Cartography Unseen is an interactive game prototype in which visitors explore an AI-generated landscape that never settles into a fixed map. Each movement steers a walk through the model’s latent space, continuously changing the world ahead. The player becomes an explorer of the AI’s black box, navigating and mapping an uncanny landscape that shifts across viewpoints, moments and play sessions.
The work treats AI’s algorithmic opacity as an artistic medium, exploring what it can hide, reveal and transform. Informed by Philip Galanter’s theory of Complexism, it asks how play emerges through the interaction of human and computational agents. By giving the player influence over a world they cannot fully predict or control, Cartography Unseen questions conventional map and level design and examines how agency is shared between player and computer.
Demonstration
A landscape generated by the explorer’s movement.
Prototype in real time · 00:36
Exhibition maps
Explore the recorded journeys.
Explore the 100 longest visitor journeys recorded from 21 to 27 September 2026, ranked by distance travelled.
Drag to orbit. Scroll to zoom.
Exhibition photographs
The installation and its visitors.
Projected journey map and generated landscapeVisitor playing Cartography UnseenPlaying within the projected worldVisitor viewing the installation
Research
Context, contribution and significance
Research context and approach
Generative AI’s latent space has attracted interest from artists and computer scientists as a black box whose outputs emerge through interpolation across many dimensions of data. Without contextual spatial awareness, the model produces images that sit between the familiar and the unknown. The project connects this instability to Freud’s Unheimlich, the return of the familiar in defamiliarised form.
Philip Galanter’s theory of Complexism (2016) provides a framework for considering interacting systems and agents whose collective behaviour exceeds the sum of their parts. Within this framework, the prototype maps player movement to latent-space walking. Rather than generating a level that remains persistent once created, the system generates the view continuously in response to the player’s controls.
Research contribution and significance
The prototype explores a method of procedural map and level design in which generative AI participates in the act of play. Its contribution lies in connecting a technical experiment in movement-driven generation with an artistic inquiry into uncertainty, perception and player/computer agency.
This approach opens possibilities for experimental game design and interactive media practice by treating the model’s emergent behaviour as material for exploration. The demonstration, recorded journeys and exhibition photographs document the work and its presentation.
Credits and funding
Authors: Nhat Pham Hong, Agnieszka Kiejziewicz, Ricardo Arce and Kok Yoong Lim. Technical development: Tom Nguyen.
Part of Emergent Play, supported by the School of Communication and Design’s 2025 School Research Seed Funding, RMIT University Vietnam.