The empirical studies of the arts have been keeping the focus on the appreciation and creation of art even though the concept of art and the circumstance of the artistic experience is constantly changing. One of the new challenges came with the digital era. What kind of changes are detectable in the aesthetic process, if there’s any? Aritificial intelligence (AI) – assisted generative art, such as text-to-image systems, promises to democratize art creation, enabling individuals to visualize their inner worlds in ways previously inaccessible. But does this promise hold? Can AI-assisted generative art truly capture the nuances of our inner worlds and personal identity?
In our study, 230 participants (mean age 21.54 years, SD = 2.12) were asked to create self-representative images using the text-to-image software Midjourney, guided by natural language prompts. Participants were given 45 minutes to create images. The image creation session was followed by a semi-structured interview, where participants reflected on their experiences, the creation strategies they used, and the images they produced. The study employed both quantitative methods (such as sentiment analysis of the image creation prompts and predicting psychological questionnaire results using machine learning algorithms) and qualitative ones (such as thematic and phenomenological analysis of the interviews and the images themselves). Together, these analyses suggest that AI-assisted art holds potential for both self-exploration, and it opens up new perspectives for empirical aesthetics.