This research investigated diffractive practice as art-based method to prototype machine learning materials.
Diffraction—a concept introduced by queer philosopher and physicist Karen Barad—intends to displace reflection, which assumes pre-existing subjects and objects interacting with each others, as a dominant model of inquiry. It does so by assuming that humans and non-humans are bound together within complex socio-material practices, which are fluid and ever evolving. We were interested in exploring the potential of diffractive practices to prototype machine learning in ways that resist normative framings of computation. To do so, we adopted a crafting approach, using machine learning as material in several art and design projects—specifically, in somasticks, and The Appprentices.
Our work let us identify five socio-technical conditions for art-based machine learning prototypes to produce constructive interferences between design and engineering practices: situational whole, small data, shallow model, learnable algorithm, and somaesthetic behaviour. It also enabled to sketch what we named intra-active machine learning—a process of becoming-with others, assembling disciplinary concepts, techniques, and practices of machine learning, as well as diverse roles of artists, designers, and engineers, into one transformative, fluid and ever evolving socio-material phenomenon.