pith. sign in

arxiv: 2604.10978 · v1 · submitted 2026-04-13 · 💻 cs.HC · cs.AI

Enabling and Inhibitory Pathways of Students' AI Use Concealment Intention in Higher Education: Evidence from SEM and fsQCA

Yiran Du , Huimin He This is my paper

Pith reviewed 2026-05-10 16:22 UTC · model grok-4.3

classification 💻 cs.HC cs.AI
keywords AI concealment intentionhigher education studentsfear of negative evaluationpsychological safetyenabling and inhibitory pathways
0
0 comments X

The pith

Perceived stigma and risk push students to hide AI use through fear, while self-efficacy and support reduce hiding by building safety.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines why university students might choose to conceal their use of AI tools in academic work. It identifies an enabling route in which concerns about stigma, risks, and unclear rules heighten fear of negative judgment and thereby increase the intention to hide AI assistance. An opposing inhibitory route shows that confidence in using AI, perceptions of fairness, and social support build a sense of psychological safety that lowers the desire to conceal. A sympathetic reader would care because these routes suggest concrete ways institutions could shift student behavior toward greater openness about AI use in learning.

Core claim

Students' intention to conceal AI use is shaped by two opposing mechanisms: an enabling pathway in which perceived stigma, perceived risk, and perceived policy uncertainty raise fear of negative evaluation and thereby promote concealment, and an inhibitory pathway in which AI self-efficacy, perceived fairness, and perceived social support raise psychological safety and thereby reduce concealment. Survey data from 1346 students confirm these links and their mediation effects through two complementary analysis methods, while also showing that different combinations of the same factors can produce concealment or non-concealment.

What carries the argument

The dual pathways mediated by fear of negative evaluation (enabling concealment) and psychological safety (inhibiting concealment).

If this is right

  • Reducing stigma around appropriate AI use would lower fear of negative evaluation and decrease concealment.
  • Clear institutional policies on AI would reduce policy uncertainty and thereby weaken the enabling pathway to hiding.
  • Increasing students' confidence in using AI tools would strengthen psychological safety and promote openness.
  • Providing social support and fair treatment would enhance safety and reduce the likelihood of concealment.
  • Multiple different combinations of these factors can lead to the same outcome of either concealment or transparency.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Universities could test whether targeted campaigns that normalize AI assistance change the balance between the two pathways in practice.
  • The emphasis on fear and safety suggests that interventions focused on emotional climate may matter more for disclosure than technical training alone.
  • The same pathways could be examined in workplace settings where employees decide whether to reveal AI assistance to colleagues or supervisors.

Load-bearing premise

That students' answers to a single questionnaire about their perceptions and plans accurately reflect the real causes behind whether they will actually hide their AI use.

What would settle it

A study that tracks the same students over several months, records their actual AI-assisted assignments through system logs or direct observation, and checks whether shifts in reported fear or safety levels predict real changes in concealment behavior would test the pathways.

Figures

Figures reproduced from arXiv: 2604.10978 by Huimin He, Yiran Du.

Figure 1
Figure 1. Figure 1: The Conceptual Model [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Structural Model Results [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
read the original abstract

This study investigates students' AI use concealment intention in higher education by integrating the cognition-affect-conation (CAC) framework with a dual-method approach combining structural equation modelling (SEM) and fuzzy-set qualitative comparative analysis (fsQCA). Drawing on data from 1346 university students, the findings reveal two opposing mechanisms shaping concealment intention. The enabling pathway shows that perceived stigma, perceived risk, and perceived policy uncertainty increase fear of negative evaluation, which in turn promotes concealment. In contrast, the inhibitory pathway demonstrates that AI self-efficacy, perceived fairness, and perceived social support enhance psychological safety, thereby reducing concealment intention. SEM results confirm the hypothesised relationships and mediation effects, while fsQCA identifies multiple configurational pathways, highlighting equifinality and the central role of fear of negative evaluation across conditions. The study contributes to the literature by conceptualising concealment as a distinct behavioural outcome and by providing a nuanced explanation that integrates both net-effect and configurational perspectives. Practical implications emphasise the need for clear institutional policies, destigmatisation of appropriate AI use, and the cultivation of supportive learning environments to promote transparency.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript investigates students' AI use concealment intention in higher education by integrating the cognition-affect-conation (CAC) framework with structural equation modelling (SEM) and fuzzy-set qualitative comparative analysis (fsQCA). Based on a survey of 1346 university students, it identifies two opposing mechanisms: an enabling pathway in which perceived stigma, perceived risk, and perceived policy uncertainty increase fear of negative evaluation leading to higher concealment intention, and an inhibitory pathway where AI self-efficacy, perceived fairness, and perceived social support enhance psychological safety thereby reducing concealment intention. SEM is used to test the hypothesized relationships and mediation effects, while fsQCA identifies multiple configurational pathways highlighting equifinality.

