Explore connections between fields.

Both maps use the same 11,339 eligible source records from 2019–2023 as the main TRACE view, or the same 201 provisionally title-screened AI records when AI focus is selected. A year filter narrows those source records. No newly discovered candidate is added. Relevance screening, including separation of technical trust and irrelevant uses of “trust”, remains pending.

The citation map reproduces the headline Research connections indicator: 31,751 cross-field citations / 81,570 mapped citations = 38.9 per 100. AI focus reproduces 1,117 / 1,702 = 65.6 per 100. The denominator includes within-field citations, which are not drawn. Citation targets come from eligible cached records published from 1990 through the source paper's year, including targets outside the selected years and outside the AI subset. Only part of each source paper's bibliography is covered.

The shared-paper map counts each distinct field once per paper, and each unordered pair of its fields once. A paper may support several nodes and pairs; these counts do not sum to the cohort size. All retained OpenAlex topic fields are used, without an extra topic-score cutoff. Low-scoring alternative topic labels can create links. This exploratory map is not a new headline score.

These are broad OpenAlex fields, not the Web of Science research areas in the published map. The network groups are weakly separated (modularity Q = 0.083 for shared-paper links and 0.061 for citations), so the colours should not be read as firm disciplinary divisions. Cluster colours are calculated from the full 2019–2023 network using weighted greedy modularity at resolution 1. The force layout uses seed 42. Colours and positions stay fixed within each relationship while filters change weights; group numbers are not comparable between the two maps. Distances are illustrative. Line widths and circle radii use a square-root scale; circles have a minimum visible radius. No field or pair threshold is applied in the full map.

Metadata overlap and citation links do not establish integration of methods, knowledge co-production, or the quality of human or AI trust. Human assessment remains necessary for a transdisciplinary claim. Read the indicator methods.