Pictures that were painted around 700 to 500 years ago were decisive for the art-historically-oriented development of methods. It is based on an idea of imagery that was developed from one single picture / one subject, the European panel painting (Panofsky, 1975, p. 38; cf. critique Pilarczyk & Mietzner, 2005; Müller, 2012). The frescoes by Giotto (see fig. 1) and the works of Michelangelo and Raphael, which inspired Imdahl (1980) and Panofsky (1975) to develop their methods, are originally part of an architectural and painterly Gesamtkunstwerk and are narratively and formally related to the pictures arranged alongside, above, and below them. In the history of its reception and analysis, however, the individual image has prevailed (e.g. The Creation of Adam in the Sistine Chapel), which at most includes the composition with other images in the analysis retrospectively. At first glance, the frescoes are maximally contrasted with those from an influencer’s Instagram profile (Fig. 2: Partial view of Kylie Jenner’s Instagram profile).

Although at second glance the visual language of the Renaissance is reproduced here, the key moment for our argument is not the individual image, but the media framing that Instagram offers in this case. The images published on social media platforms always appear together with other images, in the aesthetic and algorithmic ordering logic of the respective platform.

Challenges of computable imagery

If images that appear and are shared in the context of social media are the focus of research interest, the algorithmically arranged, framed, and personalized (i.e. user-specific) synopsis of the images by the platform requires further development of the image science analysis categories. The following formal aspects are crucial in this regard:

At the interface level, the images are formally framed, i.e. the platform not only specifies the size ratio, but also separates the images from each other with windows or grids. In the currently dominant platform interfaces, feeds dominate the display, i.e. vertically ordered grids that are optimized for the practice of scrolling with smartphones held upright (vertically). This gesturally oriented optimization is accompanied by the fact that not the single image, but the arrangement of several images next to each other or below each other is the most common view.

Each image is also framed by attention-economic information about the image, which provides information about how often the image has been seen, liked, and reposted. Images that have gained a lot of attention so far are algorithmically preferred, i.e. displayed at the top of followers’ feeds. This means that at the level of the algorithmically personalized feed, the images of the followed social media users appear in a predetermined order. The individual image therefore always appears to be (pre)set in relation to other images.

With these formal-informational framework provisions of the platforms in mind, three challenges for the interpretation of computable imagery can be determined in relation to the classic analysis grid:

First, the analysis of planimetric composition in the context of social media is challenged by determining what the focused image is in the first place. Is the platform-specific framework or interface of the app included in the image analysis? Is the attention-economic information (e.g. number of likes) also interpreted? Are the pictures shown together framed as a collaged overall composition? The character of imagery is further complicated by the fact that text-image compositions are often shared (such as memes) and pure text panels and videos are part of the respective image tableau (Przyborski, 2018; Langford, 2021).

Secondly, the perspective identification loses its meaningfulness or must be contextualized when taking into account the display options and additional information that are placed over the image like an additional layer. This is because in the collaged overall composition, the perspective of the image producers, who position the image recipients in a certain way in relation to what is shown via the visually conveyed horizon line, can no longer be clearly determined. There is no longer a single point of view, no vanishing point, no central perspective; accordingly, the platform images appear flat in a two-dimensional sense. Nevertheless, it is possible to analyze a frequently adopted perspective, a perspective typical of an account (for example, in the case of “Neo” in Schreiber (2023), the close-up shot, the fragmented body).