Drawing a satisfying picture: An exploratory study of human-AI interaction in AI Painting through breakdown–repair communication strategies

Artificial intelligence (AI) painting represents a burgeoning field in AI-generated content; however, there is limited research on users’ preferences and biases concerning AI painting compared to traditional forms of art. This study seeks to shed light on the communication and interaction dynamics between painters and AI tools in the process of creating a satisfactory image. A self-report study involving 52 volunteers and 312 reports was conducted, analyzing users’ communication strategies amid AI painting-related communication breakdowns. The investigation aimed to unveil users’ experiences, behavioral patterns, psychological motivations, and needs. The study identified three primary communication barrier categories in AI painting and proposed 10 corresponding repair strategies. It established a user experience framework within the evolving dynamics of user interaction, encompassing behavioral patterns and psychological motivations. Users’ perceptions of the AI painting system significantly impacted the interaction process, leading to an anthropomorphic view of the system as a “potential assistant” and the continuous use of effective strategies to “train” the AI painting system. Conversely, perceiving the system as a “flawed collaborator” prompted the adaptive expectation strategy. Persistent repair failures may lead to users attributing responsibility to the limitation of AI system, culminating in interaction termination. The study also explores the influence of user interaction experiences, including user background and interaction efficiency. By modeling users’ behavior patterns and underlying psychological cognitions, this research enhances the understanding of user psychology and preferences, an understanding that can be used to improve the efficiency of human–computer communication.