DIVE-PL

A reproducible AI tool-chain methodology for building survey research studies.

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.