← Prism overview
Prism · Worked example

When consensus is amplified.

In the same consultation, a group that looked like the clearest case of agreement turned out to be one advocacy form-letter filed a dozen times under different names. By every summary figure it read as organic support. Traceability is what gave it away.

The appearance.

One group in this docket — Support for Proposed COPPA Rule Amendments Generally, the most internally cohesive group in the run — presented as its clearest case of consensus: strong agreement, a single dominant posture, and, on the face of it, many distinct people behind it, distributed broadly rather than concentrated in a few hands. By every summary figure, it reads as organic agreement.

Cluster: “Support for Proposed COPPA Rule Amendments Generally”
Raised by 23 distinct submitters across 28 extracted positions · distributed broadly across submitters.
Support 24 Modify 3 Non-substantive 1
cohesion 0.78 · isolation 0.20 · the tightest agreement in the run

Twenty-three different names. Broadly distributed. The highest cohesion of any group. If you stopped at the figures, you would report robust, widely-held public support.

What it actually was.

Its strongest evidence was the same sentence, filed under different names. Reading the verbatim text confirms it — one advocacy organisation's form-letter template, the bodies word-for-word identical, differing only in the signatory's name and the occasional added personal line:

“I support the Commission's Children's Online Privacy Protection Act (COPPA) proposal, especially the Commission's goal to safeguard the privacy of young people.”

Filed word-for-word, among others, by:
Jessica Roer · FTC-2024-0003-0103 Arielle Leipham Ellis · FTC-2024-0003-0102 Lucia Castellano · FTC-2024-0003-0113 Mary Williams · FTC-2024-0003-0114 Mary Pliska · FTC-2024-0003-0112

A second templated line runs through the same group — “Please enact these commonsense rules.” — filed verbatim by Stephen Wishart (0111), Julia Hebner (0084) and Gregory Kimber (0031).

Coordinated commenting of this kind is common in public consultations and is not in itself improper. The analytical task is simply to recognise it — so that one drafted text filed a dozen times is not read as a dozen independent voices.

Why the obvious signals missed it.

This is the instructive part. The very figure you would reach for — the count of distinct submitters — is exactly the figure a varied-name campaign is built to defeat. Each copy carries a different name and town, so by identity alone the filings read as many independent citizens, and the "distributed broadly" check passes.

The source platform's own mass-mail marker was no more reliable: of three identical copies examined, it had been applied to only one.

What exposed the campaign was not a summary statistic but traceability — the same verbatim content, plainly visible under several different identities, which no amount of varied metadata could disguise. Because every position is tied to the exact words it was drawn from and to the comment that filed it, the repeated template is impossible to hide once you read the evidence.

The campaign was larger than one group.

Because a single comment speaks to several matters at once, the copies of this one letter did not stay in one place. The same template surfaced across more than half of the docket's topic groups — supporting one point in one group, a different point in another — so that, read group by group, it appeared as a dozen people independently endorsing several different positions. It was one text, filed repeatedly.

A reading that takes each group on its own therefore overstates not only the depth of support on any one point, but its breadth across the whole consultation. (That a single comment legitimately contributes to several groups is by design — you can see one comment span multiple topics on the overview. The same mechanism is what makes a multi-topic campaign visible.)

Breadth, corrected.

The legitimacy of a consultation rests on telling genuine breadth of opinion from amplified repetition. A tally that counts filings — or even nominally distinct names — reports robust, broadly distributed support. The defensible reading sets the one campaign's amplification aside and reports the independent remainder: a materially narrower, and more honest, figure.

Prism supports that correction through traceability. Because every position is tied both to the exact text it was drawn from and to the comment that filed it, an analyst can identify the repeated template, set it aside, and report distinct people rather than distinct names. "A dozen submitters support this" becomes a defensible, campaign-adjusted count — and the adjustment is itself auditable, since the basis for it is visible in the source.

Two cautions, the same that apply throughout. Prism reports structure, not intent: identical text under many names establishes coordination, not motive — and no claim is made about the latter. And convergence on shared language can reflect a genuine groundswell as readily as an organised campaign; the structural signal marks a candidate for review, and timing, filing patterns and provenance are what confirm it. Making coordination visible and checkable to a reviewer is what the method does today; a systematic screen across a whole docket is the active direction of development. The judgment remains the analyst's.

The receipts are the point.

A clean reading of a contested question sits in the companion example, reading one contested question; the Prism overview shows the whole method. If you're weighing large volumes of free-text responses and need to tell real breadth from amplification, we'd be glad to walk you through it.

info@szlamkaconsulting.com

Source: PRISM cluster 20260607-165457-cc24 · clustering run 20260607-165457 (k=25) · docket FTC-2024-0003. Figures reproducible from the same run.