Tracing AI-assisted research with research-graph
AI-assisted research becomes difficult to trace once several agents start producing, reviewing, and revising the same result.
research-graph is an open-source graph-engineering tool that connects research files with the checks, roles, and revision paths around them. When a result changes, you can follow its supporting records, see who is listed as producer and reviewer, and identify the parts of the workflow connected to it. It also checks whether the recorded workflow meets its traceability rules. A passing check does not mean that a scientific result is correct.
How do you keep track of review and provenance in your own AI-assisted research workflow?