“Machine Learning-based Approaches in Biometric Data Art” presented by Han


Session Title:

  • Nonhuman Creativity and Agency

Presentation Title:

  • Machine Learning-based Approaches in Biometric Data Art




  • This paper explores issues of bias on biometric data and anxieties about identifications through audiovisual interpretations of the biometric data artworks. As seen in previous historical approaches, many people have been concerned about the reading of race, character, and narratives into genetic traces. Recent history and current trends in biometrics show that biometric data relates to statistical prediction, and bias is an inevitable part of biometrics. The bias has been observed in my previous exhibitions from spectators who experienced their biometric-driven audiovisual outcome. The artist created an experimental version of the two biometric data artworks using machine learning (ML) methods of artificial intelligence (AI) to investigate the bias on biometrics. The AI-driven interface analyzes the input data, improves predictions, and extracts visual features based on sample data. Two biometric data (iris and finger-print) are used in the artworks, and an informal user study is dis-cussed. This project investigates the possible artistic approaches in using biometric data and attempts to find unbiased solutions for biometric data interpretations.