Significance. If the causal interpretations can be substantiated or appropriately qualified, this work makes a meaningful contribution to the literature on AI adoption in education by conceptualizing concealment intention as a distinct behavioral outcome and by demonstrating the value of combining net-effect (SEM) and configurational (fsQCA) analyses. The identification of both risk-enhancing and protective factors provides actionable insights for higher education institutions seeking to promote transparent and ethical AI use. The large sample size and dual-method approach are notable strengths that enhance the robustness of the empirical findings beyond single-method studies.

major comments (2)
  1. [Abstract and Results (SEM analysis)] The paper interprets the SEM results as confirming directional 'enabling' and 'inhibitory' pathways with mediation effects (e.g., fear of negative evaluation mediating the effect of stigma/risk/uncertainty on concealment). However, given that the data are from a single cross-sectional survey using self-reported measures, the analysis cannot establish temporal precedence required for causal mediation claims. Reverse causation or bidirectional relationships remain plausible alternatives. This directly affects the central claim of distinct pathways.
  2. [fsQCA analysis section] The fsQCA results are presented as identifying 'configurational pathways' and 'causal recipes' leading to concealment or non-concealment. Since these configurations are derived from the same cross-sectional self-report data, the causal language and equifinality claims carry the same limitations as the SEM mediation analysis, potentially overstating the explanatory power without additional validation such as out-of-sample testing or longitudinal data.
minor comments (2)
  1. [Abstract] Key statistical details such as model fit indices for SEM (e.g., CFI, TLI, RMSEA, SRMR) and consistency/coverage thresholds for fsQCA are not reported in the abstract, which would help readers assess the strength of the evidence immediately.
  2. [Discussion] The manuscript could benefit from a more explicit discussion of potential common method bias and social desirability bias inherent in self-reported concealment intentions, and how these were mitigated (if at all) through procedural remedies or statistical controls.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed and constructive comments on our manuscript. We particularly appreciate the emphasis on the limitations of causal inference in cross-sectional data. We have revised the manuscript to qualify our interpretations accordingly while preserving the core contributions.

read point-by-point responses

  1. Referee: [Abstract and Results (SEM analysis)] The paper interprets the SEM results as confirming directional 'enabling' and 'inhibitory' pathways with mediation effects (e.g., fear of negative evaluation mediating the effect of stigma/risk/uncertainty on concealment). However, given that the data are from a single cross-sectional survey using self-reported measures, the analysis cannot establish temporal precedence required for causal mediation claims. Reverse causation or bidirectional relationships remain plausible alternatives. This directly affects the central claim of distinct pathways.

    Authors: We fully agree that our cross-sectional survey data cannot establish causal relationships or temporal precedence for the mediation effects. The CAC framework provides a theoretical basis for the proposed directions, but we recognize that the empirical analysis shows associations consistent with these pathways. In the revised version, we have modified the abstract and results sections to use more cautious language, such as 'suggests' and 'is associated with' instead of 'confirms' and 'increases'. We have also expanded the limitations section to explicitly discuss the inability to rule out reverse causation and the need for longitudinal studies to validate the directional claims. This revision addresses the concern without altering the empirical findings. revision: yes

  2. Referee: [fsQCA analysis section] The fsQCA results are presented as identifying 'configurational pathways' and 'causal recipes' leading to concealment or non-concealment. Since these configurations are derived from the same cross-sectional self-report data, the causal language and equifinality claims carry the same limitations as the SEM mediation analysis, potentially overstating the explanatory power without additional validation such as out-of-sample testing or longitudinal data.

