Extract the graph
Convert papers into directed causal-claim edges. This is the only stage where AI is allowed.
Deterministic metascience for economics
FrontierGraph ranks what economics should work on next.
The system turns the literature graph into a ranked research agenda. AI is used only to extract text into graph structure. The ranking itself is deterministic, inspectable, and built for busy researchers who want the evidence quickly.
856
concept nodes
30,896
economics papers
387,555
candidate missing links
Why trust it
Not a black-box recommender
FrontierGraph is designed to be legible. Users can inspect the mediators, paths, supporting papers, and prior contact behind every recommended research direction.
- AI only extracts graph structure from text.
- Ranking is deterministic once the graph exists.
- Each idea comes with evidence you can inspect.
- The tool is meant to assist judgment, not replace it.
One live example
Exchange rates -> human capital
priority
0.602
mediators
219
prior co-occurrences
7
This is the kind of result the tool is built to surface: a plausible bridge where the graph already finds many indirect routes, but the direct link remains thin in the existing literature. The point is not that the system knows the answer. The point is that it can quickly show researchers where the field may be leaving structured opportunity on the table.
How it works
A short pipeline
Score missing links
Use underexplored pairs, path support, motif completion, hub penalties, and field distance.
Inspect the evidence
Move from a ranked shortlist to mediators, paths, papers, and exportable research briefs.