For modern visual analysis, images are no longer pure images, they have become agents. Their actions and mechanisms no longer represent something, as traditionally studied in the field of semiotics, history or art history, for instance, for images actually do something, they produce various types of relationships and thus meanings inside and outside the merely visual space. More concretely, they morph into “boundary images”, a concept that pays tribute to the work on WJT Mitchell and his interest in what pictures “want” but also a concept that can be referred to the broader definition of Susan Leigh Star and James R. Griesemer, for whom a boundary object is “an entity that links networks, elastic enough to be adapted to a new context and robust enough to keep its main characteristics” (p. 2). Boundary images are a subcategory of such boundary objects, and they are themselves capable of accepting new subcategories, like for instance “ontological boundary images”, which transgress the usual frontiers between the ontological domains of reality and fiction.

As the four authors of this book state in their joint theoretical introduction, reflecting on the meaning of the notion of ‘boundary’, such an object is “the exact opposite of ‘an end’ or marker of strict impermeability. Rather, a boundary object is a thing, entity, or any other type of object that can be shared or used differently by varying groups, with each holding its own interpretation, understanding or normative practices with that object” (p. 7). The concept of boundary is thus closer to the idea of contact than of frontier, although the examples of this book will show that there are also strong limitations, at least in practice.

Boundary images are a cutting-edge dimension of contemporary visual culture analysis, not only in that they help nuance and deconstruct conventional dichotomies (literal versus symbolic, rational versus affective, past versus future, single versus plural, subjective versus objective–the list is almost endless), but also in that they foster a reading of images that tackles the dynamic and productive way of their social life. In that sense, the analysis of boundary is political, or unavoidably becomes so. Boundary images aim at challenging the status quo, while also showing that their effect and impact is far from being as progressive and empowering as they may or should be.

The three essays that follow, written by closely collaborating researchers with very different backgrounds (visual arts, curation, urban studies, anthropology, yet always with a very strong interest in digital culture), live up with the high expectations raised by the general introduction of the book. They all do it convincingly, while at the same time–perhaps unwillingly?–highlighting a fundamental problem that I will return to in a concluding remark.

The first essay by Winnie Soon and Magdalena Tyżlik-Carver is both a case study and a far-reaching exploration of how images are either made visible or withdrawn from the visible public domain (in this case the internet). Taking as its starting point a cartographic metaphor (meaning that the image is not seen as the direct visual reproduction of a visual object out there but as the machine translation of a set of encoded data), the essay documents the ways in which a certain image, “Lego Tank Man” (a Lego reconstruction of the famous solitary 1989 Tiananmen Square demonstrator, a world-famous yet in China heavily censored icon of political resistance), appears on screen. During almost one year, artist Winnie Soon daily submitted the same Google search query, taking a screenshot of the image’s presence (or not) on screen and whitening all other results of the query appearing on the same screen. The shifting results of this operation show the underlying presence of mechanisms that disclose direct as well as indirect censorship. In the case of Chinese censorship, the mechanism is blatant (thanks to the Lego mask, the image could however penetrate the digital sphere in China for a couple of hours, before being blocked). In the case of the logarithms outside China that influence and determine how encoded data are translated into images on a screen, the mechanism is less visible but no less effective. Thus the essay draws attention to the manifold biases and manipulations that intervene at the level of data encoding and on the visibility or invisibility of images, the key problem being the fact that all the verbal and visual criteria that are used to describe images before they become visible on screen involve a kind of statistical average that is far from being a natural given, although the AI mechanisms and procedures in this domain are among the best kept secrets of the industry (the notion of ‘face recognition’ for instance is based on models that rely on Western ideas of “normal” faces). The curation of “Lego Tank Man” expands on this message, emphasizing the role of machine and technology interfaces in the process of making visible or invisible. Both the art work as such as its curated version are very efficient in showing the concrete results of these mechanisms. They destroy the illusion that what we see, even if it is what we get, is not what there is to be seen in the real world.

The second and third essay equally expand on the basic insights of the joint introductory essay with its strong focus on ideology critique of technology, but they move from the artistic environment to direct political analysis and action. Melody Devries offers an anthropological field study of the use of ontological boundary images in US and Canadian far-right propaganda, often with a strong evangelical twist, and its attempt to produce belief in non-rational ideas and behaviors. The essay also makes more general claims on belief systems, which are not defined as one-time decisions to accept this or that set of values but as the result of a repeated, if not permanent exposure to certain objects, images, messages, and behaviors (in this sense, screen addiction is directly shown as leading to some kinds of complot theories, apocalyptic thinking and more generally apparently “crazy” belief systems). The power of this exposure is taken very seriously by the researcher, who observes and acknowledges the limits of her own rational thinking in certain circumstances. The final essay by Giselle Beiguelman, artist and professor of architecture, urbanism and design in São Paolo, establishes a thought-provoking analysis of the standardizing aspects of AI and related technologies, which are all reframed in light of the history of eugenics. At the end, but very briefly and schematically, the essay also makes room for non-Western approaches of visual analysis, image collection as well as image presentation and use.

This last point hints at what I see as the major question this book not really addresses. On the one hand, boundary image theory and analysis are introduced in the opening essay as a world of new opportunities (agency, empowerment, political action, inclusion, democracy, etc.). However, this stimulating horizon is systematically contradicted by each of the three case studies, which all heavily insist on the dangers and evils of technology and AI (the notion of boundary thus moves back from contact to frontier). Granted, art and critical theory are shown here to possess some degree of counterhegemonic power, but this power seems fragile–and perhaps even somewhat elitist?–in comparison with the bulldozer effects of modern technological mainstream culture, both Western and non-Western (China, Russia, Iran and the like perfectly know as well how to use AI for their own needs). At the end of this exciting publication, the gap between the optimism of the introduction and the pessimism of the case studies leaves a strange taste in the mind.