Co-Rationality Mechanisms in AI Decision Support

AI-enabled decision support systems shape how users interpret evidence, allocate attention, calibrate trust, exercise control, and justify decisions. Although research has examined trust, transparency, explainability, automation bias, workload, fairness, and accountability, these constructs remain dispersed across theoretical traditions. This study develops co-rationality as an integrative mechanism framework for analyzing how human judgment, algorithmic inference, and governance conditions are configured in AI-supported decision systems. We conducted a systematic computational review combining database retrieval, deduplication, relevance screening, phrase extraction, clustering, GPT-4o-assisted annotation, source-text auditing, and human validation. The final corpus comprised 1,086 abstracts and 6,434 coded mechanism instances, consolidated into 53 mechanisms across cognitive, emotional, and social domains and individual, interaction, and system levels. The study contributes a mechanism-level framework, clarifies how co-rationality extends distributed cognition, hybrid intelligence, sociotechnical systems, calibrated trust, and conjoint cognitive systems, and proposes indicators including override rates, explanation requests, trust repair, auditability, and contestability.