Subject to Interpretation: An Interactive Portrait of Perception
A photograph feels like a record of a moment. Underneath, a camera produces data: the metadata written into the file, and the gestures and expressions that machine-vision systems read and score. Tech companies annotate and train on these signals at scale, by hand and by algorithm, to decide what an image contains and how far it should travel. For the Jakarta International Photo Festival (JIPFest), the challenge was to make this invisible process something a visitor could feel, without asking them to follow the technical machinery behind it.
The research began with what a camera extracts: EXIF data, facial expression, gesture, the raw material of algorithmic classification. The concept turned that into a transaction. A visitor enters a booth and takes a self-portrait. An AI trained to run locally analyzes the image and prints a receipt that itemizes them: what the machine detected in their face, how it classified them, how viral it judged the picture likely to be.
The familiar form carries the message. Everyone understands a receipt, so the unsettling content arrives without instruction, reaching visitors who would never sit down with a paper on machine vision. The portrait a visitor composed and the analysis the machine returns rarely agree, and that gap is the reflection the piece was built to provoke.
Once an image becomes data, its subject stops controlling what it says.