Animistic interactions in conversation with LLM-powered robots

Abstract

This thesis examined how large language models (LLMs) can be embodied into physical prototypes to evoke animistic interactions. Using a Research through Design (RtD) approach, I created and tested several prototypes that linked LLM outputs to robotic movements. One robot responded with poetry, another tracked objects, and others danced or recognized themselves in images. Each prototype combined movement and personality to give the impression of animism. The experiments showed that even basic robotic gestures became expressive when paired with language. People often projected personalities onto the robots, interpreting their movements as intentional. Interactions unfolded through co-creation and negotiation. Technical limitations opened up for playful interpretations of rebellious or autonomous behaviors. In this design space, movement proved to be a powerful tool: both failure and ambiguity invited users to interpret and engage.