Paper Type
Short
Paper Number
PACIS2026-1735
Description
A novel form of unethical consumer behavior has emerged in e-commerce after-sales contexts, where consumers use generative AI to fabricate product damage evidence for fraudulent refund claims. This study investigates how perceived AI agency, defined as consumers’ subjective attribution that AI acts as an autonomous proxy executor rather than a passive tool, may facilitate this behavior. Drawing on Moral Disengagement Theory (MDT), we propose that perceived AI agency activates displacement of responsibility at the agency locus of MDT, triggering anticipatory moral disengagement, and thereby increasing unethical behavioral intention. Victim identifiability and warning message type are introduced as boundary conditions to moderate this process. Four scenario-based experiments are designed to test these propositions. This study contributes a three-dimensional conceptualization of perceived AI agency (goal interpretation, means selection, and outcome execution), advances understanding of how AI proxy execution reshapes consumer moral cognition, and offers implications for platform governance and AI ethics design.
Recommended Citation
wang, chen; Zhang, Jiacheng; Liu, Xincan; and Wang, Xi, "From Tools to Proxy Executors: The Role of Perceived AI Agency in Unethical Consumer Behavior" (2026). PACIS 2026 Proceedings. 12.
https://aisel.aisnet.org/pacis2026/ai_ethic/ai_ethic/12
From Tools to Proxy Executors: The Role of Perceived AI Agency in Unethical Consumer Behavior
A novel form of unethical consumer behavior has emerged in e-commerce after-sales contexts, where consumers use generative AI to fabricate product damage evidence for fraudulent refund claims. This study investigates how perceived AI agency, defined as consumers’ subjective attribution that AI acts as an autonomous proxy executor rather than a passive tool, may facilitate this behavior. Drawing on Moral Disengagement Theory (MDT), we propose that perceived AI agency activates displacement of responsibility at the agency locus of MDT, triggering anticipatory moral disengagement, and thereby increasing unethical behavioral intention. Victim identifiability and warning message type are introduced as boundary conditions to moderate this process. Four scenario-based experiments are designed to test these propositions. This study contributes a three-dimensional conceptualization of perceived AI agency (goal interpretation, means selection, and outcome execution), advances understanding of how AI proxy execution reshapes consumer moral cognition, and offers implications for platform governance and AI ethics design.
Comments
03-EthicsSocietalImpact