Introduction
This paper presents a creative collaboration between two design researchers using text-to-image prompts to think across two key ideas. First, we explore the relationships between collage practices and prompt-based image-generation – both modes of image-making that create images with images. Second, we take an ‘anarchival’ approach to thinking about absence in historical archives through creative practice. We define ‘anarchival practice’ as an approach which aims to trouble what is missing or omitted from historical archives, how these archives are organised, and how archival materials might be reactivated in critical and creative ways.
In the section ‘Creating Images with Images’, we discuss a set of experiments which aimed to generate images of the extinct King Island dwarf emu, specifically, an emu taken to live in Empress Josephine’s Malmaison estate outside Paris. There is little visual record of dwarf emus, and what remains is ambiguous and factually inaccurate. An initial experiment involved creating paper collages from archival photographs and illustrations, followed by an experiment in writing prompts with DALL•E 2 (Ramesh et al. 2022) to generate images of the emus in Paris. We reflect that the scarcity of visual reference material on dwarf emus provides an interesting case study to explore how generative image models might attempt to visualise images of a historical event and an extinct species.
In the section ‘Working with Absence in Historical Archives’, we consider prompting with large-scale image-generation models as a way to think-with and speculate, rather than to merely generate. Through our work of thinking-with generative models, we asked: What happens when a generative model is used in an anarchival practice? How is image generation similar and different to collage within an anarchival practice, particularly one that focuses on absence in historical archives?
We frame this process of ‘thinking-with’ after Donna J. Haraway (2016), for whom thinking and knowing are relational processes. Haraway’s idea of thinking-with acknowledges that we are always thinking and knowing collectively with an array of humans/non-humans or in the mix of more-than-human entanglements. Maria Puig de la Bellacasa argues that the relational interdependency of thinking is ‘not a contract, nor a moral ideal – it is a condition’ (2017, 70). In other words, a condition of thinking is relation, and further relations condition what thinking is, how it works, in different ways. To use text-to-image models to think-with is to acknowledge the expansive relationality of such image-making processes across human and non-human, as well as the specific, contingent and situated ways of thinking-with them.1 This means understanding text-to-image models beyond an ability to fill a ‘gap’ in terms of what is missing or to offer a ‘coherent’ visualisation of the dwarf emu’s experience in France. Rather, text-to-image models generate images according to both their dataset and how models, in their work with images, enact specific ways of non-visually re-organising the visual (Munster and Mackenzie 2019). As a form of thinking-with, prompting in our project becomes a way to think-with text-to-image models about extinction, absence in the archive, and the models themselves as image-making processes.
The section ‘Further Experiments to Think-With’ discusses a set of experiments designed to critique the images from the first experiment. Without access to DALL•E’s training dataset, we first tested what a reverse image search might reveal. This method of visual content analysis uses ‘features by proxy’ (after Offert’s 2023 ‘attribution by proxy’) to understand what qualities, or features, of images in the training dataset are repeated in the generated images. Second, we conducted a comparative analysis of images created by other models, given the same prompts, to interrogate how different models visualise the emu.
In the final section, we reflect on how this creative practice has led to new insights about how we might think-with models, how we might think about absence in the archive, and the limitations imposed by the prompts and models used in the experiments. We differentiate between creating and critiquing images in a formal way – considering how they are composed from line, shape, pattern, colour, etc – and creating and critiquing them from a cultural and ethical perspective – considering what they are doing in the world. We conclude by arguing that, at a time of both deliberate and accidental miscommunication, it is important for those with expertise in how images ‘work’ to critique and analyse these image-generating tools.
The extinction of dwarf emus
Before reporting on our suite of image-making experiments, we first need to introduce the strange and tragic story of the extinct dwarf emus: the King Island emu (Dromaius novaehollandiae minor) and Kangaroo Island emu (Dromaius novaehollandiae baudinianus). Although our experiments focus on the King Island emu, the confusion in historical records makes it relevant to introduce both subspecies.
In 1804, as Napoleon Bonaparte was crowned emperor of France, his wife Josephine acquired several dwarf emus for the grounds of her Malmaison estate, which was a living ‘cabinet of curiosities’ from French colonial expeditions.2 The emus were collected from what are now known as Kangaroo Island in the Great Australian Bight and King Island in the Bass Strait, during a scientific expedition to Australia led by Nicolas Baudin.3 Not long after the dwarf emus arrived in their artificial mise en scéne in Josephine’s estate, their remaining kin were hunted to extinction by colonial settlers, mostly for food but also for sport.
The dwarf emus in the French menagerie were later moved to the Jardin des Plantes in Paris, where they died in April and May 1822. They are thought to be the endlings of their kind.4 There is little visual documentation in historical archives of these two emu subspecies, in part due to how quickly after colonial ‘discovery’ they were driven to extinction. The specimens and documentation that exist are fragmented and misleading.
The remains of one emu are dislocated; the skeleton is held in the Musée Nationale d’Histoire Naturelle in Paris and its skin (a taxidermy mount) in the Natural History Museum of Geneva. Other remains are scattered around European museums, including a worn skeleton in the Museum of Natural History, Florence, which is a composite of two birds. The South Australian Museum has a model constructed in the 1940s based on remains, but the feathers are from mainland emus. A single King Island emu feather was gifted to the Tasmanian Museum and Art Gallery in 2000, from the Paris museum, and is the only confirmed dwarf emu feather in Australia (Pfennigwerth 2013).
King Island emu (Dromaius novaehollandiae minor) in the collections of the French National Museum of Natural History. The chick is likely a mainland emu.
