Resources: methods notes
Short, plain notes on the methods behind IAMB.AI, for your own methods section and for the reviewer who asks.
Notes
Every box in the PRISMA 2020 flow diagram is counted from the review’s own records: identified per source, duplicates removed, screened, excluded, sought, not retrieved, assessed and included.
Records excluded by the AI on its own are counted as excluded by automation tools. Records found by citation searching get their own boxes. Because the counts come from the records, the diagram always adds up.
You write the search once. The app parses it and writes it in each database’s own syntax by rule, runs it where the database has an API, and hands you the query for the rest.
Each run is logged, and the log goes into the methods for PRISMA-S.
For each record the app tries a cascade of legal sources in turn: the source’s own PDF, OpenAlex, Semantic Scholar, arXiv, Unpaywall, Europe PMC, CORE, and IEEE and ACM open access. An AI web search is optional.
Closed papers get an OpenURL link to your library. Shadow libraries are blocked. Whatever is still missing is marked not retrieved.
Cohen’s kappa is taken pairwise on each reviewer’s independent vote, recorded before any discussion. Changing your mind in a conflict discussion does not change your kappa.
The manuscript reports the basis of the kappa, so readers know exactly what was measured.
IAMB.AI is invite-only for now. Say who you are and what you want to research.
Assistant
AI answers. A person can step in.
AI answers can be wrong. We keep conversations for 90 days.