INTERNATIONAL OPEN CALL FOR ARTISTS

Cluster A · Conceptual and artistic · A1

ARCHIVAL SILENCES

Every archive is the result of decisions about what was collected, catalogued and kept. These decisions are shaped by national and regional contexts, institutional priorities and the availability of funding. Historically, some voices are overrepresented while others are excluded. GenAI systems trained on such material inherit and amplify these asymmetries, presenting a partial record as a complete one.

Suggested experimentation

• Identify a specific absence, exclusion or under-description in a partner collection and make it perceptible.
• Use GenAI tools to expose bias in classification, description or retrieval.
• Work with communities or practitioners connected to the missing material, where appropriate.

What does this archive not hold, and what happens when a machine is asked to speak on its behalf?

Cluster A · Conceptual and artistic · A2

Creative misuse

Media art has a long history of productive misuse: pushing tools beyond their intended function to reveal their flaws. Applied to GenAI systems, misuse becomes a method for making the logic of the model visible: through failure, saturation, refusal, recursion or deliberate mismatch between material and model.

Suggested experimentation

• Build work around the failure modes of a GenAI system rather than around its successful outputs.
• Feed a model material it handles badly, such as poor-quality or damaged video, minority languages, technical documentation or non-standard notation.
• Document the misuse as a reproducible procedure, not only as a result.

What does the system reveal about itself when it breaks?

Cluster A · Conceptual and artistic · A3

From document to behaviour

Media art frequently exists as behaviour, process, interaction or computation rather than as a stable object. Archives therefore preserve installation procedures, source code, dependencies and tacit knowledge alongside files. This documentary residue is itself artistic material: an installation manual, a conservation report or an artist interview describes a work that may no longer be executable anywhere.

Suggested experimentation

• Respond artistically to a work that survives only through its documentation, without necessarily reproducing it.
• Treat technical documentation, conservation records or restoration histories as primary source material.
• Produce a new work whose subject is the gap between the record and the experience it describes.

If the only thing that remains of a work is the description of how to install it, what can be made from that?

Cluster B · Technical · B1

Heterogeneous archival data

There is a lot of diversity in archival documentation: some institutions apply recognised metadata standards; others have built bespoke systems for media art. Levels of description range from basic inventory entries to full technical and conservation records. Any artistic pipeline that assumes clean, uniform, richly described data will fail on real archival material – the diversity is not a defect to be corrected but a property to be worked with.

Suggested experimentation

• Build a way of working (a tool, process or pipeline) that still functions when archival descriptions are incomplete, inconsistent or non-standard.
• Make the structure of the metadata itself visible or audible in the work.
• Add to or align archival metadata (for example, writing new descriptions or matching terms across collections) in a way that is reversible and clearly separated from the original records.

How can a work be built from an archive whose records are incomplete, inconsistent or described in different ways?

Cluster B · Technical · B2

Small collections, large models

Large generative AI models are trained on enormous amounts of data, while an archival collection may hold only a few dozen or a few thousand records. How can an artist work with a large model without the model overriding what makes the collection specific?

Expected experimentation

• Compare approaches (for example, giving the model access to the collection at the moment it answers, lightly retraining an existing model on the collection, or building a small model from scratch) and document the trade-offs.
• Test where the specificity of the collection survives the process and where it is lost.
• Make the size and limits of the source material explicit in the work.

At what point does the collection stop speaking and the model start interfering?

Cluster C · Ethical and legal · C1

Authorship
and attribution

When archival works by identified artists are processed through a GenAI system by a third artist, authorship becomes distributed across original creators, the institution that holds and describes the material, the residency artist and the systems used. The residencies ask for positions on this question that are made explicit in the work rather than resolved silently in a credit line.

Suggested experimentation

• Develop and state an attribution model for the resulting work (a clear statement of who is credited for the work and how: the original authors, the archive, the artist and the AI systems used).
• Make the contribution of source authors visible to audiences, not only to institutions.
• Address the position of the archive itself as a co-author of meaning.

Whose work is this, and how would an audience know?

Cluster C · Ethical and legal · C2

Contextual integrity

Contextual integrity means keeping an archival record connected to the context that gives it meaning: where it was made, by whom, for what purpose and how it was kept. When materials are detached from these documentary frameworks, historical meaning is lost. A GenAI system operates by removing context: it flattens provenance, date, medium, intention and controversy into features. Work in this challenge deals with what is lost in that operation, and whether anything can be built to resist it.

Suggested experimentation

• Identify what a specific record loses when it becomes training or conditioning material.
• Design forms of presentation that keep source context accessible alongside generated output.
• Work with politically or historically sensitive material, highlighting the care this work requires.

What did this record mean before it became training data?

Cluster D · Preservation and documentation · D1

Reactivating ephemeral works

Many works in media art archives are ephemeral: installations, interactive pieces and performances that survive mostly as documentation, and that depend on equipment, spaces and sometimes software that may no longer be available. This challenge asks artists to reinterpret such a work from the host’s collection so that it can live again in a new format, and to record what is kept, what changes, and what a future re-presentation would need. GenAI is part of the process: for example, to reconstruct, regenerate or translate elements of the work that can no longer run as originally made. Unlike From document to behaviour (A3), which uses documentation as material for a new work, this challenge is about bringing a specific work back into operation.

Suggested experimentation

• Distinguish what is fixed in the work from what is generated at each execution.
• Record the work in operation with an audience, following the audiovisual documentation practices used across the partner archives.
• Specify what a legitimate future re-installation would need to reproduce, and in what kind of space.

What must be kept for an ephemeral work to live again in another format, and what can change?

Artist Fee

€5,000

Fixed fee per artist, paid in instalments linked to the stages of the residency, as set out in the artist contract.

Travel and Accommodation

Further information regarding travel and accommodation arrangements for the residencies will be provided once officially confirmed.

Environmental Responsibility

Artists are encouraged to consider the computational cost of their methods and models, as well as low-carbon travel and responsible production choices where feasible.