The science · the casebook

The Casebook.

Every paper we run through our instruments — assessed, dated, and on the record.

01 · What it is

The literature, weighed in the open.

The Radar catalogs what patients build. The Casebook catalogs what researchers claim — and grades it.

Every article, preprint, review, and meta-analysis we assess enters the Casebook as a case: brought before the instruments and judged, with its verdict, its date, and its reasoning attached. It is the standing record behind the Standard's promise — not "we have named methods," but "here are the verdicts, on the record, where anyone can check them."

It exists because the claims made about patients — what they can understand, whether they should use AI, what happens when they do — are setting policy right now, often on evidence that never observed a patient. Those claims deserve to be weighed as carefully as the things patients actually do.

02 · How a case is made

Run through the instruments, line by line.

Each paper is read in full and run through the relevant frameworks: CAIHL to establish whose interests it serves, REAL-PAIR to test whether its patients were real patients, ASSAY to separate what it demonstrates from what it asserts, CLAIM-CSN to check whether the claims it inherits stand on evidence or on a citation cascade, and PEER-REV to judge whether it is sound and useful. Not every instrument applies to every paper; each case names the ones used, and every judgment traces back to published criteria.

03 · The verdicts

Plain, and falsifiable.

Each case carries a plain verdict and the reasoning behind it. Broadly:

  • Sound. The claims are supported by the evidence presented, and the methods fit the conclusions.
  • Mixed. Some claims hold; others outrun their evidence, or rest on proxies, small samples, or stale models.
  • Manufactured authority. The apparent consensus is a citation artifact — amplification, diversion, or invention — not a body of data.

Verdicts are dated. When a paper is superseded or a verdict revisited, the change is recorded in the open, with the original preserved. A verdict here is an argument, not a decree: if you can show where it is wrong, that correction is part of the record too.

04 · Browse

By instrument, by verdict, by claim.

The Casebook is built to be filtered — by which instrument was applied, by verdict, and by the specific claim under examination — so you can find every paper that made a given claim about patients and see how each one held up. The collection grows with every assessment we publish.

The Casebook index is being populated from our existing assessments. To submit a paper for assessment, or to contest a verdict, write to hello@patientsuse.ai.