Dataset — Differentiated Emphasis, Shared Silences: A Tension-Aware Academic Balanced Scorecard Analysis of the Strategic Plans of the World's Top-50 Universities.

This deposit contains the coding outputs, derived datasets, codebook, and analysis script required to reproduce the quantitative findings of the study. It deliberately excludes the verbatim texts of the coding units: these are excerpts from copyrighted institutional documents that are publicly available at the sources listed in the source register, and the segmented corpus is available from the corresponding author upon reasonable request for verification purposes. Contents: 01_perspective_votes.csv — Perspective votes per unit (KIMI K2.6, Manus 1.6, Fable 5) and the strict-majority result (3,982 units). 02_leadlag_votes.csv — Lead–lag votes and HIGH/LOW confidence flags (KIMI 2.6, Manus 1.6, GLM 5.1) with majority result. 03_tension_votes.csv — Strategic-tension votes (0, T1–T4) and confidence flags with majority result. 04_university_compositions.csv — Per-university perspective counts, proportions, balance index, ilr coordinates, and cluster labels (50 universities). 05_human_validation_sample.csv — The 200-unit blind human validation sample (fixed random seed = 42). 06_source_register.csv — Register of the 58 analyzed institutional documents (QS 2027 rank, university, country, document title, source URL); documents sourced from official university websites as of July 2026. CODEBOOK.md — Full coding rules for all three dimensions and variable definitions. reproduce.py — Recomputes the headline statistics (reliability coefficients, validation agreement, compositional means, ilr variances, clustering and the Foundation-dominant archetype, lead–lag and tension distributions) from the deposited files. Requires Python ≥ 3.10 with pandas, numpy, scipy, scikit-learn. Notes: All stochastic procedures use fixed seeds (sampling seed = 42; k-means seed = 42, 100 initializations). Unit identifiers are opaque and consistent across all files. Data files and codebook: CC BY 4.0; reproduce.py: MIT License.