1. Introduction

In 2008, Okwui Enwezor curated an exhibition titled Archive Fever: Uses of the Document in Contemporary Art at the International Center of Photography. This exhibition surveyed thirty years of artists exploring the photographic and filmic archive. Some, like Zoe Leonard, explored biographies of fictitious persons; others, like Walid Raad, reflected upon absurd historiographies of national events; and still others, like Christian Boltanski, would blur the line between fiction and history in photomontages. Through this show, Enwezor sought to chart a constellation of artists who think through photography and film as the quintessential archive, regardless of the source material’s origins (Enwezor 2008).

14 years later, we still see this interest in the archival turn, though it has mutated with the presence of computer-generated (CG) media, algorithms, and artificial intelligence (AI). Heesoo Kwon’s Premolt (2019-present) series encourages viewers to review her lenticular light prints, where 3D-rendered models of her ancestors shimmer holographically through family photographs. On the other hand, Jon Rafman’s six-minute short, You, the World, and I (2011), invites suspicion of its archive, as the narrator adds CG landscapes and panoramas to supplement a single still of an unnamed lover found through Google Street View. Still others, like Rashaad Newsome’s Being: The Digital Griot (2019-present) feature an AI program trained on the works of abolitionist, queer, and feminist texts and vogue performance to develop an anti-tour guide, a therapy phone application, and eventually decolonization workshops. Jacolby Satterwhite, Lu Yang, and Hito Steyerl also draw attention to the digitized, motion-captured body. In their separate ways, they use dancing CG avatars as a medium to represent fictional fabulations to build, expand, and sustain existing archives through alternative modes.

These artworks use generative computational methodologies to revisit personal archives, and in doing so they transform our experience of what we know to be true. The pieces do not simply illustrate unquestionable traces of the past; rather, they call attention to the fact of their own digital generation, and as such, to the instability of the truths they portray. They illustrate various, sometimes variable, iterations of history, allowing each interaction between the artwork and the viewer to change with each viewing. The artworks are self-referential yet at the same time interested in acting as a historical referent. They are interested in the continual creation and preservation of documents in their archives.

For this paper (and eventually, my dissertation), I collate viewer-object encounters with such works into what I call the “generative archive.” Building upon Enwezor’s exhibition, the generative archive examines artworks that use CG and AI to make a claim for historicity in relation to pre-existing collections of objects and data. The archives range from personal souvenirs of public databases; they encompass Google Street View stills, family photographs, literary theory, motion-capture data, and more. I observe three general tendencies within these artworks that contour the generative archive. First, the generative archive underscores the act of recognition, where all documents within the archive, regardless of origin, are homogenous in their claim to historicity. Secondly, the generative archive emphasizes remembrance, as it highlights the processuality of computational mediation and the act of revisitation. Finally, the generative archive invites repair, where CG gestures towards utopic futurities while simultaneously addressing existing gaps. Through these tendencies, the generative archive points to the unique ways in which computational mediation translates and transforms occidental epistemologies and perceptions of time. This theory of the generative archive seeks to answer the question essential to these artworks and ongoing debates on AI-generated content, which is ultimately perceived through CG: How does CG shape, cut, and fit into the archive and its organization of truth?

2. The tendencies of the generative archive

2.1. Recognition

Enwezor’s Archive Fever covers the archive and its fictions as represented by analog photography and film. What happens, however, when the CG fictions are woven into the archive? Wolfgang Ernst asserts that, in a digital archive, the contents of individual documents are not as important as the new set of relations developed within the digital archive. The digital archive makes clear the relationships between statements within the larger field, rather than focusing on individual statements themselves (Ernst 2002, 89, 94, 134). Speaking on databases, Lev Manovich also notes that in a database, individual documents of the archive are homogenous to each other in significance (Manovich 2007, 39). What is more notable is the collection itself, and the way in which the collection is structured within the program and within its hardware storage.

