The installation 0.51 sad, 0.23 happy, 0.17 surprise investigates the classification and visualisation of emotions through artificial intelligence. Its point of departure is an image dataset used as training material for AI systems designed to recognise emotions, based on Paul Ekman’s contested theory of universal basic emotions. This model reduces emotions to a small set of clearly legible facial expressions and continues to underpin many applications of emotion recognition today.

Bergelt analyses the algorithmic evaluations of these images and translates the results into abstract colour fields. This reduction to clear forms makes the apparent objectivity of AI visible while questioning its normative mechanisms.

A photographic wallpaper featuring portraits from the training dataset brings the source material of emotion recognition into the exhibition space. Alongside it, a series of seven AI-generated portraits depicts standardised, stereotypical facial expressions. While the colour fields reveal how AI classifies facial expressions as emotions, the generated portraits expose how generative AI reproduces conventional representations of emotion and reinforces cultural norms.

Coding: Lars Bergelt

Installation views, Kunstquartier Bethanien, Berlin