LLM personalization is becoming increasingly accessible, with little critical inquiry into the extent to which these models are actually capable of representing personal identity. We present two methods for training LLMs on a personal corpus using open-access tools and evaluate the usefulness of such processes for encountering narrative identity.

This paper outlines the use of locative literature (a location-responsive narrative accessed via mobile app) to digitally represent materials from the Stuart Hall Archive. Hall’s work on media serves as conceptual inspiration for both the narrative and interface design. It invites users to navigate the archive through sites of significance.

We analyze associations between fictional characters and emotions in a corpus of fanfiction about Greek myth, using vector space models. We examine the similarity in the VSM between six basic emotions -- sadness, joy, anger, fear, surprise and disgust -- and popular characters, and compare patterns across character-genders and fanfiction-genres.