Integrity Review: "Research on Computing Power Guarantee of the 'East Data, West Computing' Project" — DOI:10.13713/j.cnki.cssci.2024.01.006

Academic fraud report · Geng Detector

Summary

This report examines a 2024 CSSCI article by Huang Chaochun, Luo Yihong and Sun Jinliang (Guizhou Academy of Social Sciences) on computing-power guarantee for China's ten national data-center clusters. Verdict: substantive integrity concerns, treated as effectively confirmed ('实锤' in the original). The strongest evidence comes from massive misalignment between descriptive text and the underlying tables: the authors repeatedly cite standard deviations as means (e.g., 6.91 cited as the mean of the Eastern cluster's computing-power index while the mean is 45.51; 15.71 cited as a mean while the mean is 41.77); values in the text for public-service sub-index standard deviations (11.82/13.12) do not match those in Table 5 (11.40/13.07); and in the development-ecology paragraph the phrase "natural environment sub-index" is left over from a copy-paste. Additional serious problems include substituting province-/city-wide statistics for cluster-level indicators (an ecological-fallacy issue), and a missing square root in the TOPSIS distance formulas (Eqs. 8–9). A potential conflict of interest is flagged because the authors' home institution sits inside the Gui'an cluster, which ranks top in their results. Image-level checks were not possible. Confidence is high for the data/text mismatches and the unit-of-analysis problem; image forgery cannot be ruled in or out.

Verdict

🔴 Substantive concerns effectively confirmed. The article shows clear, repeated inconsistencies between reported numbers in the text and the tables the authors themselves present, indicating either careless fabrication, copy-paste assembly without review, or both. These errors are not typographical but systematic. The unit-of-analysis problem (province-level data used as cluster-level indicators) is a fundamental methodological flaw. Image-based checks were not possible.

Key findings

Descriptive text cites standard deviations as means: e.g., Eastern cluster "mean" computing-power index reported as 6.91, but Table 2 shows 45.51 (6.91 is the SD); Eastern cluster "mean" computing-level sub-index reported as 15.71, but Table 3 shows 41.77 (15.71 is the SD).
Numbers in the public-service sub-index narrative (SDs 11.82 and 13.12) do not match Table 5 (11.40 and 13.07).
The development-ecology paragraph still contains the phrase "natural environment sub-index" and the closing sentence copied verbatim from the previous natural-environment section — a clear copy-paste residue.
Unit-of-analysis substitution: indicators such as "talent reserve" (R&D full-time-equivalents of industrial enterprises above designated size in the cluster's province), "policy intensity" (policies of the cluster's province), and even temperature/distance are computed at the province or city level rather than for the cluster itself. Example: Wuhu cluster's "talent reserve" uses Anhui province's R&D FTE of 170,421 person-years.
TOPSIS distance formulas (Eqs. 8–9) omit the square root in the Euclidean distance; ranking invariance limits impact, but the omission signals that formulas were copied without understanding.
Potential conflict of interest: all three authors are affiliated with the Guizhou Academy of Social Sciences, while the Gui'an cluster (Guizhou) tops the authors' composite rankings and dominates the natural-environment and policy-intensity sub-indices.

Evidence highlights

DOI: 10.13713/j.cnki.cssci.2024.01.006
Journal: 贵州社会科学 (CSSCI), Jan 2024
Fund: 22GZZB04
Authors: 黄朝椿, 罗以洪, 孙金良 (贵州省社会科学院)
Quantitative anchors preserved from the report:
Table 2: Eastern cluster mean = 45.51; SD = 6.91 (text mislabels 6.91 as the mean).
Table 3: Eastern cluster mean = 41.77; SD = 15.71 (text mislabels 15.71 as the mean).
Public-service SDs cited as 11.82 / 13.12 vs. Table 5 values 11.40 / 13.07.
Natural-environment SDs (21.13, 16.53, 21.16) misattributed to the development-ecology paragraph.
Wuhu province-level R&D FTE = 170,421 person-years used as a cluster-level input.

Notes

Image reuse / splicing checks (Figures 1–6) could not be performed; the report explicitly flags this gap, so any image-related fraud is neither confirmed nor excluded.
The missing square root in TOPSIS distances does not change rankings but is indicative of formula copying without verification.
Conflict-of-interest concern is logged as suspected ("存疑"); the report does not assert deliberate bias.
All numerical evidence above is taken verbatim from the source report; no new findings have been added.
Final determination of academic misconduct remains with the journal, the authors' institution, and any formal investigation; AI-assisted findings can contain false positives or miss issues.