I often think about the cost of efficiency.

Technology offers speed, precision, infinite iteration, yet my tension with AI has never been about the tool itself, but about the logic behind it: the quiet substitution of reflection with optimization.

Yet optimization itself is never neutral. The architectures of AI: its data centers, training models, and platform economies are built upon massive flows of capital and labor, optimized not for collective insight but for profitability and scale. What appears as technical inevitability is, in fact, a reflection of institutional intent. In this light, AI is less a tool of intelligence than an agent of power, a system that encodes who gets to see, decide, and create.

When machines learn to create, what remains human?

Negotiated Intelligence is a research exhibition that unearths the creative archives of our time: prompts, drafts, hesitations, revealing the sedimented traces of human decision within algorithmic processes. Here, technology is not background infrastructure but protagonist: a force that grants new freedoms while quietly eroding others. What we are negotiating is not just authorship, but the boundaries of being human. Artists do not automate their labor but choreograph with it, navigating the tension between algorithmic randomness and human coherence, between prompts as inputs and aesthetics as outputs.

This exhibition does not ask whether AI can create. It asks: when creation becomes computable, what new forms of creative agency emerge? What does it mean to negotiate with intelligence that is neither fully tool nor fully collaborator?

More than documentation, this is an archaeological site, one that captures a pivotal moment before these negotiations become automated themselves.

Every project in Negotiated Intelligence carries a negotiation record:

• Human Creator Ledger (artist intentions)
  
• Negotiation Arena (exchanges with AI)
  
• Emergent Hybridity (synthesized outcomes)

Through these strata, the exhibition performs a kind of digital archaeology, excavating how agency circulates between human and machine. The resulting map is less a taxonomy of styles than a record of micro-decisions of when to yield, when to overwrite, when to let the algorithm speak.

In the fragments artists shared, I sensed something deeply human: a desire to locate selfhood amid machine logic, to feel through abstraction, to build meaning in the ruins of efficiency. Daria Belkina's Elsewhere Gallery exemplifies this tension through recursive negotiation. Envisioning gallery space as "a liminal state of transition" where paintings open into other dimensions, Belkina battled AI video models (Kling and Sora) that "repeatedly substituted random imagery into the paintings, producing alien elements." To achieve her vision of seamless portal-like movement, she developed an elaborate method, then repeat across 12 transitions. The AI contributed unexpected aesthetic elements she chose to preserve and the final continuous loop became possible only through this exhausting choreography between algorithmic randomness and human coherence.

Aaron Oldenburg pursued a radically different constraint using only found AI outputs from MidJourney's public archive. His interactive narrative Fractured Star Projections wove other users' prompts into prose, creating "blind" drawings of AI-generated images — a process that "put attention on how my brain takes in information and translates that into arm movement." He described the contrast between "AI's limitations, seen in the back-and-forth between prompter and generator," and his own "lack of any imaginative limitations," ultimately foregrounding biological over digital neural networks.

Marcie Begleiter's Chimera of the Post-Anthropocene moved slowest of all: collecting biological debris from stressed ecosystems, assembling hybrid sculptures, photographing dioramas, writing haiku in response, then feeding those poems to Kling for animation. When the AI "ran off on flights of fancy" with hallucinations, Begleiter rejected these outputs to maintain coherence in her speculative world-building. Her sculptures remained "assembled from nature but not of it" — a phrase that could describe the exhibition itself.

Val Yang's a.ikea offered a radical repositioning: what if we stopped correcting AI and started celebrating its ungoverned imagination? Working at the threshold of AI's early creative development, when outputs were "rough but always full of potentials", Yang deliberately used vague, open-ended prompts to let AI "fill in the blanks uninstructed by human guidance." This inverted the dominant paradigm that values AI outputs by how closely they follow instructions. Instead, Yang embraced AI's surprising and unconventional results, using storytelling and craft to contextualize them into coherent design. The project arrived at a poignant temporal juncture: by late 2024, engineers had already "fixed" common visual errors like malformed fingers, threatening to erase the very premise of Yang's approach. "Errors would be fixed," Yang noted, "but creativity isn't always about correction. Perhaps AI wasn't given enough right to imagine and dream." By shifting AI from assistant to "head of design," a.ikea redefines the human-AI relationship: not as correction of machine failure, but as curation of machine dreaming.

Across these works, artists did not uniformly resist mechanization. Rather, they negotiated with it — sometimes exhaustively correcting through recursive cycles, sometimes rejecting hallucinations as noise, sometimes using AI as emotional scaffolding through creative crises. They redefined creativity not as production, but as choreography: knowing when to lead, when to follow, when to stop the music entirely.

Within this exhibition, artists reclaim the human slowness of making — the warmth of imperfection, the tenderness of a decision that could have been automated but wasn't. In doing so, they redefine creativity not as production, but as resistance to mechanical inevitability.

This exhibition isa living research corpus. A collective story of adaptation, the materials we gather in this exhibition will support new frameworks for understanding creative agency today. This work investigates how artists actually work with AI through two approaches: tracking how artists make decisions and assert control using frameworks like the "Human Creator Ledger" and "Negotiation Arena" to classify their roles (director, editor, curator, or interpreter); and examining how artists respond to AI's unexpected behaviors, analyzing whether they view surprising outputs as failures, happy accidents, or true collaborative moments. Ultimately, this research aims to replace oversimplified arguments with a data-informed understanding of creative agency in human-AI partnerships.

In the long run, I hope Negotiated Intelligence becomes more than an exhibition but a verb:

to negotiate intelligence is to remain awake in an automated world.

---