The method
A conclusion should
show its working.
Sources, studies, claims, and assumptions have different jobs. Keeping them distinct makes it possible to see where evidence ends and interpretation begins.
- 01SourceA stable record
- 02StudyThe underlying experiment
- 03ClaimWhat the source supports
- 04AssumptionWhat analysis requires
- 05ReviewWhat has been checked
- 06VersionA replayable record
This is the organizing workflow. The prepared demonstration uses curated source records; it does not automate every step.
Trace
Count studies carefully.
Keep reports connected.
A new publication may describe a new experiment, a follow-up, or an existing sample. These are different kinds of evidence.
The current packet preserves explicit source-to-claim links. Reviewed study and sample relationships are a development priority. Shared authors alone do not establish shared participants or independence.
Follow a source-linked claimExample from the prepared packet
Saved result · independent review pendingSource-linked claim
The saved primary analysis pools 12 trials: Hedges’ g −0.90, with a 95% confidence interval from −1.26 to −0.55.
Limit recorded beside the claim
Negative values favour psilocybin. The dataset combines several clinical populations and controls; this is not an MDD-only effect, a durability estimate or a treatment recommendation.
Challenge
Make the assumption
part of the result.
A result can be precise under one set of choices and fragile under another. The interface should show which choice changed and which conclusion it affects.
In this preview, the PsychDoc control replays an assumed additive bias shift against a saved result. It does not estimate actual bias, refit a model, or assign a probability that a claim is true.
Replay the saved analysisA bounded demonstration
Hedges’ g- Saved result
- Assumed shift
Negative values favour psilocybin. The original interval is retained and translated with the assumed shift.
Transfer
The same record.
Different scientific constraints.
Source identity, claim links, preparation history, and review status can travel across domains. Statistical compatibility has to be established for each question.
The materials packet retains conditions such as illumination, temperature, and the stability endpoint. No pooled result is presented when the recorded inputs do not justify one.
Compare the materials conditionsPsychDoc
Saved statistical replay
One pinned psilocybin analysis; assumption control with declared limits.
- Claims
- 3
- Sources
- 4
Perovskite stability
Condition comparison
Different test contexts remain visible; no pooled statistical endpoint.
- Claims
- 5
- Sources
- 5
A shared presentation contract and existing source-reference checks have been exercised with both packets. This does not demonstrate general extraction or quantitative portability. Download the check receipt.
What exists.
What comes next.
The interface is only useful if its status is as inspectable as its sources.
Two curated source packets, claim-source links, condition comparison, versioned downloads, and a bounded saved-analysis replay.
Independent verification of scientific inputs and methods. A clear presentation is not evidence that this review has happened.
Reviewed study/sample relationships, source-linked assumption boundaries, and evidence-change impact. These are directions, not shipped guarantees.
Inspect the record.
Use the two packets to see what stays consistent and where the science needs different treatment.