Â鶹´å

The lab’s flagship project, AI Style, uses research from literary criticism, narratology, and the digital humanities to bring a humanistic attention to the formal properties of text to AI research and development. The first research paper from the project, "'Written in the Style of': ChatGPT and the Literary Canon," is  in the Harvard Data Science Review. A second paper, "No humans-in-the-loop: The people-less stories generated by GPT" is in Narrative Inquiry. 

Direct discourse, an umbrella category for the parts of fictional text that are the verbatim report of words of entities other than the primary narrator (usually in the form of direct speech, or dialogue), is of interest to many literary critics. However, it is notoriously tricky to detect computationally. This project aims to compare, assess, and develop new computational pipelines for detecting direct discourse, including LLM-based approaches. A paper based on this work has been accepted to and will be published in the proceedings.  

Keeping You Close employs computational analysis alongside close reading to examine reader address in a corpus of over 3,000 nineteenth- and twentieth-century Anglophone novels. To “distant read” these interjections, Kirilloff employs a computational method that separates dialogue from narration and detects address at scale. Research from Keeping You Close has been published in  and is forthcoming in the Willa Cather Review.

Reading Minds uses machine learning to argue that descriptions of characters’ social behavior are central to the history of character realism, because of how they engage readers’ real-life social cognition. The project incorporates research and theory from literary studies, narratology, computer science, psychology and cognitive science to offer a new theory of character realism that takes into account the social psychology of the reader. Research from the Reading Minds project has been published in and is forthcoming in Digital Scholarship in the Humanities.