Investigate expert judgment
What turns a practiced observation into a claim others can inspect?
Read at your own pace; optional source checks take longer
Start with these sources
Your first move: Open What Does a Hand Know? and write down the starting assumption before reading the evidence.
Without AI
- Define one judgment, the people it concerns and what would count as a useful or accurate finding.
- Open original sources; record exact methods/results locations and anything you cannot retrieve.
- Compare a starting claim with the evidence, write a revision and state what would change it.
With GenAI
Bring your own AI tool and the source material. This atlas does not send your data anywhere or run an AI model.
- Give your AI tool the bounded question and ask for primary sources with identifiers; verify that each exists.
- Supply inspected passages and ask it to distinguish agreement, validity, utility and causal explanation.
- Challenge its strongest conclusion, inspect the cited passage yourself and retain the reason for your revision.
A prompt to make your own
Using only the supplied question and source passages, distinguish measured observations, interpretations and missing evidence. For each claim give the source identifier and exact section/table. Separate within-examiner repeatability, between-examiner agreement, reference validity, decision usefulness and learning over time. Do not infer individual benefit from a mean. Offer a strong counterargument and a revision trigger. Do not invent testimony or offer medical advice. Question and passages: [Paste here.]
Make something: A claim-and-revision record and an argument that names its strongest objection and revision trigger.
Keep your footing
- A convincing explanation still needs evidence suited to its scope.
- A search portal, a reference method and a care comparison answer different questions.
- A fictional teaching table cannot establish clinical effectiveness.
Why follow this question?
Start with a patient’s question
What Does a Hand Know? is a prepared Patient / AI collaborator conversation. It shows questions and revisions before presenting a completed essay. Both voices are authored; no clinical history or practitioner testimony is invented.
Two kinds of learning to investigate
Across patients, ask how expectations are checked and corrected over years. Within one patient, ask what later observations change about the first interpretation. A proposed connection between those loops is a question to test, not a result to assume.
Without AI
Use a paper notebook with columns for starting claim, source location, wording now and unresolved questions. The cited studies and the synthetic exercise support the same reading process without any AI service.
Go a little further
Take the same questions to another field of expertise while finding that field’s own evidence.
Worked examples: How do you know what to do just by touching my back?