a statement about the group's honesty. A Low-confidence rating does not imply the group is hiding anything; many ordinary religious organisations would have Low-confidence ratings if we tried to score them, simply because no one has investigated them.
Confidence is not a confidence interval in the statistical sense. We are not claiming a 95% probability that the true score lies within ±n of the published score. The rating is editorial: it describes how settled the evidence base feels rather than a calculated uncertainty band.
Confidence is not the same as score band. A group can have High confidence in the Minimal band (we have a lot of documentation that the group does not exhibit control patterns) just as readily as in the Destructive band.
Confidence does not adjust the score. The numeric CLCI is calculated from the BITE axes and Modifier per the BITE model page; confidence is reported alongside but does not change the arithmetic.

How confidence changes over time

Confidence ratings move as the public-record trail grows or thins. New court proceedings, government inquiries, peer-reviewed studies, or major investigative pieces can move a profile from Medium to High. Retractions, successful appeals, or successful defamation defences by the organisation can move it the other direction. Where a profile has been reassessed, the confidence change is logged in the public changelog alongside the date and the source that drove the reassessment. See Score Updates for the changelog mechanics.

What readers should do with the confidence rating

Read confidence first, then score. If a profile is High-confidence and Extreme band, the published material on coercive control is unusually substantial and a sceptical reader can verify the core claims via the linked sources. If a profile is Low-confidence and Moderate band, the rating should be treated as a starting hypothesis for further reading rather than as a settled summary.

Where confidence is Low, we particularly welcome additional sources from people who have direct knowledge — current or former members, journalists who have reported on the group, academics who have studied it. The Corrections route is the standard intake.