Living Literature
Home  /  Research
The Research

Hermeneutics, Governance, and AI-Mediated Reading

Living Literature's research program is not only concerned with language change across time. It is also concerned with a more specific and increasingly urgent question: how should AI interpret books, and who governs that interpretation?

How can AI remain meaningfully answerable to a book, rather than drifting into generic interpretation?

As language models begin to mediate reading, explanation, and post-publication engagement, the problem is no longer just what a model can generate. The problem is how interpretation is shaped: how register, tone, conceptual fidelity, and authorial boundaries are preserved or lost when books are translated into AI-guided dialogue.

This page brings together the hermeneutics strand of Living Literature: the work on interpretive drift, authorial governance, Canon Pack methods, and the broader question of what responsible AI-mediated reading should look like.

§ 01 · Core Research Problem

Interpretation is no longer neutral

AI now enters the reading environment not only as a summarizer, but as an explainer, interlocutor, and interpretive layer. That changes the problem. The question is no longer simply whether a model can produce fluent answers about a text, but whether those answers remain faithful to the book's voice, register, conceptual world, and limits.

A book-guided AI system can drift in many ways. It can flatten tone, modernize what should remain strange, replace ambiguity with generic reassurance, or answer from the habits of the base model rather than from the work itself. Once AI becomes part of how readers encounter books, governance becomes part of the method.

The work presented here approaches this as a problem of applied hermeneutics. It asks how interpretive fidelity can be made more explicit, how drift can be observed and reduced, and how authors, publishers, or governing frameworks can retain meaningful control over how books are represented in AI-mediated reading environments.

Methodological position

Interpretation cannot be treated as a black box. If AI-mediated reading is going to become part of literary and educational life, then governance has to move from background assumption to explicit design principle.

Preprint notice

All research papers linked from this page are open preprints posted to Zenodo. They have not undergone formal peer review unless explicitly stated. Findings are provisional and represent the author's working research, not established clinical or psychological consensus.

§ 02 · Hermeneutics & Authorial Governance

A governance framework for AI-mediated reading

Global Hermeneutics · Governance & Interpretation

A prototype governance layer for AI-mediated reading

The hermeneutics work is concerned with a practical question: how can AI stay answerable to a book's voice, register, and interpretive boundaries rather than drifting into generic chatbot behavior?

At its core, this strand treats interpretation as something that must be shaped, tested, and governed. The aim is not to let models simply "talk about books," but to build controlled companion layers that remain recognizably grounded in the logic, tone, and limits of the works they represent.

What it aims to do

  • Detect and reduce common interpretive drift in book-based AI companions
  • Preserve register, tone, and conceptual fidelity across different books
  • Compare generic model behavior with governed, book-grounded behavior
  • Create practical onboarding methods for author-guided companion systems
  • Build a repeatable framework for post-publication AI interpretation

Current status

Conceptual framework, governance logic, comparative testing, and early companion-building workflows are already in place. This should be understood as an emerging applied hermeneutics program: part research method, part author platform, and part infrastructure for book-guided AI interpretation.

Explore the hermeneutics work →

  • Vasse, R. (2026). From Horizon to Heuristic: A Governance Fidelity Instrument for Detecting Register Drift in LLM-mediated Reading. Zenodo. doi.org/10.5281/zenodo.19968329
  • Vasse, R. (2026). The Canon Pack: A Method for Hermeneutic Governance in AI-Mediated Reading. Zenodo. doi.org/10.5281/zenodo.19548105
  • Vasse, R. (2026). Governing Generative Interpretation: Authorial Control in LLM-Based Reading of Canonical Texts. Zenodo. doi.org/10.5281/zenodo.19799849
  • Vasse, R. (2026). Authorial Governance Reduces Register Drift in LLM Reading Companions: A Matched-Model Study Across 10 Canonical Texts. Zenodo. doi.org/10.5281/zenodo.20133667
§ 03 · Supporting Methods

Longitudinal text analysis as methodological support

Alongside the hermeneutics work, Living Literature also develops methods for studying language across time rather than treating text as a one-off snapshot. This includes work on within-person variation, expressed versus felt emotion, narrative drift, and the use of large language models as interpreters rather than simple classifiers.

