Brief — Part VIII "Scientific Knowledge Networks" (file manuscript/sections/part08_networks.tex; bib manuscript/references_part08.bib)
Target 25–35 pages. Follow WRITER_COMMON.md. Analyse authors, institutions, countries, citations, interdisciplinarity and
knowledge diffusion (CLAUDE.md §9) using ONLY the corpus v2 result files below (copies in the repository; every number via the
macros of manuscript/tables/numbers.tex — regenerate with python3 tables/make_numbers.py . — or, when a value has no macro,
cite the result file and row in a footnote and add the value to a LaTeX table generated by tables/make_tables.py — you may
extend make_tables.py with new tables built from these CSVs, re-run it, and \input the generated tables/*.tex):
- results/bibliometrics/team_size_yearly.csv, international_collab_yearly.csv, industry_share_yearly.csv, institution_types_yearly.csv, countries_yearly.csv (fractional and full counting), country_pairs_by_period.csv (edges ≥ 50 joint works), top_institutions_by_period.csv (OpenAlex institution ids; names are not in the file — refer to ids or omit names), top_sources_by_period.csv, venue_hhi_yearly.csv, citation_windows_yearly.csv (c12/c36 windows with completeness flags; Gini), top_cited_by_year.csv (titles of the 50 most cited per year by c36 — may be quoted), indexing_lag.csv (explain why created_date cannot measure lag), dblp_only_yearly.csv.
- results/graph/field_flows_yearly.csv (references from the corpus to works inside it vs outside by OpenAlex field), cited_age_yearly.csv, reference_coverage_yearly.csv. NOTE: results/graph/ is being regenerated for corpus v2; if the files are missing or dated before 2026-11-03 18:00, write the citation-flow subsections against the figure files figures/fig_field_inflows.pdf / fig_cited_age.pdf (generated from v1) and mark the numbers \pending{v2 graph}. Category-level flows and modularity (results/graph/category_flows_yearly.csv, cross_area_share_yearly.csv, modularity_yearly.csv) do NOT exist yet: write subsection "Knowledge flows between categories" as a short placeholder paragraph marked \pending{category flows after classification} that the director will fill.
- Figures available in manuscript/figures/: fig_collaboration.pdf, fig_country_shares.pdf, fig_venue_concentration.pdf, fig_citation_inequality.pdf, fig_field_inflows.pdf, fig_cited_age.pdf (reference them with \includegraphics and \label). Hypotheses to confront (protocol P1): H4 (cross-area citation share rises; modularity falls — only the field-level part is testable now), H5 (citation inequality higher post for AI — only the aggregate Gini is available now; say so), H6 (industry affiliation share and China share rise; international co-authorship share does not rise). Report what the tables show, including null or opposite results, with the right-truncation caveats (2025–2026 windows incomplete; 2026 partial). No causal language. Methods are in Part IV (cite its labels by name). Cite the networks/bibliometrics literature from manuscript/references.bib (keys exist for Newman, Wagner, Waltman, Nielsen & Andersen, Porter & Rafols, Stirling, Leydesdorff, Priem…); verify any new key.