SESSION: AI and the Artworld: Art History and the Generative Imagination

With artists like Pindar van Arman using agentic AI while Holly Herndon and Sasha Stiles forge transmedia human / technological collaborations, is it time to move away from definitions of creativity that are rooted in the human experience? How might we approach AI God, for example, which was painted by humanoid robot Ai Da, and sold for over one million dollars at Sotheby’s? What frameworks, then, can help us approach AI art critically?

This is a state-of-the-field session that considers computer-generated visual culture in the broadest sense, with a particular interest in methodological issues. Can we establish new criteria for critique in a world of obscure datasets, unpublished algorithms and blockchain initiatives? Can we find solutions for methodological and provenance issues? We encourage papers that examine algorithmic justice, copyright issues, potential risks, biases, and backlashes presented by AI entering the artworld. Gender disparities, ethnic and racial prejudices are well documented, and there is a need to consider questions around cultural appropriateness and representations. Whose aesthetics are being represented, and who is forming the canon? Might AI act as another form of colonialism?

We also invite provocations around collaborating with AI, especially around training and development processes, to ensure rights are considered as fairer AI ecosystems are built. Can we use AI as a tool to reclaim power, support marginalised voices, and address structural biases in the art world?

Session Convenors:

Lorna Dillon,Hong Kong Babtist University

Melody (Zixuan) Wang,University of Edinburgh

Speakers:

Micol Hebron, Chapman University

Malyssa Shaw, Chapman University

From Fembots to “Women Shot AI”, how the patriarchal structures of computing and AI are designed to objectify and colonize the female form

This presentation will feature a series of AI-generated artworks (by the author) that strategically expose and highlight the ways in which generative AI models perpetuate and amplify the sexual and gender biases of cis-hetero-patriarchal colonialism and capitalism. There is a long history of male dominance in the history of computing and of AI, which often omits the contributions and experiences of women, trans and nonbinary people. This has led to a heavily male-dominated perspective – found in the demographics of the history of computing, computer programmers and software engineers, the CEOs of technology companies, the authors of the large language models and datasets, and even the philosophers that are cited when theorizing the role of AI today. This paradigm leads to radical deficits and perversions in the representations of female-bodied experiences and histories in generative AI models. There has been a legacy of femme-presenting chatbots and service bots; a large market for female sex robots; a YouTube channel titled “Women Shot AI” (consisting only of AI generated videos of women being shot); a rising crisis in self-harm and suicidal ideation among teen girls; male-authored tutorials on how to create female AI models for Only Fans accounts; and support for creating “Erotica” (pornography) on OpenAI and other AI platforms. Colonialist and Orientalist strategies have been employed to perpetuate the domination, objectification, hyper-sexualization and denigration of the roles and images of women through AI. This presentation will show that by tracing the misogyny inherent in the history of computing, leading to AI, we can understand the source of the alignment problems in the visual outputs of generative AI, and thus propose countermeasures to address them. 

Melody (Zixuan) Wang, University of Edinburgh

How Does AI Represent Social Concepts? Examining the Visual Representation of Care in Text-to-Image Tools

What does “care” mean in our lives, and how is it represented by AI? 

Developments in text-to-image (T2I) AI tools enable users to generate highly realistic images from textual descriptions. These AI-generated images are increasingly produced and used across creative, professional, and everyday contexts — yet the assumptions embedded in AI outputs are not fully explored.

This study investigates how the concept of “care” is (mis)represented in AI-generated images through a visual analysis of 140 images of “care” generated by Midjourney. The findings indicate that Midjourney is reproducing stereotypical, reductive, and inaccurate representations of care that often conflate care with older age, reduce care to formal care and a one-to-one relationship, and visualise care as feminine and through touch. Through the analysis, the study invites critical reflection on the biases and misconceptions about care in our society, while considering how AI technologies are mediating and shaping our cultural understandings of care. The study further explored the potentials and pitfalls to mitigate these biases and create alternative representations through prompt engineering, and developed a reflexive prompting framework guiding practice. Overall, the study aims to raise public awareness and spark discussions about AI, care, and visual culture. 

Michiel Willems, University of Leuven

Speculative Archiving in the Age of Artificial Intelligence

This paper delves into the realm of speculative archiving, conceptualizing it as an emergent practice situated at the nexus of artificial intelligence (AI), augmented reality (AR), and digital museology. The paper explores the manner in which artists and institutions utilize digital technologies to reimagine and contest the colonial architectures of the archive. Projects such as the Virtual Museum of Stolen Objects, developed by UNESCO, exemplify the potential of computational reconstruction to function as both a preservation tool and a critical lens. This approach utilizes AI-driven modeling to facilitate the virtual restitution of looted artefacts, thereby offering a novel approach to addressing the issue of cultural patrimony. A similar phenomenon can be observed in a series of guerrilla AR-interventions in the Metropolitan Museum. In this case, digital overlays of Indigenous artworks are activated directly on top of canonical displays, thereby performing a counternarrative gesture that reclaims visibility within institutional space. Works by Auriea Harvey, Linda Dounia, Rodell Warner, and others expand the speculative archive into artistic realms. The paper will argue that AI functions as both a medium of reconstruction and imagination. AI can be considered a technology that transforms the ontology of the archive from a static repository to a dynamic process. This approach, known as speculative archiving, challenges the authority of traditional archival practices and proposes a reparative, plural reconfiguration of cultural memory. The paper posits the following research question: What new forms of truth, fiction, and futurity emerge from this engagement with the past?

Miao Wang, University of Exeter (Online Presenter)

From Aura Loss to Collaborative Generation: A Critical Framework for Algorithmic Bias in AI Art

BThis paper examines the methodological challenges for art criticism presented by AI-generated art. Proposing a shift from Walter Benjamin’s aura to what Byung-Chul Han terms an aesthetics of smoothness, it contends that AI’s data-driven output is visually seductive but critically deficient. A comparative analysis of Ai-Da and the OpenAI Ghibli-style filter reveals that meaning is not intrinsic to the work but is produced through public “collaborative generation.” This collaborative mechanism remains systematically constrained by three layers of algorithmic bias, empirically evidenced in our case studies as: (1) cultural-representational bias, manifest in the Ghibli filter’s reduction of diverse cultural scenes into Western-centric animation tropes; (2) creation-theory bias, evident in how AI art reduces creative practice to computational mimicry through data fitting and style transfer; and (3) systemic bias, operating as a form of “data colonialism” where tech giants discipline artistic generation by stripping cultural symbols from their native contexts. In response, this paper constructs an operational “Participatory AI Art Critique Framework” comprising two complementary analytical pathways: social discourse mapping, which traces how meaning is negotiated through public engagement with AI art; and aesthetic value assessment, which evaluates works through the dual criteria of cultural extraction and generative respect. This theoretical apparatus transforms abstract critique into actionable research practice, providing methodological support for constructing a more equitable AI art ecosystem.