SESSION: Curating with AI: Risks and Opportunities

Computational technologies falling under the umbrella term ‘artificial intelligence’ have been embraced and critiqued by art practitioners and cultural institutions around the world. This session explores the potential uses and risks of incorporating AI into the curation of art exhibitions. The idea that a ‘self-learning human-machine system’ could curate an exhibition was debated by the Liverpool Biennial and the Whitney Museum of American Art (Impett and Krysa 2021). Combined machine curation and audience interaction was incorporated into the 2023 Helsinki Art Biennial (Del Caeillo et al., 2023), and in 2024 the Nasher Museum of Art used ChatGPT to curate a supposedly ‘groundbreaking exhibition’ (Richardson 2024). Researchers have also explored whether computer vision can distinguish between exhibitions curated by a human and by an algorithm (Van Davier et al., 2024). By taking curation out of the hands of humans, can AI be used to decolonize museum practices and propose innovative connections between artworks? Might the sheer use of technology in curation have a democratizing effect by stimulating the interest of new audiences? Alternatively, does the delegation of curation to a machine erode a fundamentally social aspect of museum practice? Aside from its environmental impact, does the use of AI create a hierarchy that privileges those institutions with access to sophisticated technical resources and staff capacities? As public and private funding floods into technology rather than arts and heritage, there is a pressing need to determine the gains and losses of AI curation and the implications of a new techno-curatorial imaginary.

Session Convenors:

Kathryn Brown, Loughborough University

Alison Kahn, Loughborough University

Speakers:

Serena Iervolino, King’s College, London

Alasdair Milne, King’s College, London

Curating AI-Driven Art: Organisational Change, Expertise, and Institutional Dynamics

While debates on “curating with AI” often focus on machine learning as a tool to support curatorial decision-making, this paper examines the distinct challenges of curating AI-driven art – a term we use expansively to refer to artistic practices that incorporate AI technologies, particularly machine learning, as part of the creative process, attributing to these tools roles such as pattern recognition and decision-making. Drawing on findings from the Curatorial Strand of the AHRC-funded project Creative AI: Machine Learning as a Medium in Artistic and Curatorial Practice (2020–2022), which combined workshops with curators, producers, and policymakers alongside interviews, the paper explores how these works destabilise established curatorial and organisational logics. Exhibiting AI-driven art requires institutions to confront distributed authorship and data-dependent practices, while developing infrastructures to preserve and present process-based or adaptive works. The research highlights that these demands trigger significant organisational shifts – museums increasingly rely on interdisciplinary collaboration, technical partnerships, and novel forms of curatorial expertise that blur traditional boundaries. These shifts present opportunities for innovation in cross-disciplinary practice and interpretive frameworks, while also introducing risks related to authority, accountability, and resource inequality. Far from a neutral technological adjustment, the curation of AI-driven art becomes a site where institutional and professional roles and responsibilities are actively renegotiated. By situating these transformations within broader debates on organisational change, the paper argues that AI-driven art functions as a lens to reassess how museums define expertise, organisational structures, and institutional identity in the age of artificial intelligence.

Seunghye Sun, Korean Cultural Centre, London

Techno-Heritage Curatorial Care: Human–AI Collaboration in Digital Heritage Exhibitions at KCCUK

This paper presents a curatorial case study on how large language models (LLMs) have been applied to re-contextualise non-English cultural heritage in the digital exhibitions at the Korean Cultural Centre UK (KCCUK) —Digital Heritage: AI with You (2024) and Endless Bonds: AI & Heritage (2025). These projects explored how AI can act as an auxiliary curator in translating Korean heritage for global audiences while expanding curatorial imagination. In 26 Your Korean Words (2024), Midjourney-generated images visualised Korean words listed in the Oxford English Dictionary, revealing AI’s image biases towards Korean language and culture. The 2024 exhibition employed three modes of collaboration: (1) Dragon-Phoenix Incense Burner used AI as a supplementary interpretive tool offering comparative Eurocentric readings; (2) Pensive Bodhisattva invited contemporary artists Shin Seung-back and Kim Yong-hun to convert emotion-recognition data from audience responses into sea sounds, embodying the metaphor of crossing an ocean of reflection; and (3) AI-generated descriptions of British Museum objects were corrected and annotated by human curators referencing data from the National Museum of Korea. In Endless Bonds (2025), AI narratives compared Silla crowns with British regalia through human–AI dialogue in a web-novel format. The study concludes that when curators possess deep contextual knowledge, LLMs amplify curatorial creativity; when not, AI explanations tend to become generic and mechanistic. The paper proposes “Techno-Heritage Curatorial Care” as a new framework to integrate AI collaboration into heritage curation, contributing to UNESCO’s future approaches to digital memory and global cultural accessibility.

Chandi Jeswani, Independent Researcher, Dubai

Sacred Absences and Algorithmic Gazes: Towards a Critical Ethics of Machine Curatorship (Online)

As museums and biennials increasingly experiment with AI-driven curation, the question of who—or what—sees, selects, and interprets art becomes newly fraught. This paper investigates how algorithmic systems reconfigure curatorial authorship, visibility, and ethics through three intertwined lenses: absence, the sacred, and labour. Building on research into digital archiving and diasporic memory, the paper begins by examining algorithmic curation as a new regime of visibility—one that translates aesthetic judgment into data-driven pattern recognition. While framed as democratizing, such systems often reproduce historical exclusions, codifying absence by privileging datasets aligned with Western epistemes. At the same time, when AI encounters artworks embedded in sacred or ritual temporalities—from the cyclical spatialities of Varanasi to the metaphysical aura of ritual objects—it exposes a deeper incompatibility between computational linearity and sacred multiplicity.

These case studies reveal the algorithmic gaze not as neutral perception but as an infrastructural worldview that flattens plural ontologies into calculable form. Yet behind this “machine vision” lies a hidden network of human labour—data annotation, content moderation, and curatorial oversight—whose invisibility sustains the myth of automation. By tracing these visible and invisible processes, the paper proposes a framework of critical techno-curatorship: an ethics of seeing that acknowledges both the sacred opacity of the uncuratable and the social infrastructures underpinning algorithmic creativity. AI curation, far from autonomous, becomes a mirror through which the museum confronts its own hierarchies of knowledge, power, and presence.

Giselle Beiguelman, University of São Paulo

Ana Gonçalves Magahães, University of São Paulo

Thiago Hersan, Parsons School of Design

Meta-Acervos: AI, Metadata, and the Politics of Visibility in Brazilian Museum Collections

We present our methodology and prototype for a meta-collection system called Meta-Acervos that uses different AI models to recombine existing archive metadata. This work was carried out within the context of a digital humanities research group (Digital Collections and Archives) that aims to create strategies for the visibility of underrepresented narratives in the fields of art, architecture and design. The current interface focuses on combining and expanding information from 17 Brazilian museum collections available in public datasets and APIs like Wikimedia and the Brasiliana Museus project, with 4000 artworks from the XIII to XXIst century. Our methodology makes use of open-source multimodal models to annotate, cluster and classify paintings and drawings according to their visual and semantic characteristics. This can be used to visually explore aggregate characteristics and patterns from works. The diversity of the collections available in Meta-Acervos leads to significant ambivalences in the use of AI for the treatment of museum collections and, consequently, the field of curation. The system enables image-based searches that allow combinations that break with the traditional canons of art-historical databases. These features suggest speculative and exploratory curatorial approaches that bring forth alternative aesthetic and historical narratives. On the other hand, the results of the visualization and search filters reveal the fragility of AI models when dealing with modern and contemporary artworks and the weight of biases in the interpretation of images depicting marginalized people and histories.