In collaboration with Colins McSwiggen, this project introduces a large language model (LLM) whose understanding of the world is based on differences between concepts rather than on an identification between them. We call this an empathy machine for it challenges the paradox that confines traditional empathy: the tension between difference and similarity. (And their constitutive paradoxes: self/other, interior/exterior, subject/object, mind/body, visible/invisible, life/death, light/dark, masculine/feminine…etc) Studies have shown that empathy is more fluid when the other resembles ourselves, and more solid when they appear different.

Our approach stems from a critical observation: empathy often fosters a tribal mentality, as noted by psychologist Paul Bloom and further theorized by Saidiya Hartman, who suggests that this limit of empathy is a feature and not a bug. Most people can only empathize with others that they can imagine themselves into. In order to empathize, one displaces the other to feel their pain in a form that is legible to oneself. Unsurprisingly, this legibility hinges upon sameness. Hartman critiques that empathy requires the negation and obliteration of the other into a phantasy of ourselves.

At the heart of this project is the desire to hack into the procedures and categories that lead to a proprietary relationship with the earth and each other. LLM represents words as vectors, spatializing language. Our work performs geometric transformations on this latent space, constructing a machine that speaks from an alternative geometry of relationality. With this attempt we offer an opening to the artificialisation of language (hidden behind the veil of representational opacity)  while proposing a poethics of empathy(an irreducible space of misordered  relations).

© 2023 HSURAE