Algorithmic Cultural Vandalism (ACV) is an ongoing research project that explores what happens to symbols when meaning is deliberately suspended. It started from questions around cultural appropriation, AI image generation, and the violence implicit in classification systems, and gradually became a long-term investigation into the morphological roots of language, drawing, and signification.

ACV takes the form of an interactive installation and workshop format built around a drawing robot trained on a large, heterogeneous archive of symbols, glyphs, scripts, talismans, diagrams, and visual signs collected from across global cultures. This archive is intentionally anarchivistic. All captions, classifications, metadata, and cultural labels are removed. Provenance is erased. Symbols are reduced to pixels and treated as pure form. The goal is not interpretation, but exposure: freeing signs from fixed meanings and allowing them to operate as traces, referring only to other signs, in line with Derrida’s notion of arche-writing.

The archive itself emerged through a process that is openly extractive and ethically unstable. Images were gathered through “wild downloading,” using browser searches conducted primarily in English, reinforcing the asymmetries and colonial biases of digital access. Sacred and magical symbols were collected without context, raising questions about appropriation and power. The archive reflects the privilege of knowledge without ethics, a chaotic accumulation that mirrors how contemporary technologies consume culture.

The AI is trained in an unsupervised manner on this material and connected to a mechanical drawing arm. Human participants interact with the robot by drawing in turn, producing pseudo-symbols in a continuous exchange. The system operates through reinforcement learning: a visual “Critic” evaluates the drawings in binary terms, while the AI learns how to manipulate morphology to score better. Meaning collapses into pixels. Humans, meanwhile, instinctively try to reintroduce sense, narrative, and intention.

Workshops and installations revealed recurring patterns. In social or mediated settings, participants tended toward figurative, controlled drawings. In isolated, silent contexts, they were more willing to scribble freely, trusting the process and the machine. People appeared more comfortable attributing non-representational scribbling to a robot than to themselves. Participants without formal artistic training often engaged more openly than professionals, suggesting that the capacity to suspend intentionality is unevenly distributed.

These observations inform the current phase of research, which shifts the focus from symbolic critique to the hypothesis of scribblese: a proto-linguistic, non-referential form of graphic exchange (see Ontogeny and Abstraction). ACV does not claim empirical certainty. It operates as a speculative diagram of Technic, where symbols are cut, flattened, and recombined, and where meaning emerges, disappears, and reappears in the act of drawing itself.


ACV won the Grand Prize at the Campus Award of Ars Electronica 2024.