The psychological mechanisms by which algorithmic authority influences moral disengagement in Human–AI collaborative decision-making

Abstract: As artificial intelligence (AI) is increasingly used in high-stakes domains, including healthcare, judicial decision-making, and organizational management, human decision-makers may deflect moral responsibility by attributing their actions to algorithmic recommendations. Algorithmic authority has thus emerged as a key factor influencing moral disengagement. However, empirical findings regarding the direction of this effect are mixed. In some contexts, algorithmic authority exacerbates moral disengagement; in others, it inhibits moral disengagement by heightening moral vigilance. To account for these divergent findings, this paper proposes a Dual Cognitive Appraisal Model of Algorithmic Authority and Moral Disengagement. The model posits that algorithmic authority does not directly determine moral disengagement. Instead, it reshapes individuals’ self-responsibility attribution through two forms of cognitive appraisal: technical cognitive appraisal and moral cognitive appraisal. When moral conflict and explainability are both low, individuals are more likely to defer their judgment to algorithms. This weakens self-responsibility attribution and induces moral disengagement through the displacement of responsibility. Conversely, when moral conflict and explainability are both high, individuals are more likely to scrutinize the basis for algorithmic recommendations and recognize the moral significance of decision outcomes. This strengthens self-responsibility attribution and thereby inhibits moral disengagement. By clarifying the psychological mechanisms and boundary conditions through which algorithmic authority influences moral disengagement, the model provides new theoretical insights and practical implications for the design of responsible human–AI collaborative decision-making systems.