Project Volano is carrying out cross-disciplinary research to understand how we can value and utilize community perspectives throughout the whole of the AI pipeline. Our first use case is to support historically marginalized communities shape their cultural representation in AI generated media. This raises a host of research challenges from efficient training of AI models on the small amounts of data that people can provide to the development of new methods of evaluation for cultural representation. In each of these questions, we engage with the shifting and specific nature of cultural representation that challenges are common static, one-size-fits all approaches to scaling AI development.
Our focus has been on new methods for community-led processes of meaning-making, data curation & annotation, as well as community-specific evaluation that allow us to develop better models and evaluations. We have developed the Community Library Creator to help communities define their preferred representation and embed that in AI-interpretable data. Based on a Community’s Library, we have developed a Community Scorer, an automated measure for how well an AI-generated image aligns with a community’s preferred representation.

Register your organization’s interest in building a Community Library in the future
Open-sourced Community Libraries
Community Libraries provide training and evaluation resources to improve cultural representation in AI media generation models.