Training an Artificial Body: A Feminist Dataset
2026 - 2027
Research & Development Project
Supported by Arts Council England National Lottery Project Grant
This is an R&D project exploring how AI can be trained through feminist understandings of emotion, embodiment and lived experience, rather than standardised and extractive data models.
Rather than building a functional AI system, the project develops speculative methods for ‘feminist datasets’ - where voice, breath, gesture, emotional memory, and reflective speech are treated as valid forms of knowledge and creative material. These inputs are used not as training data for machine learning models, but as material for shaping the behaviour, movement, and expressive logic of digital bodies in animation and sound work.
The project responds to a key limitation in current AI systems: emotional and embodied experience is often flattened into measurable signals, reinforcing bias and reducing complexity. In contrast, this research asks how emotional ambiguity, care, consent, and situated knowledge might reshape how digital bodies are constructed and animated.