Gallery of Emotions

The photo-collection 'Coding Happiness' is a commentary on our increasing reliance on machines in areas of life which were once considered the exclusive domain of human sensibility, intuition, and creativity. Using personal photographic memorabilia, the artist attempts to reverse-engineer the process of translation of human affect into straightforward data units that can be fed into computer algorithms. The exhibited photographs have been analysed by an online mood assessment tool and present the audience with an opportunity to see the verdict of the algorithm in the context of their own reading of the photographed moment.
The tool used for the reading of the photographs is one of many emerging applications that apply biometrics research and machine learning to automate emotion detection in online content such as photographs and videos. Most of them base their analysis on the premise of seven basic emotions: happy, sad, angry, surprised, scared, disgusted, and neutral (P. Ekman, 1992) which are matched to generalized cues for facial expression. The percentiles in the returned evaluations correspond to the confidence of the algorithm of the presence of a given attribute. The photographs shown are thus grouped according to the detected emotion and lined up in an increasing order of algorithmic confidence. The collection of analysed facial expressions is large enough to create a sense of the embedded criteria for emotion recognition, but also question the amusing incongruities between images and their readings.

While it is known that neural networks or AI are dependent on human feedback loops for training, their self-correction processes are virtually untraceable. However, before we ask whether mood assessment algorithms can be trusted, we should probably ask: why do we need a machine to tell us what we feel?