This door requires you to express happiness through your smile in order for it to open.

In today’s world, facial recognition technology is becoming increasingly pervasive, allowing our built environment to not only recognise our presence, but also identify our emotional state and other personal attributes such as gender, age and ethnicity.

Doors, symbolic features of building control, typically have two binary states: “open” or “closed”. Similarly, this particular door classifies the visitor’s emotional state, specifically whether they are “happy” or “not happy”, before allowing them entry.

This artwork creates a scenario whereby emotions are dictated and judged by a non-human entity. Each visitor’s smile is evaluated and given a score based on its conformity to a training dataset, and the door only opens upon achieving the minimum score. This aspect of the artwork critiques dataset biases within machine learning, whereby unrepresentative or poor datasets often leads to discriminatory results.

The inspiration for this project stemmed from the artist’s personal experience at an automated photo-taking booth at a UK visa application centre, where his picture was rejected multiple times due to its “non-conformity”.

The AI perception system of this installation was constructed using a convolutional neural network (CNN) algorithm and trained with the popular FER-2013 dataset for facial expressions. FER-2013 comprises approximately 30,000 faces. Despite the dataset’s size, there were occasional incidents during the three-month exhibition where visitors of certain ethnicities encountered difficulties having their smiles recognized. Fortunately, I was present during some of these moments and took on the responsibility of explaining why these challenges occurred. Engaging in extensive conversations, I discussed the implications and how we should interpret such occurrences.

The installation was made possible with the support of Studio Egret West and was part of their exhibition “co-creAIte: AI everywhere all at once,” which premiered during the 2023 London Festival of Architecture.

Photographed by Jarrell Goh