<p>Organizational decision-making has been fundamentally reshaped by advances in artificial intelligence and data analytics. Hybrid configurations, in which human judgment operates alongside algorithmic computation, are increasingly recognized in contemporary scholarship as offering decision-quality advantages that neither purely human nor purely automated systems can achieve alone. Yet the rationality that governs such configurations remains bounded, and the evaluative frameworks available to scholars and practitioners remain inadequate for hybrid arrangements in which emergent properties arise from human–machine interaction. These arrangements also raise specific ethical stakes, concerning fairness, accountability, and the distribution of moral agency, that conventional evaluation frameworks are ill-equipped to address. This paper reports an integrative literature review designed to synthesize fragmented disciplinary perspectives on hybrid decision-making and to derive a theoretically grounded evaluation framework from that synthesis. The review drew on five databases: Scopus, Web of Science, Google Scholar, IEEE Xplore, and ACM Digital Library, and applied a structured selection protocol. An initial search yielded 847 publications; after deduplication, two-researcher title-and-abstract screening, independent full-text review, and quality assessment, a final corpus of 89 peer-reviewed publications, supplemented by 23 pre-2010 foundational sources identified through citation snowballing, for a combined working corpus of 112 items. The formal analytic corpus comprises the 89 peer-reviewed publications; the 23 supplementary sources are incorporated as conceptual anchors but are clearly distinguished throughout the articles was retained. These sources were analyzed across three theoretical streams: algorithmic rationality, behavioral and cognitive dynamics, and organizational governance. Systematic synthesis of these streams identifies significant gaps in how hybrid decision quality is conceptualized and measured. In response, this review proposes the Triadic Evaluation Model, which suggests that hybrid decision quality should be assessed across three interdependent dimensions: Technical Auditability (encompassing algorithmic transparency and performance validity), Cognitive Alignment (addressing mental-model accuracy, trust calibration, and human-AI complementarity), and Contextual Viability (covering organizational fit, accountability structures, and ethical alignment). The model challenges prevailing substitution-oriented paradigms by positioning value as emerging from synergistic interaction between human and algorithmic agents rather than from automation alone. Ethical dimensions, including fairness, accountability, and moral agency, are treated as constitutive of, rather than supplementary to, this evaluative framework. This conceptual review thereby contributes theoretically coherent vocabulary for cross-disciplinary discourse on hybrid intelligence evaluation and identifies priority directions for future empirical research.</p>

Human-AI hybrid rationality in decision-making: an integrative literature review

期刊论文
发表日期：

2026年6月26日

查看全文

Asmae Sbai,
Abdelmoumen Berjaoui

摘要

Organizational decision-making has been fundamentally reshaped by advances in artificial intelligence and data analytics. Hybrid configurations, in which human judgment operates alongside algorithmic computation, are increasingly recognized in contemporary scholarship as offering decision-quality advantages that neither purely human nor purely automated systems can achieve alone. Yet the rationality that governs such configurations remains bounded, and the evaluative frameworks available to scholars and practitioners remain inadequate for hybrid arrangements in which emergent properties arise from human–machine interaction. These arrangements also raise specific ethical stakes, concerning fairness, accountability, and the distribution of moral agency, that conventional evaluation frameworks are ill-equipped to address. This paper reports an integrative literature review designed to synthesize fragmented disciplinary perspectives on hybrid decision-making and to derive a theoretically grounded evaluation framework from that synthesis. The review drew on five databases: Scopus, Web of Science, Google Scholar, IEEE Xplore, and ACM Digital Library, and applied a structured selection protocol. An initial search yielded 847 publications; after deduplication, two-researcher title-and-abstract screening, independent full-text review, and quality assessment, a final corpus of 89 peer-reviewed publications, supplemented by 23 pre-2010 foundational sources identified through citation snowballing, for a combined working corpus of 112 items. The formal analytic corpus comprises the 89 peer-reviewed publications; the 23 supplementary sources are incorporated as conceptual anchors but are clearly distinguished throughout the articles was retained. These sources were analyzed across three theoretical streams: algorithmic rationality, behavioral and cognitive dynamics, and organizational governance. Systematic synthesis of these streams identifies significant gaps in how hybrid decision quality is conceptualized and measured. In response, this review proposes the Triadic Evaluation Model, which suggests that hybrid decision quality should be assessed across three interdependent dimensions: Technical Auditability (encompassing algorithmic transparency and performance validity), Cognitive Alignment (addressing mental-model accuracy, trust calibration, and human-AI complementarity), and Contextual Viability (covering organizational fit, accountability structures, and ethical alignment). The model challenges prevailing substitution-oriented paradigms by positioning value as emerging from synergistic interaction between human and algorithmic agents rather than from automation alone. Ethical dimensions, including fairness, accountability, and moral agency, are treated as constitutive of, rather than supplementary to, this evaluative framework. This conceptual review thereby contributes theoretically coherent vocabulary for cross-disciplinary discourse on hybrid intelligence evaluation and identifies priority directions for future empirical research.