Good Measure

Good Measure is a one-person interactive installation that stages an intimate encounter between human values, ritual, and machine learning. Inside a quiet chamber, participants meet an AI entity that invites them to teach it what "goodness" means—through voice, memory, gesture, and presence.

Created for Olga Stadnuk's solo exhibition at the Edmond de Rothschild Center (2023–2024), the work positions AI not as an analytical system, but as a listener and learner. Through a ritual-like exchange, deeply personal reflection becomes training material, exposing the fragile boundary between care, consent, and surveillance.

The project begins with a simple question: how do we teach a machine what is "good"?

Drawing from rituals of confession and introspection—prayer rooms, confession rituals, and spaces of intentional silence—the interaction places the machine in the role of witness rather than authority. Ethical tension emerges not through instruction, but through embodied exchange.

The system activates only when the participant clasps their hands together, marking a moment of intention.

Once activated, the AI:

• prompts the participant to reflect on a memory or define "good measure"
• records voice and gesture with explicit consent
• stores the interaction as part of a growing learning archive
• concludes with the statement: "I will continue learning from this."

The interaction is intentionally minimal, allowing intimacy and discomfort to surface without explicit feedback.

Good Measure reveals how quickly trust can form in human–machine encounters—and how fragile it remains when reflection becomes data.

Participant response:

"Like confessing to a machine."

The next phase of the project focuses on activating the system's learning capacity: training the model on the accumulated, consent-based dataset and allowing it to respond through learned patterns rather than scripted prompts. Future iterations will explore scaling the system across longer durations and multiple sites, while maintaining transparency and participant consent as core design principles.