How the claims were sorted

What counts as a claim

Anything asserted as a reason for or against the proposal, including statements of preference offered in support of a position.

Not counted: questions, procedural remarks, restatements of the proposal itself, and anything said outside the six recorded sessions.

A claim counts once however many times it was made. One of them was said in some form on nine separate occasions and appears in the record once. A count of claims is therefore not a count of support, and reading the record as a vote would be wrong.

The sorting question

Each claim was sorted by one question:

What would have to be true for this to be settled?

• Empirical if some observation would make it false.
• Value if nothing observable would, because it states what matters.
• Interest if it is settled by naming who bears a cost.

What each was answered with

Recorded as the kind of thing offered in reply: evidence, a study, data, a principle, a statement of what matters, a statement about who pays, or nothing.

Where the kind of the answer matches the kind of the claim, the exchange is recorded as answered in its own kind. Where it does not, it is recorded as crossed. Where nothing was offered, it is recorded as unanswered.

The totals

Of the nine empirical claims: two could be settled with data the organisation already held, three were answered with studies measuring something else, and for four no evidence exists either way.

Of the five interest claims: one was stated aloud, in a different session; four were never stated at all.

What is not in the record

Anything identifying. No names, no job titles beyond "the people setting the rule", no sector, no city, no figures that would locate the organisation.

Also absent: how many people held each claim, how long each session ran, and anything said privately or outside the six sessions.

The two arguable sortings

A record that admits no doubt about itself should not be trusted. Two of the twenty-four could reasonably be sorted differently, and both are flagged on the page that lists the claims in full. Moving both changes the headline split from nine-ten-five to eight-eleven-five and changes nothing about the count of crossed answers, because both were answered with a study, which cannot settle them under either reading.

How the drawings work

Each panel adds up only what it displays, and declares anything it has declined to draw. The reasons differ — a claim whose kind could not be read is left undrawn because the panel is about whether the kind of the claim and the kind of the answer agree; a non-empirical claim is left out of the evidence panel because asking what evidence would settle it is not a hard question but a wrong one; and a cost whose two shares do not divide one whole is left undrawn because that is a fault in the record rather than an unusual case.