62 - Getting the Story Wrong: Reading Literature with Machines
Description/Abstract/Artist Statement
Generative AI has become a useful tool for professional copywriters, and composition researchers have understandably begun to study AI’s impact on the composition processes of expert and non-expert writers. But what might the impact of AI be on the reading processes of expert and non-expert readers, particularly in the realm of literature? For the purposes of this study, we developed a method for identifying material over the course of three successive readings of Isabel Allende’s short story, “And of Clay We are Created”-- material that could then be used to make and support claims regarding two questions: “What is the story about?” and “What’s the beginning, middle, and end of this story?” Allende’s story was selected because the work underwent three major revisions, each shifting the perspective from which the story was told. In addition, while “And of Clay” is a “whole” story unto itself, that “whole” is also a part of a larger narrative/collection of stories. We then developed a corpus of readings of this popular story generated by ChatGPT from online study notes and other online materials. We discovered that all of these readings identified the parts of the story in terms of plot, not in terms of the narrator’s action (the telling of the story). Our question: Given that machine learning has greatly enhanced the computer models of writing to grasp nuances of language (for example, humor and idioms), what challenge does “perspective” represent for machine readers?
Faculty Advisor/Mentor
David Metzger
Faculty Advisor/Mentor Department
English
College Affiliation
College of Arts & Letters
Presentation Type
Poster
Disciplines
Computer Engineering | Modern Literature
62 - Getting the Story Wrong: Reading Literature with Machines
Generative AI has become a useful tool for professional copywriters, and composition researchers have understandably begun to study AI’s impact on the composition processes of expert and non-expert writers. But what might the impact of AI be on the reading processes of expert and non-expert readers, particularly in the realm of literature? For the purposes of this study, we developed a method for identifying material over the course of three successive readings of Isabel Allende’s short story, “And of Clay We are Created”-- material that could then be used to make and support claims regarding two questions: “What is the story about?” and “What’s the beginning, middle, and end of this story?” Allende’s story was selected because the work underwent three major revisions, each shifting the perspective from which the story was told. In addition, while “And of Clay” is a “whole” story unto itself, that “whole” is also a part of a larger narrative/collection of stories. We then developed a corpus of readings of this popular story generated by ChatGPT from online study notes and other online materials. We discovered that all of these readings identified the parts of the story in terms of plot, not in terms of the narrator’s action (the telling of the story). Our question: Given that machine learning has greatly enhanced the computer models of writing to grasp nuances of language (for example, humor and idioms), what challenge does “perspective” represent for machine readers?