Following this, I argue that the generative archive allows the rendered to remain equivalent to the recorded as a “trace” of the past. The generative archive is dependent upon not just the relation between the archival documents themselves, but between the viewer and object as well. For example, Heesoo Kwon highlights the absence of women in Korean genealogical records, which only preserve the names of cisgender men. To counteract this, she collects stories about her female ancestors and maintains her own shamanist-inspired religion, Leymusoom. Throughout her oeuvre, she fleshes out woman-centered family photographs drawn from her personal archive with 3D renderings of her ancestors, femme god spirits, and imaginations of herself.

This method comes through in Kwon’s Premolt series, of which features fifteen composited images. Here, the 3D figures glimmer into place. At one angle, we see only the original family photograph in Premolt 2: a smiling group of three woman, standing beside each other on sandy shores in the middle of the frame (Figure 1). At another, with a slight shift in perspective, we can see the final composite with her ancestral inserts. Two naked women stand beside the original trio, beaming at the viewer with a ghostly translucence. A third lounges about on the sound in the front, luxuriating under the warm sun. Just as apparent as the 3D renders are at one angle, they just as easily disappear in the next.

The images exist as prints and installations, but the spectrality of the 3D-rendered figures comes through best in her lenticular lightboxes. When we look at the Kwon’s lenticular lightboxes, we find that the image shifts into a multistable relation with the viewer. Like Don Ihde’s observations of the Necker cube, the recorded and the rendered become engaged not in a two-step, frame-by-frame relation, but rather, a figure-ground one (Ihde 2012, 37-40). The rendered shimmers into and recedes from the foreground dependent upon the subjective position of the viewer. Only through self-reflexive observation is one able to perceive of the shifting World in relation to the Self. The essential ambiguity of the multistable object invites and ejects the viewer a series of oscillations of looking through, at, and into a technological mediary.

Fig. 1. Heesoo Kwon, 2 out of 15 in Premolt series (2019), lenticular lightbox 22 x 32’’.

As a result, when we regard Heesoo Kwon’s own revisitations of the archive, we are faced with a paradox of recognition. In order to experience both the quick and the dead in the same frame, we must engage in the cracks apparent between the composited images. To face the utopic future of a recorded-rendered composite, we must also acknowledge the absence that divided the images to begin with. CG does not so much fulfill a longing for what is lost, but instead calls attention to what Allyson Nadia Field calls the “archival gaps, lacunae, redactions, and blindspots” (Field 2022). In this way, the composite’s acknowledgement of the seams between the rendered and the recorded – found in the perceived absence of the ancestors –underscores how the generative archive is not so concerned with filling the gaps as it is interested in exposing them. In so doing, the generative archive openly acknowledges the fact of its generative origins, in the name of previous and future generations. To see the “full picture” of Kwon’s composites, one must regard both the rendered and recorded.

2.2. Remembrance

After recognition comes the second tendency of the generative archive: the urge towards remembrance. This is exemplified in Jon Rafman’s six-minute short, You, the World and I (Figure 2). Here, an anonymous, male-coded narrator reports that he lost contact with the love of his life. After scouring the Internet, he finally stumbles upon a photograph of Her — an unnamed, nude woman facing the Adriatic Sea – on Google Street View. Then, hoping that the Google Street View camera captured more accidental photographs of Her, he revisits the areas that they have been to together before. CG images of the lands they visited together appear on screen. The Atacama Desert. Chicago highrises. Nondescript parks. His quest comes to a screeching halt, however, when he finds the photograph has disappeared. The film then cuts to a blacked-out screen, that claims that “this image is no longer available” (Figure 3). After ten seconds of mournful pause, the video becomes available to play again.

Fig. 2. Jon Rafman, still (photograph of Her from Google Street View) from You, the World, and I (2011), 6’’ video.

