Author(s) :

Claire Daeun Kim.

Volume/Issue :

Abstract :

As artificial intelligence systems increasingly assist or replace human decision-making in high-stakes domains, transparency has become a central design principle intended to support user understanding and trust. However, the effectiveness of transparency may diminish as users may not process the provided information entirely due to cognitive load. Thus, it is unclear whether detailed explanations of AI systems always lead human users to evaluate the system in a positive light. To answer the question, this study conducted an online experiment where 181 participants were given AI-assisted decision-making scenarios with different levels of decision transparency and cognitive load. Contrary to the expectation that transparency may backfire under high cognitive load, the findings show that transparency about the decision process consistently increased perceived trustworthiness and ethicality of AI systems regardless of cognitive load. Moreover, the experiment results find that the level of explanation detail or cognitive load does not alter users’ responsibility attribution, pointing to the rigidity of moral responsibility in the AI-assisted decision-making contexts. This study contributes to the literature of AI-human interaction by empirically demonstrating the limited role of cognitive load in shaping human perceptions of AI decision-making. In addition, this study hints at important practical implications for how AI-human interaction should be designed to foster the effectiveness of the interaction.