Critical Thinking in AI-Assisted Deontologically-Governed Professional Decision-Making: When and How Explainability, Reliability, and Transparency Matter

Korosec-Serfaty, Marion; Léger, Pierre-Majorique; Parent-Rocheleau, Xavier et Sénécal, Sylvain (2026). « Critical Thinking in AI-Assisted Deontologically-Governed Professional Decision-Making: When and How Explainability, Reliability, and Transparency Matter ». Information Systems Frontiers.

Fichier(s) associé(s) à ce document :

Prévisualisation	PDF Télécharger (4MB)

Résumé

Critical thinking is a central safeguard for responsibility and accountability in deontologically-governed professions. Artificial intelligence (AI)-assisted decision-making is increasingly being integrated within these professional workflows. However, AI introduces autonomy, learning, and inscrutability, disrupting critical, reflective, decision-making processes. Given these challenges, this research systematically unpacks how embedding explainability, reliability, and transparency into AI fosters critical thinking. Employing a multi-method approach combining cognitive neuroscience, behavioral and self-report measures, we conducted three experiments with practicing professionals tasked with realistic AI-assisted scenarios. Experiment 1 assessed varying levels of AI-generated reconstructive causal explanations under consistent reliability, Experiment 2 introduced variable reliability. Experiment 3 added transparency through model confidence scores. Results reveal that minimal reconstructive explanations enhance analytical reasoning under reliable conditions, while epistemic uncertainty drives critical engagement when reliability varies. Transparency offers limited restoration of explainability benefits. These findings suggest AI reliability primarily drives critical thinking, informing AI design that preserves professional responsibility and accountability.

Type:	Article de revue scientifique
Mots-clés ou Sujets:	Reconstructive causal explanations, Epistemic uncertainty, Model confidence scores, Critical thinking, Deontologically-governed professions, AI-assisted decision-making, Cognitive Science, NeuroIS, Neuroscience, Philosophy of Artifical Intelligence
Unité d'appartenance:	École des sciences de la gestion > Département d’analytique, opérations et technologies de l’information (AOTI)
Déposé par:	Marion Korosec-Serfaty
Date de dépôt:	05 mars 2026 08:35
Dernière modification:	05 mars 2026 08:35
Adresse URL :	https://archipel.uqam.ca/secure/id/eprint/19747

Modifier les métadonnées (propriétaire du document)

Statistiques