AI evidence extraction

Ethical Judgments and Weakness of Will in the Misuse of Generative AI: The Case of Academic Essay Production among Higher Education Students

Antonio Pérez-Portabella, Jorge de Andrés-Sánchez, Mario Arias-Oliva, Graciela Padilla-Castillo · Innovative Higher Education · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

2/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1007/s10755-026-09885-6

Methodology & findings

Study design

Cross-sectional survey design with partial least squares-structural equation modeling (PLS-SEM).

Sample

N = 151, 3 groups

Primary method

Partial least squares-structural equation modeling (PLS-SEM) using SmartPLS 4.1.1.6. Analysis proceeded in steps: (1) validity assessment of measures (internal consistency via Cronbach's alpha and composite reliability for reflective constructs; convergent validity via AVE; discriminant validity via Fornell-Larcker criterion); (2) For formative construct USE_AI: VIF collinearity testing and indicator relevance assessment; (3) standardized factor scores computed and structural model estimated using percentile bootstrapping with 10,000 replications; (4) predictive power tested through Cross-Validated Predictive Ability Test (CVPAT) following Sharma et al. (2023). Power analysis conducted using G*Power 3.1.0.

Main result

The study found that "moral justice and relativism emerge as significant predictors, while contractualism shows no relevant effect" of students' intention to misuse GenAI for essay writing. Additionally, "legitimate uses of GenAI, as considered in this paper, can serve as a bridge toward misuse," with the construct of legitimate use of AI yielding β = 0.283 (p < 0.001; f² = 0.177), demonstrating "the largest effect size among all explanatory variables." The model explains "approximately 70% of the variance" in misuse intentions.

Reports effect sizes and confidence intervals.

Research paradigm

Positivist/empirical-analytical (quantitative survey with SEM analysis)

Author conclusions

"The results confirm that ethical evaluations are decisive in explaining the predisposition to misuse GenAI. Moral justice and relativism emerge as significant predictors, while contractualism shows no relevant effect." Furthermore, "the study demonstrates that legitimate uses of GenAI, as considered in this paper, can serve as a bridge toward misuse. This challenges the notion that formative and fraudulent uses belong to separate domains, and indicates instead that familiarity with the tool and the lowering of psychological barriers foster the transition toward questionable practices." The authors emphasize that "the misuse of GenAI should not be understood solely as an individual problem of dishonesty, but rather as a complex phenomenon in which ethical judgments converge with dynamics of temptation, potentially triggered by prior patterns of legitimate use of the technology."

Risk of bias

Selection bias: purposive (non-probability) sampling from two Spanish universities only; Social desirability bias: self-reported measures on sensitive topic (academic misconduct); Hypothetical scenario bias: behavioral intentions from vignette, not actual behavior; Generalizability limitations: restricted to social science students in Spain; Temporal bias: cross-sectional design cannot establish causal inference or capture behavioral evolution; Selection bias: Purposive (non-probability) sampling from two Spanish universities limits generalizability; Social desirability bias: Study relies on self-reported measures of ethically sensitive behavior; Hypothetical scenario bias: Behavioral intentions may not translate to actual behavior; Lack of randomization: Non-probability sampling prevents causal inference; Geographic/cultural bias: Limited to Spanish universities and social science programs; Temporal limitation: Cross-sectional design cannot establish causality or temporal relationships; Selection bias: non-probability purposive sampling from two Spanish universities only; Social desirability bias: though mitigated by anonymous online format, self-reported measures of sensitive behavior (academic misconduct) are inherently prone to underreporting; Hypothetical scenario bias: responses about intentions to engage in misconduct in a vignette may not reflect actual behavior; Generalizability limitations: sample restricted to social science students at Spanish universities; Lack of temporal validity: cross-sectional design cannot capture behavioral change over time

Open questions raised

Longitudinal or experimental studies needed to verify whether intentions translate to actual practices and how attitudes evolve over time
The non-significance of contractualism may reflect early stage of GenAI integration; future studies should revisit this in more mature regulatory environments
Extension to diverse disciplines and international/cultural contexts beyond social sciences in Spain
Investigation of moderating roles of psychological variables (self-regulation, tolerance of ambiguity, perceived pressure)
Alternative methodologies needed: field experiments, analysis of digital traces, faculty perceptions triangulation
Impact of GenAI on other assessment modalities (mathematics problem-solving, programming, technical reporting) beyond essays

Extracted from: pdfAgreement 53%

Explore related topics

Responsible AI Use Research Integrity Empirical Study