Algorithmic Accountability in High-Stakes Automated Decisions: Explainability, Auditability, And Human Oversight
Computer and Decision Making: An International Journal · 02.11.2026 · 3 (6S) · 3139-3155
DOI: 10.1922/comdem.v3i6s.18936 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 02.11.2026. Verfasst von Fadi Almasalha und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
This study develops and tests an integrated framework for perceived algorithmic accountability in high-stakes automated decision-making by examining explainability, auditability, and human oversight as three complementary governance mechanisms. A quantitative cross-sectional survey model was specified in which explainability, auditability, and human oversight were modelled simultaneously as exogenous constructs of perceived algorithmic accountability. The measurement framework comprised 23 reflective indicators evaluated using partial least squares structural equation modelling (PLS-SEM). The study analysed 300 valid survey responses from AI, risk, audit, and domain professionals collected through an online survey between August and September 2026. Recruitment used virtual channels including Google Forms, direct e-mail invitations, LinkedIn and other professional networks, and institutional/professional mailing groups. The analysis included descriptive statistics, correlations, measurement-model assessment, discriminant-validity tests, collinearity diagnostics, bootstrapping, effect-size analysis, explained variance, and Q² predictive relevance. All three mechanisms showed positive and statistically significant associations with perceived algorithmic accountability. Auditability showed the strongest association (β =.417), followed by human oversight (β =.242) and explainability (β =.241). The model explained 48.7% of the variance in perceived algorithmic accountability and demonstrated satisfactory reliability, convergent validity, discriminant validity, and predictive relevance. Organizations deploying high-stakes AI should combine understandable explanations with traceable evidence, systematic audit records, and empowered human reviewers capable of challenging, overriding, and correcting automated outcomes. The findings highlight accountability as a multidimensional socio-technical capability rather than a consequence of transparency alone. The study advances algorithmic-accountability research by jointly estimating, within one PLS-SEM model, the distinct associations of explanation quality, audit evidence infrastructure, and human review authority with perceived algorithmic accountability. This moves beyond treating transparency, auditing, and oversight as separate governance checklists and provides a comparative empirical account of their relative associations with answerability, contestability, responsibility attribution, and corrective capacity.
Zitieren
APA
Almasalha, F., & Santos, P. J. T. D. (2026). Algorithmic Accountability in High-Stakes Automated Decisions: Explainability, Auditability, And Human Oversight. Computer and Decision Making: An International Journal, 3(6S), 3139-3155. https://doi.org/10.1922/comdem.v3i6s.18936
Vancouver
Almasalha F, Santos PJTD. Algorithmic Accountability in High-Stakes Automated Decisions: Explainability, Auditability, And Human Oversight. Computer and Decision Making: An International Journal. 2026;3(6S):3139-3155. doi:10.1922/comdem.v3i6s.18936
BibTeX
@article{Almasalha2026,
author = {Almasalha, Fadi and Santos, Paulo Junior Trindade dos},
title = {Algorithmic Accountability in High-Stakes Automated Decisions: Explainability, Auditability, And Human Oversight},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {3},
number = {6S},
pages = {3139--3155},
doi = {10.1922/comdem.v3i6s.18936}
}
EUVIMED