A research project, born from real friction.
LitFlow began with the friction of running systematic literature reviews — and the unresolved question of how, and whether, LLMs can and should be used in that work.
Three principles.
Show your work
Every AI suggestion carries its confidence, its justification, and its source.
Recommend, don't decide
The system is a collaborator, not an autonomous agent. Inclusion, exclusion and interpretation remain decisions belonging to the human reviewer — always.
Build in the open
Code, prompts, evaluation data and roadmap are public.
Four iterations.
LitFlow is built in echeloned cycles — each iteration framed, designed and evaluated before the next.
Problem scoping
Framing the gap: SLR tools cover parts of the workflow, but transparency and configurability of AI support are missing.
Survey
A survey of 178 researchers on SLR tooling and unmet needs grounds the design requirements.
Web platform
The end-to-end LitFlow platform — search, screen, extract and report in one document model.
Formative evaluation
A think-aloud evaluation with five information systems researchers across the full workflow.
Iteration dates to be confirmed before launch.
Where LitFlow sits.
Information Systems · University of Münster
University of Münster, Germany