International Journal of Management and Organizational Research
The Role of Artificial Intelligence System Reliability and Transparency in Reducing Algorithm Aversion and Enhancing Auditors' Trust in AI Outputs
- 1
This is an author-deposited copy. The version of record was originally published elsewhere: Originally published in International Journal of Management and Organizational Research. ISSN 2583-6641. Volume 5 Issue 5. Pages 17-23. Published 2026-09-01. DOI 10.54660/IJMOR.2026.5.5.17-23 . Source: https://www.themanagementjournal.com/article/1443/the-role-of-artificial-intelligence-system-reliability-and-transparency-in-reducing-algorithm
Abstract
Artificial intelligence (AI) is being used in audit processes. However, auditors may ignore algorithmic advice, a condition known as algorithm aversion. This study examines the association between the perceived dependability and transparency (explainability) of standard artificial intelligence systems and their effects on algorithm aversion and trust in AI outputs; and the association between trust and auditors’ desire to rely professionally. The questionnaire was issued to 140 external and internal auditors and consisted of 25 items, measuring five constructs on a 5-point Likert scale. All constructs showed good internal consistency (Cronbach's α = 0.848–0.882). Reliability and transparency were negatively related to algorithm aversion (β = −0.29 and −0.30) and positively related to trust (β = 0.27 and 0.31), while aversion was negatively related to trust (β = −0.28). Trust had the strongest standardized coefficient in the professional-reliance-intention model (β = 0.40). The models explained 24.7%, 45.6%, and 38.6% of the variance in aversion, trust, and reliance intentions, respectively. The percentile bootstrap analysis (5,000 resamples) revealed favorable indirect associations between reliability and transparency with reliance intention through trust. The confirmatory factor analysis (CFA) indicated a good fit (χ2(265) = 283.74, p = 0.205; CFI = 0.99; TLI = 0.99; RMSEA = 0.02). Composite reliability was ≥ 0.85, average variance extracted was ≥ 0.53, indicating a good convergent validity. The Fornell–Larcker criterion confirmed the discriminant validity. Analyses of diagnostics suggested sample adequacy (KMO = 0.90), no multicollinearity (maximum VIF = 1.84), and a Harman single factor of 37.3% for an initial assessment of common-method bias. All 8 hypotheses were empirically supported. Our results show that trust mediates the relationship between dependable and explainable AI design and auditors’ intention to rely on AI outputs, but the cross-sectional design does not allow for causal conclusion.
Keywords: Management Research, Organizational Dynamics, Business Management, Organizational Behavior, Leadership Studies, Strategic Management, Business Innovation
How to cite
| BibTeX | Download .bib |
|---|---|
| RIS | Download .ris |
| EndNote | Download .enw |
Licensed under CC BY 4.0.