The installation comprises both hardware and software. The software enables the installation to infer simple characteristics of observers and to measure how these characteristics change. The physical output of the installation evolves though machine learning in the software. This is driven by the information inferred from changes in the characteristics of observers relative to the behaviour of the installation itself.
By inferring differences in observers’ characteristics, the intention is that over time the test installation evolves to presents itself to different observers differently; to new observers in a manner informed by interactions with previous observers with similar characteristics. Put simply sometimes it learns to smile at children and frown at grown-ups!