Finalist: Adit Dhanushkodi
ArtCenter College of Design, 2020

Website: aditd.me
Linkedin: https://www.linkedin.com/in/aditdhanushkodi/

The process of machine learning is a project of world-building facilitated by the labeling of data. Once created, these fictional worlds are deployed to make predictions about our current experiences, posing these computationally generated models as universal equivalents to our situated, locally specific experiences and knowledge. Unstable Label is a speculative and participatory civic data labeling application that proposes an alternative approach to building datasets for machine learning: rather than treating the data labels as a fixed set of categories, what if label categories were continually re-labeled, re-contextualized, and re-imagined based on each contributors' own experiences?

What would it look like to use machine learning as a space to contest meaning rather than simply as an operational tool for efficiency and optimization? Unstable Label enables local communities to collectively build an object detection algorithm as a process of negotiating meaning, helping to facilitate conversations of how we each see and imagine the world around us differently. It argues for an approach to machine learning that embraces and designs around situated data rather than striving for the impossibility of objective data.