I-Figure in a Concrete Scene: A Pattern of Addressing an Audience and Its Realisation in the Visual Arts

Images that are generated by machine learning tools are synthesised from the image archives that the programs can access. The images created are therefore close to the ‘previous’ ones. They no longer refer to factuality or truth, but to probability. In this paper, I am interested in a certain pattern of addressing the audience which consists of using image personnel who appear in the form of the ‘all-seeing’ frontal figure or the frontal portrait, as a back figure or as a small group portrait of two to three people, in order to attract the viewer’s gaze and direct it to the background of the image (or to an accompanying text). In most cases, we are dealing with one or more figures that are clearly visible in the foreground, leading the eye into the centre and background of the picture, in which a certain ‘scene’ is concretised. Millions of such images find their way into media channels such as magazines, blogs and Instagram via stock images.

These patterns become very clear when diffusion models of AI programs that generate images, such as dalle-e, midjourney or stable diffusion, are used. This is because the generation of images and the perception of images, i.e. the activity of creating art and viewing it, are intertwined in these programs. These AI-generated image media therefore also reveal the structures and patterns of our visual culture and its prevailing traditions, including stereotypes, prejudices and biases. This paper asks the following questions:

- What historical image traditions can be identified in relation to these patterns?

- How do they manifest themselves in the present and in particular in connection with AI-generated images?

- What makes these image patterns so attractive for contemporary acts of visual communication?

- How do artists adopt these patterns using repetition, exaggeration and alienation?