Direction, evidence grade, and study type are AI-generated labels (gpt-5-mini), not human-verified. Syntheses are LLM-written. "Tensions" are machine-detected candidates, not confirmed contradictions. A research-acceleration tool, not peer review.How this is built →
Co-authorship neighborhood
A bounded, provider-scoped projection of shared papers in this corpus. It is
not a map of importance, influence, intellectual similarity, geography, or
causal relationships.
Scroll the diagram horizontally to inspect every displayed node.
Circular placement is deterministic and only prevents overlap. Proximity and direction do not measure intellectual, geographic, social, or causal distance. Square = focal author; circles = collaborators; labels identify the focal and highest-weight neighbors.
Ordered author table
Scroll the table horizontally to see every column.
Accessible table equivalent for nodes in the displayed subgraph, ordered with the focal author first and collaborators by selected edge weight. Metrics use each node's full filtered provider scope, not only displayed edges.
Scroll the table horizontally to see every column.
Undirected co-authorship edges among displayed authors. Each pair is listed once.
Author pair
Shared papers
Full weight
Fractional weight
Paula Figueiredo — Carlos Gonçalves
1
1
0.333
Paula Figueiredo — Carlos Rouco
1
1
0.333
Paula Figueiredo — António Costa
1
1
0.333
Carlos Gonçalves — Carlos Rouco
1
1
0.333
Carlos Gonçalves — António Costa
1
1
0.333
Carlos Rouco — António Costa
1
1
0.333
Graph construction and limits
Each distinct paper contributes an undirected pair for every two eligible identities in its deduplicated latest OpenAlex author list. Self-loops are removed. Solo papers count toward an author's paper total but create no edge. The selected display is an ego subgraph. Paper, collaborator, weighted-degree, and active-span metrics for every displayed node use the full filtered provider view; component size describes the focal author's component.
Immutable provider payloads are stored once. A separate receipt log records a later delivery of an earlier payload, so roster reversions remain visible. Exact duplicate receipts are idempotent. Equal timestamps use receipt insertion order as the tie-breaker.
Co-authorship records shared paper attribution only. It does not establish contribution size, research quality, influence, friendship, topic similarity, or citation relationships.