What might we learn about autobiographical narrative processing from Artificial Intelligence?

Human beings are innately storytellers. Large Language Model AI algorithms (LLMs) like ChatGPT are not. But what ChatGPT produces, when instructed to generate or interpret a narrative, is powerful and can provide insight into human narrative processing, because it is trained to produce narratives from human sources. Our paper explores these ideas. In the first section we review key concepts from narrative identity theory and research, including a recent study from our multi-institution collaboration exploring similarities and differences in human and AI narrative processing. The current study extends this work by exploring people’s perceptions of ChatGPT interpretations of human personal narratives. Across these studies we find that people are adept at differentiating between human and ChatGPT-generated self-defining memory narratives and this may be because ChatGPT relies on redemptive tone and structure, a culturally dominant master narrative, in ways that seemed inauthentic and uncanny to human participants. Discussion focuses on key questions this work raises for research on narrative meaning-making, authenticity through narrative identity, and therapeutic uses of AI.