The viewer is confronted with a choice. Do we replay the video – thus repeating the search, and loss, for Her – or do we stay content with where the narrative left off last? The video’s replay reveals that technically, we can never see the piece in its original form again. A pervading sense of paranoia seeps into each re-watch. The narrator’s voice beholds an all-too-human inflection, but at the same time maintains an 8-bit timbre. The blurred lines between the animated and photographed make one question the validity of the found footage. How are we to make of this seemingly innocuous romance wherein which the narrator generates a whole archive of material around a subject that wanted to leave no photographic trace?

Fig. 3. Jon Rafman, still (“This image is no longer available”) from You, the World, and I (2011), 6’’ video.

This haunting of the past upon the present recalls Edmund Husserl’s phenomenology. He asserts that there is no single originary point for our experience of the past. However, for every point and interval that puts the past behind us, there is a “continuous generation” of consciousness of the new now. Retention enables the experience of duration, such as stringing together distinct notes of a melody (Husserl 1964, 129-132). Primary retention allows one to make sense of past information in the present, like listening to the melody at-present; secondary retention allows one to recollect, recall, or re-discover details previously unnoticed when the song is listened to a second time (44-48). Bernard Stiegler, who writes in relation to television, argues that tertiary retention (or “protention”) makes way for the record to effectively hijack consciousness (Stiegler 2011, 26-31). In other words, once one knows the end, one can never revisit the first moment of the encounter.

However, by replaying the video, the narrator’s search for Her survives a few more minutes. In Jacques Derrida’s seminal 1995 essay, Archive Fever, he writes that the archive does not just keep records, but is engaged in an act of continuous recording. This leads us to the verbal mal d’archive, or titular archive fever: the urge to generate more material to document and preserve in the archive, so as to memorize, repeat, reproduce, re-impress the past, to avoid oblivion (Derrida 1995, 14). Indeed, in medium-technical terms, the front-end screen allows for this archive fever to continue. Boris Groys notes that watching a video replay is akin to seeing a new musical performance each time; each digital reproduction is unique to its moment, surface, screen, software, and display (Groys 2016, 89-90). The aura of a reproduced digital object in this way is not in the “original copy,” but instead is in each experienced iteration. Shane Denson attributes this to the discorrelation between human and computer vision. Each “screening” is a refilming of the film, or the “generation” of images on screen. As such, “to ‘render’ the film computationally is in fact to offer an original rendition of it, never before performed, and hence to re-produce the film through a decidedly post-cinematic camera” (Denson 2020, 40). With each replay, the photograph of Her can temporarily “live again” – albeit in a different form from the first viewing – so long that the videos remain online.

Rafman’s video reveals two tensions in the act of remembrance. Knowing the end makes one suspect the beginning. By the same token, reaching the end invites a desire to revisit the beginning once again. In trying to re-access the generative archive, we are looking to generate – and regenerate again – that first moment of discovery. Paradoxically, repeating the exact experience of access – that is, to replay the video of the generative archive – only removes us further and further from the serendipitous moment of encounter, until one forgets the first viewing to begin with. In other words, the Derridean death drive in the archive is not so much a move towards primal survival as it is an act of pre-emptive mourning. It is framed by an embodied melancholy, a longing to not just see, but to re-see what has once past. As art historian Michael Ann Holly points out, desire undergirds the act of looking, an embodied will to retrieve and re-animate an artifact of the history (Holly 2007, 8-9). The impulse to remember is corrupted by its own fear of forgetting.

2.3. Repair

Resuscitating the archive can also come at a political cost, however. “To read the archive is to enter a mortuary,” Saidiya Hartman warns, “[as] it permits one final viewing and allows for a last glimpse of persons about to disappear into the slave hold” (Hartman 2007, 17). Lisa Lowe similarly cautions against rushing towards emancipatory restoration. Freedom, she points out, is a function of modern humanism that depends on “an economy of affirmation and forgetting” of its conditions of (im)possibility (Lowe 2006, 206-207). In a later essay, Lowe notes that we should thus remain suspicious of privileging recovery, as one “risks reproducing the very forms of violent erasure” that enable the continuity of archives to begin with (Lowe 2015, 98). Imagining each revisitation to the archive, and our potential freedom from it, thus requires a critical examination of what dehumanizing conditions allowed the archive to persist to begin with, in its current form.