    Authors: We acknowledge the referee's point that fsQCA, like SEM, is based on cross-sectional data and thus identifies configurations associated with the outcomes rather than proving causal recipes. We have revised the fsQCA section and abstract to replace 'causal recipes' with 'configurational pathways' or 'combinations associated with', and clarified that equifinality refers to multiple sufficient configurations in the observed data. We note that while fsQCA does not require additional validation like out-of-sample testing in the same way as predictive models, we have added a call for future research using longitudinal or experimental designs to further substantiate these patterns. These changes ensure the claims are appropriately qualified. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical analysis of survey data via standard SEM/fsQCA methods

full rationale

The paper applies the established cognition-affect-conation framework to survey responses from 1346 students, testing hypothesized enabling and inhibitory pathways through structural equation modelling for mediation effects and fsQCA for configurational patterns. All load-bearing claims rest on statistical fits to observed self-report data rather than any definitional equivalence, fitted parameter renamed as prediction, or self-citation chain that reduces the result to its own inputs. No equations, uniqueness theorems, or ansatzes are invoked in a manner that creates circularity; the derivation chain is the standard empirical workflow of hypothesis testing on independent measurements.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The claim depends on standard assumptions of structural equation modeling and fuzzy-set analysis plus the validity of self-reported psychological constructs; no new entities are postulated.

free parameters (2)
  • SEM path coefficients and mediation weights
    Fitted to the 1346-student dataset to quantify the enabling and inhibitory pathways.
  • fsQCA consistency and coverage thresholds
    Chosen to identify configurational solutions showing equifinality.
axioms (2)
  • domain assumption Linear relationships and multivariate normality assumptions hold for the structural equation model
    Invoked by the use of SEM to test hypothesized relationships and mediation effects.
  • domain assumption Self-reported survey responses accurately reflect latent psychological constructs such as fear of negative evaluation and psychological safety
    Required for both SEM and fsQCA to map measured variables to the proposed pathways.

pith-pipeline@v0.9.0 · 5500 in / 1566 out tokens · 43576 ms · 2026-05-10T16:22:38.719347+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages

  1. [1]

    Introduction Artificial intelligence (AI) is rapidly reconfiguring higher education by changing how students learn, produce academic work, and respond to assessment demands. Generative AI tools can enhance efficiency, autonomy, and personalised learning, while educators are increasingly adopting AI for adaptive instruction, automated feedback, and learnin...

    work page 2025

  2. [2]

    Literature Review 2.1 AI in Higher Education Artificial intelligence (AI) has rapidly become a transformative force in higher education, reshaping teaching, learning, and assessment practices (Qian, 2025). Tools such as generative AI systems enable students to access instant explanations, generate academic content, and personalise learning pathways, there...

    work page 2025

  3. [3]

    Theoretical Framework and Hypothesis Development 3.1 Cognition–Affect–Conation Framework The cognition–affect–conation (CAC) framework provides a useful theoretical lens for explaining how individuals’ cognitive appraisals shape affective responses, which in turn influence behavioural intention (Zhou & Zhang, 2024). In the present study, cognition is refl...

    work page 2024

  4. [4]

    The Conceptual Model 3.2 The Enabling Pathway of AI Use Concealment Intention Fear of negative evaluation is conceptualised as the central affective mechanism in the enabling pathway, referring to individuals’ apprehension about being judged unfavourably by others (Yue et al., 2022). Grounded in social evaluation theory and extensively validated in educat...

    work page 2022

  5. [5]

    Methods 4.1 Research Design The study adopted a quantitative, cross-sectional research design to examine the enabling and inhibitory mechanisms underlying AI use concealment intention among university students. Grounded in the cognition–affect–conation framework, the research employed a survey-based approach to collect self-reported data on multiple laten...

