Editorial summary. This is our text summary of an article. Charts, figures, and the author’s full voice are at the original — read it there .
Licence. The original is licensed under CC BY 4.0. This page adapts it: the summary and analysis are ours, not the authors’.
Peoplense's own analysis of this source, not the publisher's text. It was generated by machine from the source article and has not yet been read by one of our editors. Treat it as a starting point and check the original.
Editorial verdict
Methodologically ambitious but geographically narrow — the multi-method design (PLS-SEM, XGBoost, Choquet) is a genuine strength, yet the single-country, self-reported, cross-sectional sample means causal claims cannot be made; the social-cognitive framing of AI concealment is the most transferable insight.
Executive summary
This study investigates why consulting professionals conceal their use of generative AI (GenAI) from supervisors, colleagues, and clients, treating concealment as a social-cognitive and organizational behavior rather than a compliance failure. Using survey data from 564 consulting employees in Türkiye, the authors apply PLS-SEM, covariance-based SEM, XGBoost with SHAP interpretation, and a Choquet multi-criteria decision analysis to examine eleven constructs across 57 Likert-scale items. The central argument is that hidden AI use is primarily driven by AI disclosure anxiety and competence penalty fear — the concern that revealing AI assistance will undermine perceptions of expertise and originality — rather than being solely a compliance problem or intentional misconduct. Official AI tool adequacy, social intelligence and responsible disclosure were negatively associated with concealment; AI policy clarity and psychological safety were linked to its drivers (lower disclosure anxiety and lower competence penalty fear, respectively), and AI metacognition and human-AI trust were positively associated with responsible disclosure. Structural models explain 39.2% of hidden AI use variance and 34.3% of responsible disclosure variance. Exploratory Choquet analysis ranks a non-punitive AI disclosure system as the highest-priority organizational intervention. The authors conclude that enabling transparent AI use requires psychological safety, policy clarity, and fair evaluation frameworks, and acknowledge that the cross-sectional, single-country design precludes causal inference.
Key insights
- 1AI disclosure anxiety and competence penalty fear are the dominant drivers of concealment, explaining 32.9% and 24.2% of SHAP-attributed predictive importance respectively in the XGBoost model.
- 2Responsible AI disclosure and hidden AI use are empirically distinct behaviors; employees may selectively disclose depending on audience, task, and perceived professional risk, meaning concealment is not simply the absence of disclosure.
- 3Organizational conditions — AI policy clarity, official tool adequacy, and psychological safety — are as predictive of disclosure behavior as individual-level factors such as social intelligence and AI metacognition, pointing to systemic rather than purely individual determinants of transparency.
Practical takeaways
- Organizations in professional services contexts can reduce AI concealment by establishing non-punitive, materiality-based disclosure norms that distinguish minor technical AI assistance from substantive AI contributions — the Choquet analysis ranked this intervention as most robust across sensitivity checks.
- Evaluation frameworks that shift focus from surface output quality toward evidence traceability, verification quality, and demonstrated professional judgment may create conditions in which AI disclosure carries lower professional-image risk for employees.
Frameworks mentioned
PLS-SEM
Partial Least Squares Structural Equation Modeling — used to estimate structural path coefficients and test hypothesized relationships among eleven constructs.
Choquet Integral
A multi-criteria decision analysis method used exploratorily to rank eight managerial intervention alternatives across nine upstream criteria, capturing non-additive interactions among criteria.
References
- Administrative Science Quarterly (1999).Psychological Safety and Learning Behavior in Work Teams.
Source & Provenance
Not specified
Filiz Mizrak, Gozde Mert, Ahmet Erkasap, Ali Ozcan, Mehmet Mert
September 1, 2026
Research Study
Asia-Pacific
Original source metadata is preserved. AI analysis is generated separately.
Like this? Get the Monday Decision Brief — free, every week.
No spam, unsubscribe anytime.
Rate this article
Want the full article? Read it at the original source — free, no paywall.
Read original article