Review Report on 'Network Analysis of Dark Personality Traits in Adolescents' (青少年暗黑人格的网络分析)

Academic fraud report · Geng Detector

Summary

This report evaluates a 2025 article published in 心理研究 (Psychological Research) by Ren Fen and Ding Yicong (DOI: 10.19988/j.cnki.issn.2095-1159.2025.05.006). The overall verdict is 'Questionable' (yellow flag). The key finding is a basic graph-theory/arithmetic error: the authors state that 12 nodes in a network can form 60 unique edges, but the correct maximum number of edges in an undirected complete graph with n=12 nodes is n(n−1)/2 = 66, not 60. This represents a clear mathematical mistake rather than deliberate fabrication, but it indicates insufficient rigor in data handling for a published academic paper. A second concern regarding a potential mismatch between zero-order correlations in Table 1 (e.g., psychopathy–Machiavellianism r = 0.575; narcissism–Machiavellianism r = 0.502) and the sparser network visualization in Figure 1 was deemed inconclusive, as LASSO-regularized partial correlations commonly produce such patterns. Confidence is high on Finding 1; Finding 2 remains unverified due to insufficient evidence.

Verdict

🟡 Questionable — The paper contains a verified mathematical error but no conclusive evidence of academic fraud.

Key findings

Finding 1 (Verified, Yellow): Incorrect edge count for a 12-node network. Authors claim 60 edges; the correct maximum in an undirected complete graph with n=12 is 66 (formula: n(n−1)/2 = 12×11/2).
Finding 2 (Unverified, Yellow): Apparent tension between high zero-order correlations in Table 1 (psychopathy–Machiavellianism r = 0.575; narcissism–Machiavellianism r = 0.502) and the relatively sparse edge density in Figure 1. Insufficient evidence for fraud because LASSO regularization of partial correlations routinely produces this pattern.

Evidence highlights

Location of Finding 1: Results section, p. 433. Direct quote from the article: "在量表中有12个题项，因此在网络中有12个节点，任意两点之间连接，共可以形成60条独特的边线" (12 items → 12 nodes; any two points connected → 60 unique edges claimed).
Correct calculation: For n = 12 nodes, the maximum number of edges in an undirected complete graph = 12 × 11 / 2 = 66 edges. The reported figure of 60 is off by 6 edges.
Location of Finding 2: Table 1 (p. 433) and Figure 1 (p. 434). Zero-order correlations among the three Dark Triad dimensions are notably high (0.502–0.575), yet the network visualization does not display the extreme edge clustering one might naively expect from raw correlations.
DOI: 10.19988/j.cnki.issn.2095-1159.2025.05.006

Notes

Finding 1 most likely reflects a computational or typographical error rather than intentional manipulation, but the publication of such a basic graph-theory/math error suggests insufficient review of the results section.
Finding 2 is statistically explicable: psychometrics network analysis typically uses partial correlations regularized via LASSO (or EBICglasso), which shrinks weak edges to zero. High zero-order correlations do not necessarily translate into dense partial-correlation networks.
No evidence of image manipulation, data fabrication, or plagiarism was identified in this review.
Recommended actions: authors should correct the edge count (66, not 60) and ideally publish an erratum; the partial-correlation matrix could be provided to dispel any lingering doubt about the figure–table consistency.
This report is AI-assisted and intended for academic discussion only; final determinations of misconduct must come from official institutional investigations.