Searching for the Words that "Feel Right": Resonating with our Bodies and Felt Senses Through Haiku and Large Language Models (LLMs)
Claudia Núñez-Pacheco, Pedro Sanches, Jorge Olivares-Retamal
Abstract
The role of our bodies in our meaning-making has been mostly absent in discussions concerning interactions with LLMs.Acknowledging this gap, this paper explores the use of ChatGPT as a tool for somatic introspection towards finding the words that "feel right" to our bodies and emotions.We document our three-month, firstperson collaborative process using haiku-making and ChatGPT framed around Gendlin's concept of "felt sense" -a type of ineffable bodily awareness that precedes representational meaning.In uncovering the potential of LLMs to support somatic introspection and self-reflection, we contribute two design qualities, which invite designers to consider (1) Ongoing temporalities -that is, interactions in and beyond the screen and (2) Idiolectic resonance, which considers the complexity of our idiosyncratic language expression.In navigating uncertainty, designing for somatic introspection redirects trust towards our bodies, opening for less data-centric ways of designing for reflection.