Human or Artificial Intelligence: Can People Tell the Difference in First-Person Narratives?
Document Type
Article
Source of Publication
Journal of Applied Research in Memory and Cognition
Publication Date
8-1-2024
Abstract
The astonishingly rapid rise of artificial intelligence (AI) raises fundamental questions about human-generated narratives that express personal experiences. First-person narratives that emerge from autobiographical memory are shared frequently as a fundamental form of human activity and play a central role in maintaining relationships, guiding future behavior, and maintaining self-continuity. We generated first-person narratives using prompts with Chat Generative Pre-trained Transformer and tested whether human participants could discriminate AI-generated from human-generated first-person narratives. Participants (N = 101) from Prolific rated five randomly selected narratives as human- or AI-generated and explained their choices. Participants were more accurate than chance (65%) and were more accurate rating human-generated than AI-generated narratives. When participants cited grammar and writing to explain their decisions, they were highly accurate, but when citing emotional expression, they performed at chance levels. Initial results suggest that comparing human- to AI-generated narratives provides insight into how personal narratives might express qualities of an identifiably human self.
DOI Link
ISSN
Publisher
American Psychological Association (APA)
Disciplines
Social and Behavioral Sciences
Keywords
Artificial intelligence, First-person narratives, Human-generated, Chat Generative Pre-trained Transformer, Autobiographical memory
Recommended Citation
Grysman, Azriel; Mansfield, Cade D.; Singer, Jefferson A.; Camia, Christin; Booker, Jordan A.; Bauer, Jack J.; and Fivush, Robyn, "Human or Artificial Intelligence: Can People Tell the Difference in First-Person Narratives?" (2024). All Works. 6649.
https://zuscholars.zu.ac.ae/works/6649
Indexed in Scopus
no
Open Access
no