We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.AI

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Computational Lens on Cognition: Study Of Autobiographical Versus Imagined Stories With Large-Scale Language Models

Abstract: Lifelong experiences and learned knowledge lead to shared expectations about how common situations tend to unfold. Such knowledge enables people to interpret story narratives and identify salient events effortlessly. We study differences in the narrative flow of events in autobiographical versus imagined stories using GPT-3, one of the largest neural language models created to date. The diary-like stories were written by crowdworkers about either a recently experienced event or an imagined event on the same topic. To analyze the narrative flow of events of these stories, we measured sentence *sequentiality*, which compares the probability of a sentence with and without its preceding story context. We found that imagined stories have higher sequentiality than autobiographical stories, and that the sequentiality of autobiographical stories is higher when they are retold than when freshly recalled. Through an annotation of events in story sentences, we found that the story types contain similar proportions of major salient events, but that the autobiographical stories are denser in factual minor events. Furthermore, in comparison to imagined stories, autobiographical stories contain more concrete words and words related to the first person, cognitive processes, time, space, numbers, social words, and core drives and needs. Our findings highlight the opportunity to investigate memory and cognition with large-scale statistical language models.
Comments: Equal contribution from Sap and Jafarpour; in review
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Cite as: arXiv:2201.02662 [cs.CL]
  (or arXiv:2201.02662v1 [cs.CL] for this version)

Submission history

From: Maarten Sap [view email]
[v1] Fri, 7 Jan 2022 20:10:47 GMT (1350kb,D)

Link back to: arXiv, form interface, contact.