BodyBackground
Menu

Projects

Vera van der Burg

Still Life

‘Still Life’ is a research project exploring Machine Learning algorithms becoming explicitly subjective. In this project, Vera van der Burg programmed an object recognition algorithm that was trained on multiple ‘subjective’ datasets. These subjective datasets were named for example ‘love’ or ‘jealousy’, and were filled with images that, according to her own judgment, would visualize these emotional states. This training session resulted in the emerging of an ultra subjective object recognition algorithm. Then, she let this algorithm scan a still life filled with ambiguous objects, and let them interpret their emotional content.
In the research, Vera questioned how something subjective as emotions, the human gaze and object recognition could be placed on an objective, mathematical scale, as machine learning algorithms appear to be. In the past couple of decades, attempts have been made to replace certain human tasks with automated systems.
Sometimes, this seems successful. Self-driving cars ride our roads, criminal behavior is being detected, our musical taste predicted. All based on Machine Learning systems coded by programmers, trained to learn from patterns in large datasets. However, when these systems operate in the real world, Machine Learning Algorithms show to be biased. When Vera asked herself why this automated bias occurs, the answer was never ‘because the algorithm said so’.
It was never the algorithm on its own that was the culprit. It was us, feeding it with biased data and ideas in the first place. Self-learning algorithms are just a reflection of whoever programmed it, and fed it. Therefore, this object-recognition algorithm became a reflection of Vera.

The project resulted in the visualization of a thought experiment in which Vera tried to rethink what we could want of our machines, and automated her emotional experience of objects in which she focussed explicitly on her own subjectivity. What would happen if she entered the data of her soul into a machine? How can you create datasets that contain emotions? In the end, the research resulted in the exhibition of a dialogue with herself,
mediated by an algorithm, embodied in an installation of three still lifes, being analyzed by her ultra subjective, emotional algorithm.