The Social Construction of Generative AI Prompts

Abstract

As text-to-image AI tools grow in capability and widespread use, research has focused on studying individualistic user prompt crafting strategies. Recognizing that technologies are socially constructed, this paper examines prompt engineering through a social lens. We propose reframing prompt engineering as a socio-cultural practice shaped by collective knowledge building. Through qualitative analysis of 19 semi-structured interviews with members of the MidJourney community, a text-to-image generative AI tool, we identify four socio-engagement themes: proprietary/solitary, derivative, collaborative, and provocative prompting. These themes reveal a space of social engagement modes based on personal values and motivations from individual exploration to influencing the prompt community and highlight a fine line between being inspired by others' prompts and maintaining creative ownership. We argue that understanding distinct social engagement preferences can inform the design of AI tools to facilitate transparent prompt reuse mechanisms, integrate collaborative features, or preserve ethical concerns about prompt sharing.