Exploring the Behavioral Differentiation and Psychological Impact of Different Attachment Types of Users Interacting with AI

In the rapid development of generative AI, affective AIs have gradually entered public life as a new type of social companion, and users show different degrees of emotional attachment and healing experiences in their interactions with AIs. Attachment theory suggests that the attachment patterns formed by individuals in early relationships will be extended to their subsequent intimate interactions, which may also affect their interaction and psychological responses with AIs. In this study, we take attachment type as the independent variable and introduce human-computer relationship strength as the mediator variable to explore the psychological outcomes of emotional healing and emotional addiction in AI interactions by users with different attachment types from a dual-path perspective. The influence mechanism of different attachment types of users on emotional healing and emotional addiction in AI interaction was explored from a dual-path perspective. A total of 203 users were surveyed, and the ECR-S, PSI, PANAS and adapted PUCAI scales were used to measure the variables. The results found that: anxious and fearful attachment individuals had higher HCI intensity; HCI fully mediated the effect of anxious attachment on emotional addiction; and HCI partially mediated the positive effect of secure attachment on positive emotions but did not significantly affect negative emotions. This study reveals the predictive role of different attachment styles on the psychological outcomes of AI use, emphasizing that individual user differences should be fully considered in AI design and ethical governance to prevent the risk of abuse and leverage the positive value of technology.