Applicants:
- Elena Beretta, Computer Science/UCDS group, Faculty of Science
- Pia Sommerauer, Computational Linguistics and Text Mining Lab (CLTL), Faculty of Humanities
- Other team members: the project will also involve Cristina Voto, Connected World Fellowship Program Winner.
Bertillon and Galton leveraged biometric traits to identify moral traits, exacerbating prejudice toward marginalized communities. Despite such studies being largely outdated, this phenomenon still exists in a different form. Computer vision systems, for instance face recognition, rely on physiognomic features, only taking into account biological sex. Models are trained with binary gender features regardless of the wide range of gender spectrum, and recognition provides a prediction based on the visible artifacts. At the same time, people are attempting to make many areas of society more inclusive for people who do not identify with binary gender identities. For example, efforts have been made to structure and explain inclusive labels for different gender identities, e.g. The Homosaurus vocabulary. In this project, we aim to use the conceptual system captured by structured ontologies and vocabularies to study how face recognition systems could make more inclusive predictions.