In other words, Hartman and Lowe advise a serious hesitancy when confronted with the possibility of revisitation, even with the genuine interest in retroactive repair through CG and AI. Rashaad Newsome’s Being: The Digital Griot (2019-present, Figure 4) presents an apt case study here. The current iteration of the project features Being, an AI program represented by a nonbinary, humanoid avatar fashioned after the West African griot figure. Being is trained on audience feedback, vogue dance, and writings by Paulo Freire, bell hooks, Michel Foucault, Cornel West, Audre Lorde, and Dazié Rustin Grego-Sykes. Newsome’s project aims to push back against Western interpretations and dictations of history by inviting viewers to respond with their own lived experiences with various forms of systemic oppression. Through the artwork, he fashions Being into a model of resistance against hegemonic order.

Fig. 4. Rashaad Newsome, Being: The Digital Griot (2024). Performance piece, locations variable.

However, Being is positioned at a peculiar crossroads. On one hand, it actively rebels against the ways Western modes of recording knowledge have historically been weaponized to typologize the Other through its collection of individual stories. Foucault notes that epistemological endeavors are necessarily violent; he writes, “knowledge is not made for understanding; it is made for cutting” (Foucault 1971, 172). At the same time, the production of Being’s responses runs the risk of repeating the mistakes of its technologically-enframed forebears. By using the very technologies of data collection for large-scale AI systems, Being finds its words in locating patterns through its rich tapestry of human voices and individual stories. As a result, Being runs the risk of representing a generic average.

This causes a few core questions to cycle through the work, then. How does computational media learn and shape representations of the human body through motion capture and cybernetic feedback loops? How can generative media critically commemorate the past and speculate upon the future? Ultimately, through the digital, how can we recover histories and present forms of the human, a form of recognition historically denied to historically undermined peoples?

3. Some concluding thoughts: towards processuality

These questions are daunting, but lead us to a general observation of the generative archive: it seeks to disrupt concrete or universalizing modes of preserving knowledge by weaving CG and AI into its histories. This becomes clear through the three tendencies briefly discussed here. Through recognition, rendered possibilities have equal stakes to historicity as the recorded. Through remembrance, back-end modes of storage and front-end screens of replay create space to commemorate archival objects in avoidance of oblivion. Finally, through repair, CG and – eventually, AI – seek to spin alterities of future, past, and present, endowing particularities not just to the sensory body, but of the rich variety of stories that constitute the lived human experience.

As a concluding note, I would like for us to pause upon what consists of the present. Mark B. N. Hansen notes in his article, Living (with) Technical Time, that the openness and flexibility of computational mediation allows us to extend beyond the divisions of a temporal before and after, and instead remain fully integral to the ongoing process of happening (Hansen 2009, 299-302). Computational mediation is not placed so much in the unilinear conception of a past-present-future, but rather, is placed in the processual. Moreover, as other existentialist-phenomenological approaches show, the human itself is never quite complete. I would like to close here with some words by Frantz Fanon, who notes that – despite external perception of the racialized body – the individual is always in a process of becoming. He concludes Black Skin, White Masks, with an optimistic note reflective of his interiority: “In a world through which I travel, I am endlessly creating myself” (Fanon 1952; 1967, 228). In this vein, I invite us to reflect upon the generative archive as a utopian notion. Here, history is not limited to the finite preservation of the recorded. Instead, the generative archive opens up to the many infinitudes generating to construct the now.

Acknowledgments

This project is still in development, as it will lay the foundations for my dissertation. Even so, many thanks to those who have helped me with figuring out the kinks so far, including: Jean Ma, who saw the first draft of this paper; my advisor, Shane Denson; my xCoAx mentors, Miguel Carvalhais and Luísa Ribas; and of course, everyone else at the School of X.