Limitations

Stated up front, to be addressed in any paper built on this framework.

Evidence maturity

Much of the supporting evidence for the LLM-as-reviewer and AI-built-dashboard components is preprint-stage. This is flagged in Related work and should be stated explicitly when citing.

AI is a complement, not a replacement

The literature is unanimous that AI here augments expert work under human oversight. DIVE-PL inherits the risks of LLM bias, hallucination, and near-verbatim reproduction of existing items; the independent-reviewer stage and human adjudication mitigate but do not eliminate them.

Case-study validity caveat

In the diving case study, VO2max is estimated via a non-exercise test, not measured. Prior fitness–adverse-event evidence used directly measured VO2max; whether the cheaper proxy reproduces that relationship is itself part of what the case study tests, and a limitation to disclose.

Scope

DIVE-PL is a methodology for building survey studies reproducibly. It does not validate any specific clinical claim; that belongs to the individual studies built with it.