Credit: Marie-Lan Nguyen, Wikimedia Commons.
Of particular interest is the natural history illustration of 'Kangaroo Island emus', by Charles-Alexandre Lesueur, an artist on Baudin’s expedition (above). It is worth noting that this image is usually credited to Lesueur but is in fact a digital reproduction of an engraving by F. Lambert after the paintings and drawings of Lesueur. Remediation and multiple authorship within historical archives are easily overlooked when conducting surface-level searches.
The title ‘Ile Decrès’ – the French name for Kangaroo Island – implies that this is a Kangaroo Island emu family. The illustrated tableau was intended to depict a father standing protectively over a nesting mother and playful young. Yet historians now identify the standing figure as a male Kangaroo Island emu, the seated one as a male King Island emu, and the chicks as mainland emus (Pfennigwerth 2013). Three subspecies forced into an unlikely family unit, preserving an inaccurate memory of the lifeways of these extinct species.5 What makes the tableau more unlikely, and would have shocked French society at the time, is the way emus actually breed and brood. A female emu lays a clutch of large eggs, then leaves to seek a new mate while the male incubates the eggs and raises the young. Historian Stephanie Pfennigwerth described Lesueur’s depiction as ‘a political illustration, not an ornithological one’:
The reason they were depicted thus was a form of (French) Revolutionary philosophical expression: the desired natural moral order of things. Showing the birds in this tableau also fulfilled the expedition’s more mercenary aims: to find animals ‘useful’ for agricultural/economic applications.
[personal correspondence, email, 6 Sept. 2018]
Using Generative AI to Visualise
an Extinct Dwarf Emu
Dr Zoë Sadokierski and Dr Monica Monin
School of Design, University of Technology Sydney
Only known skin of a Kangaroo Island emu, held at the Natural History Museum of Geneva.
Credit: Totodu74, via Wikimedia Commons, 2018
Archival materials of the Kangaroo Island emu
Dromaius novaehollandiae baudinianus
Scroll right to view
Charles-Alexandre Lesueur’s 1807 plate of the head, wing and feathers of a possible King Island emu.
Credit: Lesueur, Charles-Alexandre, 1778–1846. Originally published in Voyage de decouvertes aux terres australes. Public domain, via Wikimedia Commons.
Only known Kangaroo Island skeleton, held at the Jardin des Plantes, Paris.
Credit: Nicolas Venner, 2019
Illustration by Charles-Alexandre Lesueur, based on life drawings made during Nicolas Baudin's journey and specimens kept at the Jardin des Plantes. Circa 1808 to 1811. The image above is a hand-coloured copper engraving made by F. Lambert after paintings and drawings by Lesueur. Lesueur was the official artist of natural history on Nicolas Baudin's voyage to Australia from 1800 to 1803.
Credit: F. Lambert, c. 1824
5. In anthropology and the environmental humanities, ‘lifeway’ refers to the way organisms are entangled within ‘interwoven patterns of living and dying, being and becoming, in a larger world’ (van Dooren et al. 2016) as part of a relational way of understanding nature.
Paper collage (left) and AI- generated image (right) of Empress Josephine with dwarf emu, created by Zoë Sadokierski and DALL•E 2, 2023–24.
2. Cabinets of curiosities (also Wunderkammer) were a troubling part of the colonial project: wealthy Europeans assembled personal collections of cultural materials from colonised places, decontextualising them from their place and culture of origin.
3. Karta Pintingga is the traditional name of Kangaroo Island for the Kaurna people, and Erobin or Yerubin is the traditional name for King Island.
4. Endling is a term that describes the last known individual of a species.
1. For a comparative example, see Kate Crawford and Vladan Joler’s ‘Anatomy of an AI System’ (2018), a visualisation that maps an expansive and hidden array of human labour, extracted planetary resources, data and processes behind a simple voice-command interaction with an Amazon Echo.
Illustration of Dromaius peroni (now a synonym of Dromaius baudinianus – Kangaroo Island emu), but it was actually based on a specimen in the Museum of Natural History in Paris, now thought to belong to Dromaius ater from King Island, also known as Dromaius novaehollandiae minor.
Confusions upon confusions.
Credit: John Gerrard Keulemans, 1907.
Digital collage altering a natural-history etching of dwarf emus with graphic and textual elements. Zoë Sadokierski 2023.
Click the image to enlarge.
Photograph of dwarf emu feather in the Tasmanian Museum and Art Gallery. Supplied by David Hocking, 2024.
Click image to enlarge.
Dwarf emus and other animals outside Château de Malmaison. Frontispiece from the Atlas of the Voyage de découvertes aux terres australes. Charles-Alexandre Lesueur, c. 1807.
Credit: Public domain, via Wikimedia Commons
Mainland and King Island emu skulls. Plates from ‘A collection of sub-fossil bird and marsupial remains from King Island, Bass Strait’.
Credit: Baldwin Spencer and J. A. Kershaw, from Memoirs of the National Museum, Melbourne, 1910.
Mainland and King Island emu pelves. Plates from ‘A collection of sub-fossil bird and marsupial remains from King Island, Bass Strait’.
Credit: Baldwin Spencer and J. A. Kershaw, from Memoirs of the National Museum, Melbourne, 1910.
Mainland and King Island emu femora. Plates from ‘A collection of sub-fossil bird and marsupial remains from King Island, Bass Strait’.
Credit: Baldwin Spencer and J. A. Kershaw, from Memoirs of the National Museum, Melbourne, 1910.
