Suspicious Behavior
Environmental & social costs
Hands-on experimentation
Bias & discrimination
What can I learn?
An interactive piece that reveals the hidden human labour behind "intelligent" computer-vision systems. Through a physical home-office set-up and an image-labelling tutorial for a fictional company, users experience the tedious, low-paid work of the outsourced annotators who teach machines to interpret human actions.
Core insight
"Suspicious behaviour" is not something a camera objectively detects — it is a category defined by underpaid human labellers following instructions. The intelligence in computer vision is borrowed human judgement, with all its bias and precarity.
How to use it in daily work
A visceral way to understand and explain that AI vision systems are built on human labour and human definitions — making the abstract "training data" concrete.
- Use the experience to explain to a group why a surveillance system's idea of "suspicious" can be biased from the very start.
- Draw on the annotator's-eye view to discuss the hidden workforce behind everyday AI.