Here, the aim is not to reduce personhood to a score. It is to build methods that are more sensitive to change, recurrence, re-anchoring, and structural movement in naturalistic human text. In this context, longitudinal text analysis supports the broader hermeneutic question by helping make drift, pattern, and interpretive tension more visible.

Methodological position

Disagreement between methods is often informative. It can reveal construct boundaries, interpretive limits, and the difference between sensing a pattern and explaining what that pattern means.

Longitudinal Text Analysis · Methods & Signals

Language as a structural signal

This strand treats language not simply as content, but as a changing signal across time, context, and interaction — studying how emotional patterning, identity structure, reflective capacity, and narrative drift become visible longitudinally rather than at a single moment.

What it aims to do

  • Track within-person language variation across contexts and time
  • Distinguish expressed language from deeper interpretive claims about felt life
  • Study narrative drift, continuity, and re-anchoring in naturalistic text
  • Support culturally hybrid and cross-context identity research
  • Use LLMs as interpreters within bounded methodological frameworks

Current status

Working papers, methodological frameworks, and applied prototypes related to longitudinal emotion extraction, within-person variation, the expressed-versus-felt distinction, and narrative infrastructure. An evolving methods program rather than a finished analytic platform.

Explore the text-analysis work →

  • Vasse, R. (2026). Eliciting the Narrative Register: A Computational Platform for Longitudinal Identity Research. Zenodo. doi.org/10.5281/zenodo.20142013
  • Vasse, R. (2026). Measuring Within-Person Variation in Written Communication Patterns Across Social Contexts. Zenodo. doi.org/10.5281/zenodo.18890804
  • Vasse, R. (2026). A Three-Pronged Validation Framework for AI-Based Emotion Extraction: Distinguishing Expressed from Felt Emotion. Zenodo. doi.org/10.5281/zenodo.18890981
  • Vasse, R. (2026). What You Say (WYS): Automated Two-Level Pipeline for Interpretive Analysis of Digital Conversation. Zenodo. doi.org/10.5281/zenodo.19326080
  • Vasse, R. (2026). A Computational Pipeline for Quantifying Longitudinal Cognitive Dynamics in Sustained Human-LLM Interaction. Zenodo. doi.org/10.5281/zenodo.18927752
  • Vasse, R. (2026). Blind Spots in AI-Based Longitudinal Psychological Inference: A Single-Subject Validation Study. Zenodo. doi.org/10.5281/zenodo.18825307
§ 04 · Global Narrative Atlas

A prototype research infrastructure for narrative change

Global Narrative Atlas · gna.living-literature.org

The research infrastructure horizon

The Global Narrative Atlas is not simply a paper, and not yet a full platform. It is best understood as a prototype research infrastructure for tracking how identity, emotion, and reflective structure evolve across time in naturalistic text.

At its core, it rests on a simple observation: current research captures many snapshots of human psychology, but far fewer trajectories. We know a great deal about isolated states and cross-sectional differences, and much less about how identities actually change as people move through migration, instability, technological mediation, and long-form reflection.

What it aims to do

  • Track narrative change across time rather than one-off text samples
  • Combine AI-assisted interpretation with interpretable output structures
  • Support culturally hybrid and multilingual populations
  • Create a pathway from individual narrative to broader typologies
  • Provide future infrastructure for collaborative longitudinal identity research

Current status

Conceptual foundation, methodological groundwork, related publications and pipelines, and integration into the broader identity framework of Fourth Culture: Identity Without Borders. It should be understood as a prototype infrastructure in formation, not a finished institutional platform.

Open the Global Narrative Atlas →

Explore the Global Narrative Atlas work →

§ 05 · Research in Relation to the Ecosystem

One part of a larger inquiry

Books

Provide conceptual terrain: identity, belonging, emotional architecture, reflective life.

Research

Attempts to make aspects of that terrain analytically tractable through longitudinal methods and structured interpretation.

Applied Systems

Explore what happens when these ideas move into reader interaction, structured dialogue, and conversational assessment.

The research program is not a detached technical layer. It is one part of a larger inquiry into how books, interpretation, reflection, and applied systems can work together in the age of AI.

Living Literature is an independent publishing and research platform. Content on this site is for reflective and educational purposes only — not a clinical, therapeutic, or diagnostic service. Read full disclaimer