What Is Hermeneutic AI?

Hermeneutics is the discipline of interpretation: the art and theory of understanding texts, actions, and forms of life. It runs from the exegesis of scripture through philosophers Wilhelm Dilthey and Hans-Georg Gadamer to the interpretive turn in the social sciences — the insistence, as anthropologist Clifford Geertz put it, that the study of culture is “not an experimental science in search of law but an interpretive one in search of meaning.”

Current developments in Artificial intelligence tend to go in a different direction: AI scientists speak in the idioms of prediction, classification, and optimization. Yet large language models are, at bottom, a machine trained on the rich and multi-layered texts of human life.

What perhaps is even less understood is how the creative potential of these models make their linguistic processes of analysis and text-generation more akin the patchy, irregular, fuzzy pace of human reading and writing, than exact robotic calculations designed to satisfy the positivist vision of their creators.

Hermeneutic AI aims to experiment with the mysterious pathways of genAI’s interpretive mind. This nascent field refers to both an object of study and a research program:

• As an object of study, it asks what kind of interpreter a large language model is. What are its prejudices, in Gadamer’s non-pejorative sense — the horizon of pre-understanding it brings to every text? What happens to the hermeneutic circle when one of its participants is a machine?
• As a research program, it asks how these systems can be put to work for interpretive scholarship: reading at scales no individual scholar can, surfacing patterns across archives, drafting the first pass of a coding scheme — while leaving judgment, context, and theoretical imagination where they belong, with the researcher.

The wager of this lab is that the interpretive social sciences should neither ignore these systems nor surrender to them, but engage them the way our disciplines have always engaged new tools for reading and analysis: critically, reflexively, and with close attention to what is gained and lost in translation.

This site will collect our notes and tools as the work unfolds.

DISCLAIMER: In the spirit of open and transparent experimentation, we have crafted this note with the help of Claude Fable 5, guiding it from the start through our main vision of Hermeneutic AI and co-authoring and editing Fable’s suggested outputs. We are solely responsible for the contents of this note.