    work page 2023

  6. [6]

    SEM was conducted using a two-step approach to assess the measurement model and the structural model

    and fuzzy-set qualitative comparative analysis (fsQCA) (Schneider & Wagemann, 2012). SEM was conducted using a two-step approach to assess the measurement model and the structural model. First, confirmatory factor analysis (CFA) was performed to evaluate reliability and validity, including indicator loadings, Cronbach’s α, composite reliability (CR), and ...

    work page 2012

  7. [7]

    Discussion 6.1 The Enabling Pathway of AI Use Concealment Intention The findings of this study provide empirical support for the enabling pathway proposed within the cognition–affect–conation framework, demonstrating that fear of negative evaluation serves as a critical affective mechanism linking threat-related cognitive appraisals to AI use concealment ...

    work page 2025

  8. [8]

    Conclusion This study advances understanding of AI use concealment intention in higher education by integrating the cognition–affect–conation framework with both symmetric (SEM) and configurational (fsQCA) analyses. The findings demonstrate that concealment intention is shaped by dual mechanisms: an enabling pathway in which perceived stigma, risk, and po...

    work page 2025

  9. [9]

    H., & Alkhwaldi, A

    https://doi.org/10.1007/s44217-024-00393-3 Alshammari, S. H., & Alkhwaldi, A. F. (2025). An integrated approach using social support theory and technology acceptance model to investigate the sustainable use of digital learning technologies. Scientific Reports, 15(1),

    work page doi:10.1007/s44217-024-00393-3 2025

  10. [10]

    A., Algodi, H., Costello, J

    https://doi.org/10.1038/s41598-024-83450-z Ans, M., Maggio, L. A., Algodi, H., Costello, J. A., Driessen, E., Oswald, K., & Lingard, L. (2026). The presence and nature of AI-use disclosure statements in medical education journals: A bibliometric study. Perspectives on Medical Education, 15(1), 212–225. https://doi.org/10.5334/pme.2431 BaHammam, A. (2025)....

    work page doi:10.1038/s41598-024-83450-z 2026

  11. [11]

    https://doi.org/10.3390/bs15040449 Li, S., Fan, H., Wang, Z., & Zhao, Q. (2025). Exploring the relationship between climate policy uncertainty perception and green technology innovation in Chinese enterprises. Economic Analysis and Policy, 86, 880–892. https://doi.org/10.1016/j.eap.2025.03.053 Lind, E. A. (Ed.). (2019). Social psychology and justice (1st ...

    work page doi:10.3390/bs15040449 2025

  12. [12]

    https://doi.org/10.2307/257999 Newton, P. M. (2025). How vulnerable are UK universities to cheating with new GenAI tools? A pragmatic risk assessment. Assessment & Evaluation in Higher Education, 50(8), 1332–1343. https://doi.org/10.1080/02602938.2025.2511794 Nguyen, H. M., & Goto, D. (2024). Unmasking academic cheating behavior in the artificial intellig...

    work page doi:10.2307/257999 2025

  13. [13]

    Look me in the eye, customer

    https://doi.org/10.3390/bs15050675 Ozuna, E., & Steinhoff, L. (2024). “Look me in the eye, customer”: How do face-to-face interactions in peer-to-peer sharing economy services affect customers’ misbehavior concealment intentions? Journal of Business Research, 177, 114582. https://doi.org/10.1016/j.jbusres.2024.114582 Pinho, M., & Colston, B. (2025). Perce...

    work page doi:10.3390/bs15050675 2024

  14. [14]

    https://doi.org/10.1186/s12888-024-06454-z Zhou, T., & Zhang, C. (2024). Examining generative AI user addiction from a C-A-C perspective. Technology in Society, 78, 102653. https://doi.org/10.1016/j.techsoc.2024.102653 Appendices Appendix A. Participant Characteristics (N =

    work page doi:10.1186/s12888-024-06454-z 2024