Appropriate Reliance on AI Decision Support Systems: A Sociotechnical Narrative Review

Artificial intelligence increasingly supports human decision making, but effective use depends on whether users assign AI advice the decision weight it warrants. This theory-oriented narrative review synthesizes 29 core studies, together with recent reviews and foundational theory, to examine appropriate reliance, overreliance, and underreliance in AI-supported decisions. The synthesis identifies five recurring sociotechnical conditions shaping reliance: system, user, task, interaction, and organization. Findings show that explanations, uncertainty cues, expertise, self-confidence, prior experience, and organizational incentives affect reliance differently across settings, while trust and reliance evolve through feedback over time. Interventions that reduce overreliance may also increase underreliance. The resulting framework centers warranted behavioral decision weight, comparative evaluation of AI and self-capability, and temporal feedback, shifting attention from increasing trust or AI use toward supporting context-sensitive allocation of decision weight.