Occupied with Traffic

Irini Kalaitzidi

Artwork Description:

A performance that explores movement undergoing AI-mediated classification. Centred around what appears to be the act of riding a bike, Occupied with Traffic invites audiences to observe and repeatedly identify this familiar human action. The process becomes an analogue echo of AI classification systems, rooted in surveillance contexts, that capture bodies and frame them with textual tags such as “walking”, “sitting” or “eating spaghetti”. How can an action be enclosed within one fixed label? Can this deterministic system find some softness in doubt and lower confidence scores? Walking with 99% confidence, or 85%, 57%, 34%, 7%. While structured like a playful experiment, the performance looks at how classification may seem harmless in closed systems, but begins to falter and eventually crash when applied to the complexity and messiness of life. In Occupied with Traffic, the body invites in classification as a game, and rejects it as a cruel reality. Parallel to the performance, the work expands into a video installation in which the performer’s body is subjected to actual AI classification systems. A human-action-recognition model and an object-detection model continuously attempt to stabilise movement into recognisable categories, producing labels that shift, overlap, fail, or contradict one another. As the work moves through different presentation contexts —a gym, a jazz club, a basketball court, a church— the surrounding environment influences the logic of classification itself. The installation exposes how these systems do not simply recognise actions, but choreograph realities through context, probability, and reduction.

Project Collaborators:

Performance and Tech Development: Irini Kalaitzidi Music and Sound Design: Thomas Rosser Technical Support: Stathis Doganis The first stage of development for Occupied with Traffic took place at ImpulsTanz ATLAS (2024) with the support of Onassis Stegi’s Outward Turn Program and Onassis AiR.