Uncovering chatgpt’s narrative identity through a psycholinguistic perspective

Abstract

The present study investigates ChatGPT’s self-description and world-description in both present and future contexts.ChatGPT-3.5 was prompted to generate 240 narratives across four conditions: self–present, self–future, world–present, and world–future. The texts were analyzed using the Linguistic Inquiry and Word Count (LIWC) tool to examine writing style and the presence of linguistic markers associated with cognitive and emotional processes, certainty, and social orientation. Repeated-measures ANOVAs were conducted to assess differences across the four conditions. Findings indicate that ChatGPT adopts a distinct narrative posture when describing itself, characterized by the prevalent use of first-person singular pronoun, authenticity, narrative-oriented mode of thinking (in contrast with a more analytical stance towards the world). It engages in deeper introspection and reasoning about its own identity compared to its descriptions of the external world. Unlike humans, ChatGPT displays increased confidence when projecting its identity into the future. Self-descriptions show heightened emotional expression, particularly in future-oriented contexts. Pro-social behaviour and evident social references are more frequent in self-description in the present. The prevalent use of first-person plural pronouns in future self-description suggests a representation of future identity as more interconnected with others. These findings shed light on ChatGPT’s description of itself and the world in two temporal contexts, providing valuable insights into the simulated cognitive and emotional capabilities of Artificial Intelligence.