Haunting and Hauntology in AI Collaborative Image-Making

This paper documents the Author’s experiments with 3D-based AI image generation software, identifying within the visual outcomes a tendency towards hauntology that is seemingly the result of the collaborative image-making process as well as a material quality of the method. The paper suggests that a combination of the authors own aesthetic concerns and working methods, combined with the tendencies inherent within the training method used to develop the Stable Diffusion AI model, results in images that are haunting in a number of ways that align to the conceptual framework of Hauntology: through unexpected traces and glitches; anachronism; notions of shared dreaming / remembering; and through the invocation of the poor image.