Searching for Similarity is an installation that continuously searches for images that most closely resemble its current candidate. AI systems often organize information as high-dimensional vectors, where semantic relationships between data points are translated into distances within a latent space: the closer two points are, the more similar they are considered to be. Yet despite appearing as an objective mathematical process, similarity is never neutral. It is shaped by the assumptions, categories, and biases embedded in the curated datasets on which these systems are trained. Search for Similarity seeks to make these hidden processes visible through an endless cycle of searching and matching.

Starting from an initial image, the system searches for the most similar image; whenever a new image exceeds a predefined similarity threshold, it replaces the current candidate and initiates a new search. The evolving sequence, together with the images’ positions within the latent space, is displayed as the search unfolds. Through this continuous process, the work reveals how the meaning of similarity can gradually shift and mutate, prompting us to question what is preserved, what is transformed, and what is ultimately left out when AI is trying to make sense of the world around us.

Exhibited during Dutch Design Week 2026 at MU hybrid art house as part of Slow & Small AI exhibition by Visual Methodologies Collective from October 2026 until January 2027.

System diagram

Installation setup

Three screens in action

Mosaics