With the rapid advancement of artificial intelligence, businesses increasingly employ AI agents to facilitate consumer consent for data collection in various domains. For instance, in financial services, AI agents assist in collecting detailed financial information to provide tailored investment advice, while in healthcare, they collect sensitive medical data for personalized treatment plans. As the importance of user data collection grows and the use of AI agents becomes more widespread, the current research investigates a critical question: How do privacy authorization agents (AI vs human) affect consumer privacy disclosure?
The research hypotheses were tested through five online experiments conducted with participants recruited via platforms such as Credamo and Connect. A series of experimental scenarios was designed to collect the necessary data. Descriptive statistics, main effect analyses, mediation analyses, and moderation analyses were performed using statistical software to evaluate the proposed hypotheses.
Results show that individuals were more likely to disclose personal information to AI agents than human agents under the high information sensitivity condition. However, no significant differences in privacy disclosure were observed between the two agent types under the low information sensitivity condition. This effect is driven by AI agents eliciting lower privacy concerns than human agents in the high information sensitivity condition. In contrast, in the low information sensitivity condition, privacy concerns remain dormant and do not affect privacy disclosure behavior. Additionally, in the high information sensitivity condition, privacy preferences moderate the effect of privacy authorization agents on privacy disclosure. Specifically, consumers who prioritize privacy protection are more inclined to disclose privacy to AI agents, whereas those who prioritize privacy usage are more inclined to disclose privacy to human agents.
This research offers key theoretical contributions to the fields of privacy disclosure, algorithmic decision-making, and the boundary conditions distinguishing algorithm aversion from appreciation. By introducing privacy authorization agents, the current study expands research on privacy disclosure. Furthermore, the current study enriches the literature on algorithmic decision-making and algorithm appreciation by examining the conditions under which individuals prefer AI over humans in privacy contexts. Finally, the findings further advance research on the boundary conditions of algorithmic decision-making by identifying privacy preference as a key moderator in high information sensitivity contexts. The findings also offer valuable insights for organizations engaged in data collection and AI implementation, including the importance of recognizing the positive role of AI agents and designing effective privacy protection mechanisms. And the research also provides guidance for developing AI privacy ethics at the societal level.
